# Borrowed Trust *Open-web native, as the recommendation and in-feed networks sell it, is a supply-rights business sold as a performance product. Its biggest sellers are expanding into display, video and AI answers. Several studies found many readers do not recognise native units as ads, and no public study with published design and results shows what native adds to sales.* Evgeny Popov · No Fluff Advisory · Research paper · September 2026 · Research cutoff 27 September 2026 ## Abstract *What native advertising is in 2026, who earns from it, and how much weight the evidence behind its sales pitch will bear. Research current to 27 September 2026, with the United States as the main market.* > **In short.** > Open-web native, as the recommendation and in-feed networks sell it, is sold as a performance product. But it is a supply-rights business. The larger listed platform, Taboola, depends more on two supply partners than on any advertiser. Both large listed platforms lock supply with long contracts and guarantees, and both are building products beyond the widget. Several studies found many readers do not know native units are ads. In the sources screened by the cutoff, this review found no public study with enough published design and results to estimate what open-web native adds to sales. The vendors' own evidence uses designs that can overstate lift. On the Native Quotient, only Taboola reaches the front rank. No vendor scores above 3 on measurement. ### The question Native advertising is sold as performance. This paper asks what the public record shows behind that pitch. How big is the market, and by whose fence? Who controls it? What does a native dollar pay for? Do readers know they are looking at an ad? Has anyone shown that native causes sales? And what are agents and AI answers changing? The scope is the open web. Social feeds, search, retail media and AI assistants are treated as adjacent systems that compete for the same budget. ### How it was done The work ran as twelve research workstreams, a vendor-universe sweep and nine vendor profiles. An AI research agent ran each one from a shared brief. A second AI agent, the verifier, then checked it, with orders to refute it. In all, 977 sources were opened and logged. The research ledger holds 642 claims, and each number in the text traces to a source. The paper scores 24 vendors on 12 dimensions. The scoring rules were written before the first scoring pass. A revision followed once the verifier records were read, and the Native Quotient summary was added after scoring, at the author's request. A separate review of an interim draft raised one hundred findings, and a second outside review of the rebuilt package raised 38 more. Each was checked against its cited source where one could be reached; one was rejected. Each change is in the corrections ledger in chapter 12. This is AI-assisted research with a published process. It has not been audited or peer reviewed. No vendor saw it before it was published. ### What it found 1. **The market has no measured size.** eMarketer puts US native display at $147.98 billion in 2026. That figure counts social feeds. They took 74% to 84% of it in each vintage whose split could be read. The industry's own revenue accounts have no native line at all. The two large US-listed open-web platforms reported $3.21 billion of revenue in 2025, about 1.6% of Meta's ad revenue. 2. **It is a supply-rights business.** Taboola's ten largest advertisers are under 10% of its revenue. Yahoo and Microsoft make up about 34% of its revenue from advertisers on publisher properties [1]. Teads Holding Co. discloses no partner at 10% of its supply cost, so this finding rests on Taboola's record. Traffic acquisition cost is its recorded cost of supply, mostly payments to publishers. Its cost of guarantees to publishers was about 18% of that cost in 2024 and 15% in 2025, up from 9% to 10% in 2021 and 2022. 3. **Many readers do not recognise it.** In three studies of sponsored articles and in-feed ads, 7% to 37% of those tested saw them as advertising. In one of those studies, 81% did so for standard ads. In the FTC's lab, about 47% did, and 68% with its recommended labels. Only the FTC study tested a recommendation widget. Better labels helped but did not close the gap. These are dated results for certain formats and samples. They are not a 2026 population rate. 4. **Conversion optimisation does not change the billing basis.** A conversion is the action an ad is meant to drive, such as a sale. In the per-click campaigns the Taboola, Outbrain and MGID help centres document, the platform's model decides which clicks to buy and the advertiser still pays for every click. The financial risk stays with the advertiser. A target cost per acquisition is a bidding setting, not per-acquisition billing. Both large platforms' filings describe per-click, per-thousand and per-acquisition pricing, and Teads prices some campaigns on incremental action. Neither discloses the share of revenue under each. 5. **The causal evidence is missing, not negative.** In the sources screened by the cutoff, this review found no public study with enough published design and results to estimate what open-web recommendation and in-feed units add to sales. Vendor evidence compares those who saw the ads with those who did not. In Facebook conversion experiments, observational designs like that overstated lift by about three to thirteen times at the median. This review found no published estimate of that bias for native brand-lift surveys. Randomised tests of native formats elsewhere, in Facebook campaigns and in native search ads, found positive effects. The gap is specific to the open-web widget. 6. **The platforms are building beyond the widget.** Taboola sells display, video and ads inside AI answers through Realize and DeeperDive, and Teads Holding Co.'s connected TV reached 13% of revenue in Q2 2026. Neither discloses how much growth those lines add. In that quarter, revenue billed through Yahoo grew faster than Taboola's total. Several publisher traffic and ad-volume indicators are falling. They measure sessions, ad requests and impressions, not one page-view series, and a shrinking supply of widget slots is an inference from them. 7. **Buyers can inspect less than they can buy.** On both large native self-serve platforms, site lists are documented only after delivery. Controls are exclusion-first. A regional platform, Readpeak, shows a pre-campaign list can be done. 8. **Agents reach native through the platforms' own doors.** Neither agentic protocol stack lists a large native platform as an implementer on the member pages opened. Taboola instead runs its own server for AI agents to manage campaigns. ### What would change these findings A holdout study of native on its own, with published design and results. A holdout keeps the ads from one group, so the gap shows what the ads caused. A platform that discloses the actual lag behind its reported conversions. A site list shown before the campaign at a large platform. Recommendation revenue reported on its own and shown to be growing. Chapter 11 gives each scenario a time window to watch and the evidence that would count against it. ### Limits The open-web native market has no measured size, and this paper does not invent one. Its economics are those of the recommendation and in-feed networks; sponsored editorial sold directly by publisher studios, creator and influencer content, and in-mail native were not measured. The buyer guidance is written for the United States, while the platforms earn most of their revenue elsewhere; Asia-Pacific markets and their rules were not covered. The financial comparisons rest on measures the companies define. The recognition studies are old. The vendor scores come from one AI-assisted scoring pass, with no second scorer. A company can score low for being private rather than weak. Access limits are logged for each source, and chapter 12 names the ones that matter. ### How to read it Chapters 1 and 2 define and size the market. Chapters 3 and 4 follow the money and the supply contracts. Chapters 5 to 7 test the evidence on readers, delivery and performance. Chapter 8 covers commerce and AI answers. Chapter 9 scores the vendors. Chapter 10 covers enterprise buying and the rules. Chapter 11 sets out scenarios with signs to watch. Chapter 12 is the method and the ledgers, and the appendix lists each download. Each chapter opens with a short summary. In the web edition, the Skim control cuts the paper down to those summaries, the figures and the instruments. ### Interests to declare The author has worked at Samba TV since May 2025. Samba TV is a TV measurement and ad data company. Before that he worked at Verve Group, where he ran the Moments.AI business. Earlier he worked at Hearts & Science and Lotame. He co-leads the Signals & Measurement working group of AgenticAdvertising.org. That body governs AdCP, a protocol discussed in chapter 11. Samba TV is listed among AdCP's members, and Teads names Samba TV among its measurement-partner integrations; Teads' connected-TV business, discussed in chapters 4 and 11, overlaps Samba TV's market. None of the author's current or former employers listed above is scored. Teads is scored and profiled; its commercial relationship with Samba TV is disclosed here. No one funded or commissioned the paper, and no vendor reviewed it. ## 1. Defining the market *"Native" names how an ad looks and where it sits. Arguments over whether the category is growing, working or worth buying almost all turn on one point. That point is which other axes are silently included.* > **In short.** > This paper splits any native placement into six axes. They are format, placement surface, targeting method, buying route, optimisation objective and commercial model. Native describes the first two. Contextual targeting is one value on the third. So native and contextual overlap, but they are not the same category. Three scope groups keep the counting honest. The core market is paid native placements and the firms that sell or optimise them. Adjacent systems, such as social feeds, search, retail media and DSPs, compete for the same budget. Emerging surfaces include sponsored answers inside AI assistants. They are tracked on their own as live, trial or announced. Only paid promotion counts. Organic recommendations and editorial links do not count as spend. And market size, addressable opportunity and a company's own revenue guidance are three different numbers. Each source this paper opened defines native advertising the same way in one sentence. In the tables, each one differs. That one sentence comes from the FTC. Native is content that bears a similarity to the news, feature articles, product reviews, entertainment and other material that surrounds it online [2]. IAB's Native Advertising Playbook 2.0 was published in May 2019. As far as this research found, nothing has replaced it. It sorts that into three types: in-feed and in-content units, content recommendation units, and branded or native content. It requires a disclosure on all three [3]. OpenRTB Native Ads 1.2 is the programmatic standard that carries native in the bid stream. It defines native as a set of components: a title, an image, a description and a link. The publisher renders them in its own style [4]. Those three are not the same. The FTC's definition turns on resemblance, IAB's on format and OpenRTB's on assets. An editorially produced sponsored article need not be built from ad components at all. This paper's definition fits all three. Native describes advertising designed to fit its surrounding content or experience. Programmatic native commonly achieves that through publisher-rendered assets. On all else, the sources disagree or say nothing. That is why the paper needs six axes, not one word. ![Native is two axes of a six-axis buy.](figures/fig-category-map.svg) *Figure fig-category-map. **Native is two axes of a six-axis buy.** A schematic of the six choices behind any native placement; a campaign can combine several placements and so several values on an axis. "Native" describes format and placement surface; contextual is one option on the targeting axis; the optimisation objective and commercial model are where the category has actually moved. Reported growth in "native" can come from any axis changing, which is why the spend chapter separates them. Units: none (schematic). Scope: definitions used throughout this paper. Sources [4], [1]. Caveat: Schematic, not a measurement. Axis values are the ones observed in the sources; the list is not exhaustive.* ### Six axes, and which two the word covers The format axis is what the unit looks like. It can be a recommendation unit, a block of suggested links at the end of an article. It can be an in-feed unit between posts. It can also be a sponsored article, a native video player or a product card. The placement surface is where the ad appears. That can be an open-web publisher page, a social feed or a search results page. It can also be a retailer's site, an AI assistant's answer or a newsletter. The word "native" describes those two axes and nothing else. The targeting method is why a given reader sees the ad. It can rest on contextual signals from the page, or on audience or behavioural data. It can use retargeting from a past visit or lookalike modelling. Or it can rest on nothing at all. The buying route is how the ad is bought. It can be a self-serve platform or a managed service under an insertion order. It can be programmatic through a DSP. Since 2025, it can also be an agent that talks to a seller's agent. The optimisation objective is what the delivery system chases. That can be impressions, clicks, conversions, return on spend or an incremental outcome. A conversion is an action the advertiser wants, such as a sale. An incremental outcome is one that would not have happened without the ad. Last, the commercial model is who pays for what. It can be cost per click, cost per thousand or cost per action. It can also be a revenue share to the publisher, a minimum guarantee or a software subscription. The axes apply to a placement or a line item. A campaign can mix formats and surfaces and use more than one route. So a campaign can carry more than one value on an axis. A single placement usually takes one value per axis, though some axes, targeting in particular, can hold more than one. Take a recommendation unit on a news site, bought self-serve and paid per click. It has a conversion objective and no targeting beyond the platform's own model. It is a native ad. So is a sponsored product card in a social feed. It is bought programmatically, paid per thousand impressions and uses audience targeting. Both are native in the broad sense, but they take different values on every axis the example names, format and surface included. A market estimate that adds them together is adding a click business to an impression business. ### Native is not contextual, and the confusion has a cause The sources most often confuse native with contextual targeting. The cause is historical. The recommendation platforms that dominate open-web native rank ads against the page and the reader. So their delivery is contextual in the loose sense. And vendors that sell contextual targeting as a data product often deliver it into native units. But the two sit on different axes. Contextual targeting is a way of choosing the reader. It can apply to a display banner, a video ad or a native card. It describes the basis for the choice, not a promise that no personal data is used. A system can read the page and also use a device ID. Native is a way of dressing the ad. It can be targeted by context, by behaviour or not at all. This series has a companion paper on contextual advertising. It treats contextual as a targeting method. Its transaction layer is a taxonomy ID [5]. This paper treats native as a format and placement. Its transaction layer is a component list. As described in chapter 6, the recommendation engine is where the two meet. It uses page context as one input among several to a click-and-conversion model. That engine is not contextual advertising in the data-product sense. It is a ranking system that happens to read the page. ### Three scope groups, one counting rule ![Three scope groups, one counting rule.](figures/fig-scope-groups.svg) *Figure fig-scope-groups. **Three scope groups, one counting rule.** The core market is paid native placements and the businesses that sell or optimise them. Adjacent systems compete for the same budget and are compared, not counted. Emerging surfaces are tracked separately as live, trial or announced. Paid promotion only; organic recommendation and editorial links never count. Units: none (schematic). Scope: definitions. Sources [4]. Caveat: Schematic. Boundary cases (social in-feed, retail-media offsite, sponsored answers) are discussed in chapter 1.* The core market is paid native placements. It also takes in the platforms or publisher businesses that sell, deliver or optimise them. It covers recommendation units, in-feed native on open-web pages and sponsored editorial. It also covers native video or commerce units where they are sold as native. Chapters 3 and 4 measure this market through filings. Chapter 9 scores it. Budget substitutes and adjacent systems compete for the same buyer's money. Or they supply something the core depends on. They are social feeds, search, retail media, DSPs, content commerce and affiliate networks. A social in-feed unit is native by format, which is why some estimates count it. Here it sits in the adjacent group. That is because its buyer, auction, measurement and rules belong to the platform. They do not belong to the open-web native market. When a figure in this paper includes a social or retail-media number, the paper says so. It also says what share is native. Emerging surfaces are places where the native definition may apply. But the market there cannot yet be measured. They are mainly sponsored answers or recommendations inside AI assistants and AI answer engines on publisher sites. Buying through agents is a buying route, not a surface. It sits on that axis. Chapter 8 sorts the surfaces into live products, trials and announcements, and keeps them apart. Chapter 11 returns to what they and the agents change. The counting rule across all three is paid promotion only. So this paper does not count an organic recommendation from the publisher's own engine. Nor does it count an editorial link, or a product listing a retailer ranks by relevance. It also leaves out an affiliate link the publisher chose to insert. But that exclusion is about counting, not about disclosure. An affiliate link carries a material commercial connection. Chapter 8 sets out the disclosure rules that apply to it. Where a boundary is contested, the paper says which side it took and why. The three cases that matter most are set out below. Four more rulings are brief. Creator and influencer sponsored posts, and podcast host-reads, are native by resemblance but sold by creators and platforms; they are adjacent here and not sized. Connected-TV home-screen tiles and pause ads are adjacent formats of the TV interface. In-app native and newsletter or in-mail native are core surfaces, but the paper documents them only in passing. > **Three contested boundaries, and where this paper draws them.** > Social in-feed ads. They are native by format, and some market estimates count them. They are excluded from the core here because the platform owns the auction, the measurement and the rules. Chapter 2 shows what happens to the market size when they are included. > > Retail media offsite. A retailer's sponsored product, served on a publisher page through the retailer's or a partner's platform. It is native by format when it shows as a product card. It is adjacent here because the buyer is a retail-media buyer, measured on the retailer's sales data. Chapter 8 sets out the routes. > > Sponsored answers in AI assistants. This is the purest version of the definition: an ad that shares the format of the content. It is also the least measured. It counts as emerging here. Chapter 8 records several live products, one retirement and one trial. Research on whether users recognise these ads is thin and not representative. ### Three numbers that are not the same number Market size is what was spent on paid native placements in a given period and place. Addressable opportunity is what could be spent if each budget that could move into native did so. What one company expects to win is a third question. The closest public answer is its own revenue guidance. That is a management forecast, not an estimate of what can be captured. The playbook that seeded this paper uses all three. So do the vendor filings in chapter 4 and the forecasts in chapter 2. Sometimes all three appear in one sentence. Take Taboola's 2025 annual report. It claims a $55 billion opportunity for its performance platform. It bases the figure on internal and external industry data [1]. That is an addressable claim, not a market size. And it covers performance advertising broadly, not native. The paper keeps market size, addressable opportunity and company guidance apart. Each time it uses a number, it names which of the three it is. ### How the definition changes the answer Chapter 2 makes the result concrete. The same word gives estimates that differ by an order of magnitude. The gap depends on whether social in-feed is inside the fence. Chapter 3 shows that the firms called native platforms now sell more and more placements that are not native at all. Chapter 5 looks at whether readers can tell the unit from the content. That is the property that defines the format. In several studies of certain formats and disclosures, their ability to do so was low. Chapter 6 shows that the platforms now compete on the optimisation objective, not the format. And chapter 11 shows that the two listed native companies are moving along different axes at once. One is moving toward display and connected TV. The other is being pulled by the page-view economics of the format it is named after. None of those findings shows up if native is treated as one thing. ## 2. Follow the spend *In the sources this project reviewed, only one public spend estimate carries the word "native". In the past, most of it was social feeds. Everything else has to be rebuilt from the platforms' own accounts.* > **In short.** > EMARKETER puts US "native display" at $147.98 billion for 2026, up 13.1%. That implies about $130.8 billion for 2025. Its vintage splits show it counts social in-feed; an AI-assisted EMARKETER FAQ also lists outstream and rewarded video, but the forecast's own scope is not published. Social took 74–84% of the total in the vintages, or forecast editions, that published a split. The last of them was in 2019. The industry's own revenue accounts, from IAB/PwC, have no native line. Native sits inside display, which was $81.6 billion in 2025. Display grew 9.8%, the slowest of the four major named formats. Social, at $117.7 billion and up 32.6%, is a channel that cuts across formats. The eMarketer native figure is about 1.6 times all of IAB's display. That tells you the two fences sit in different places. It does not mean either is wrong. The two large US-listed open-web native platforms reported about $3.21 billion of revenue in 2025. Their ex-TAC gross profit was $1.24 billion. That is their own non-GAAP measure, broadly what is left after they pay publishers. The two figures are roughly 1.6% and 0.6% of Meta's ad revenue. Google's open-web publisher line fell 1.9%. Its search line grew 13.4%. The research brief asked how much reported native growth comes from a wider category, from adding social, or from price or mix. It set that against growth in open-web native itself. An honest answer means taking each estimate apart. This chapter does that before it quotes any of them. ![One estimate reports native, and it is the one that counts social feeds.](figures/fig-inclusion-matrix.svg) *Figure fig-inclusion-matrix. **One estimate reports native, and it is the one that counts social feeds.** Which formats and surfaces each measurement source includes, from the source’s own definitions. Only eMarketer publishes a native subtotal, and its vintage splits show it counts social in-feed; an AI-assisted eMarketer FAQ also lists outstream and rewarded video, but the forecast’s own scope is not published, so that cell is marked unclear. IAB/PwC folds native into display and treats social as a cross-cutting channel; the agency forecasters publish no native line at all. Y means the source says it is included; S means included inside a larger line that cannot be separated; N means the source excludes it; ? means the source does not say, or mentions it without saying how it is counted (dentsu names no treatment of paid search, so it is ?, not N). Units: categorical. Scope: US and global measurement sources, 2025–2026 vintages. Sources [3], [7], [9], [11], [14], [15], [16]. Caveat: A question mark is absence of a statement, not evidence of exclusion. Coding is the research pass’s reading of each source’s definitions; the cell text behind every code is in data/estimates.csv. Data: data/estimates.csv* ### What the word covers in each source IAB's Native Advertising Playbook 2.0 cut native to three types in May 2019. They are in-feed and in-content units, content recommendation units, and branded or native content. The 2013 playbook had six types [3, 6]. Content recommendation units sit in the blocks of suggested links on publisher pages. The new playbook removed paid search, "even though search ads technically meet the IAB definition of native". It folded promoted listings on commerce sites into the in-feed type. It also dropped "buying and selling" and "measurement" as tests of whether an ad is native. Its reason was that programmatic buying no longer defines what is native [3]. In the playbook, in-feed native covers three surfaces: content feeds, product feeds and social feeds [3]. The playbook is broad by design. The sources that measure spend are narrower. Each one narrows it in its own way. IAB and PwC's report on US internet ad revenue for 2025 puts the total at $294.6 billion, up 13.9%. It does not report native. Native is buried inside display, at $81.6 billion. Display "includes banner, rich media, sponsorship and native revenues" [7, 8]. Social media, at $117.7 billion and up 32.6%, is reported next to the formats, not as one of them. The five formats already add up to the total. So social revenue sits inside display and video [7]. The figures are US-based earned revenue. They come from a PwC survey of sellers. Public data and estimates fill in for sellers that did not take part. PwC states it has not audited them [7]. EMARKETER's US "native display" is the only explicit native estimate in the source set reviewed. Other syndicated research may exist that this project could not open. Some may fold native into wider categories. Neither kind is compared here. Each edition of the EMARKETER forecast is a vintage. The December 2025 vintage is the latest this project could open. It forecasts $147.98 billion for 2026, up 13.1%. That implies about $130.8 billion for 2025. An AI-assisted EMARKETER FAQ lists in-feed units, sponsored content, content recommendation widgets, in-app rewarded video and sponsored outstream video as native formats, and excludes banners and pop-ups. The forecast's own published scope says only display ads that reflect the form and function of their surroundings, so whether it counts the video formats is not stated [9]. Its separate digital video line counts outstream and in-feed video. So the two totals may overlap and should not be added [10]. EMARKETER also says native's share of total US display has fallen "as connected TV and retail media expand". So under its own definition, native grows more slowly than display [9]. None of the agency forecasters publishes a native line. MAGNA's June 2025 global forecast has none. Nor do WPP Media's June 2026 update and dentsu's December 2025 forecast. IAB Europe's 2025 benchmark for 30 markets has none either. All four report search, social, retail media and video [11, 12, 13, 14, 15, 16]. The Tech Lab's OpenRTB Native standards page still says the API covers the six types of the 2013 playbook. The 2019 playbook replaced them [17, 3]. ### The definitional gulf, in one ratio EMARKETER's implied 2025 native display figure is about $130.8 billion. That is about 1.6 times IAB/PwC's whole 2025 display total of $81.6 billion. By IAB's own account, that total includes native. The two numbers do not conflict. They fence different fields. EMARKETER counts social in-feed as native, and its FAQ lists outstream video too. IAB counts social as a channel that cuts across formats. It counts in-feed video under video. A reader may quote "$148 billion of native advertising". If that reader does not add that most of it is Facebook, Instagram and TikTok feeds, they have changed the subject. ![The only explicit native estimate in the sources reviewed is a social-heavy series that is rewritten between vintages.](figures/fig-spend-vintages.svg) *Figure fig-spend-vintages. **The only explicit native estimate in the sources reviewed is a social-heavy series that is rewritten between vintages.** eMarketer’s US native values across the vintages this project could open. Social feeds took 74–84% of the total in every vintage whose split could be read, the latest from 2019. The January 2023 vintage raised 2019 from $43.9 billion to $47.3 billion; no public 2024 value was found, and 2025 is implied from the 2026 forecast. Because each vintage rewrites history, the values are drawn as bars by vintage, not as one line. Units: USD billions, nominal. Scope: United States; eMarketer definitions as stated per vintage. Sources [20], [21], [18], [22], [9]. Caveat: Values from different vintages use different definitions and eMarketer restates history between them; the 2025 point is an arithmetic implication, not a published figure. Data: data/figure-data.json#native_spend_vintages* How much is social? The last published split is from the 2019 vintage. Social was 77% of native display in 2018, falling to 74% by 2020. Native was expected to be 95.6% of social display spend. For non-social display spend, the expected share was only 30.8% [18, 19, 20, 21]. A search-indexed eMarketer page appears to put 2023 social native at $68.17 billion. That would be about 70% of that year's total. But the page could not be opened. So the number is a lead, and it is not used. In 2017 the social share was 84.2% [20]. Read plainly, "native display" as measured was mostly social-feed ads in each vintage that published a split. In 2018–2020, the non-social part was about a quarter of the total. That part also includes in-app ad space outside this paper's core. Those shares describe those vintages. They are not a measured 2026 split. They are not used here to estimate one. The public eMarketer chart from January 2023 fills most of the gap. It shows $47.3 billion for 2019, $58.7 billion for 2020, $80.6 billion for 2021, $87.0 billion for 2022 and $97.5 billion for 2023. In that chart, native was 60% to 63% of US display spending [22]. No public value for 2024 was found. Any current split sits behind a paywall that returned access errors to this project. The series is drawn as bars by vintage, not as one line. That is because eMarketer revises past years from one vintage to the next. The 2019 vintage revised 2018 up by about 7%. The 2023 vintage raised 2019 from $43.9 billion to $47.3 billion [18, 22]. **Market-size reconciliation** | Estimate | Published | Metric | Value | Year | Geography | Includes | Excludes | Combinable with | Src | |---|---|---|---|---|---|---|---|---|---| | IAB Playbook 2.0 taxonomy | 2019-05 | IAB Playbook 2.0 taxonomy (not a spend estimate) | taxonomy | | taxonomy | open-web in-feed and recommendation (In-Feed/In-Content; Content Recommendation); sponsored editorial (Branded/Native Content); native video (outstream, rewarded) (if in-feed/in-content styled); social in-feed (social feeds are in-feed); retail sponsored products (Promoted Listings folded into In-Feed) | in-stream video (not addressed as native); paid search (removed; 'technically' native); banners and rich media (unless customized to match site) | None; a taxonomy, not an estimate. | [3] | | IAB/PwC 'Display' | 2026-04 | IAB/PwC 'Display' $81.6B (2025) | $81.6B | 2025 | US, GAAP earned, gross for traffic buyers | open-web in-feed and recommendation (inside display; not separable); sponsored editorial (sponsorship inside display); banners and rich media | in-stream video (Video format); paid search (Search format); retail sponsored products (Commerce Media, cross-cutting; search-type listings in Search) | IAB/PwC display and social overlap (social is a channel inside formats); display cannot be compared with eMarketer native display. | [7] | | IAB/PwC 'Social media' | 2026-04 | IAB/PwC 'Social media' $117.7B (2025) | $117.7B | 2025 | US; channel overlaps formats | native video (outstream, rewarded) (if on social platform); in-stream video (if on social platform); social in-feed (all formats); banners and rich media (if on social platform) | open-web in-feed and recommendation; sponsored editorial; paid search; retail sponsored products | IAB/PwC display and social overlap (social is a channel inside formats); display cannot be compared with eMarketer native display. | [7] | | EMARKETER 'native display' | 2025-12 | EMARKETER 'native display' $147.98B (2026F) | $147.98B | 2026 | US; proprietary model | open-web in-feed and recommendation (content recommendation widgets, in-feed units); sponsored editorial (sponsored content); native video (outstream, rewarded) (in-app rewarded video, sponsored outstream video); social in-feed (in-feed units) | in-stream video ('in-stream video is by definition nonnative', A-S07); paid search (not in list); banners and rich media (excluded) | Not with eMarketer digital video (outstream overlaps) nor with IAB display (different fence: eMarketer native is 1.6× IAB display). | [9] | | MAGNA 'Digital Pure Players' | 2025-06 | MAGNA 'Digital Pure Players' (2025) | no native line | 2025 | global/US media-owner net revenue | social in-feed (social media); paid search (search, retail search); retail sponsored products (retail search) | sponsored editorial; native video (outstream, rewarded); in-stream video (short-form video separate); banners and rich media | Media-owner net revenue; not comparable with IAB gross earned revenue or eMarketer spend. | [11] | | WPP Media TYNY midyear 2026 | 2026-06 | WPP Media TYNY midyear 2026 | no native line | 2026 | global ad revenue ex US political | social in-feed (social); paid search (search); retail sponsored products (retail media) | | Global, ex US political; no native line. | [14] | | dentsu Dec 2025 | 2025-12 | dentsu Dec 2025 | no native line | 2025 | global, net of discounts and agency commission | in-stream video (online video); social in-feed (social); retail sponsored products (retail media) | | Global net of discounts and commission; no native line. | [15] | | IAB Europe AdEx 2025 | 2026-07 | IAB Europe AdEx 2025 | no native line | 2025 | 30 European markets, EUR | in-stream video (video > half of display); social in-feed (social EUR35.5bn); paid search (search); retail sponsored products (retail media EUR13.3bn); banners and rich media (display) | | Europe, EUR; no native line. | [16] | Secondary sources drift from the primary. One trade report turned "native is 95.6% of social display" into "96% of native ad spending goes to social". That is the same two numbers the wrong way round. Another gives two different figures for the same MAGNA forecast [19]. Each number in this chapter comes from a source this project opened. The register records which. ### What the industry's own accounts show ![Native sits inside display, the slowest of the four major named formats in the industry’s own accounts.](figures/fig-iab-formats.svg) *Figure fig-iab-formats. **Native sits inside display, the slowest of the four major named formats in the industry’s own accounts.** IAB/PwC does not report native as a line; it sits inside display, which grew 9.8% in 2025 (the residual “other” category grew about 6.9%) and compounded at about 9.5% a year from 2021 while video compounded at about 18.5%. Social media, $117.7 billion in 2025 and up 32.6%, is a channel embedded inside these formats rather than a format of its own. Units: USD billions, US, GAAP earned revenue as surveyed by PwC (not audited). Scope: United States, calendar years 2021–2025. Sources [7]. Caveat: Display includes banners, rich media, sponsorship and native; the native share within it is not disclosed. Data: data/figure-data.json#iab_formats* IAB/PwC's format series is a structured industry revenue survey, filled out with estimates. It is not an audit, but it is the most consistent view of where the money went. It is unkind to native's home. Display grew 9.8% in 2025, from $74.3 billion to $81.6 billion. That was the slowest of the four major named formats. The leftover "other" category grew about 6.9%. Video grew 25.4% to $78.0 billion, and search grew 11.0% to $114.2 billion. From 2021 to 2025 display's compound growth was about 9.5% a year. Video's was 18.5% [7]. Social's 32.6% growth to $117.7 billion sits inside those formats. Commerce media, another line that cuts across formats, reached $63.4 billion, up 18% [7]. Display's growth rate cannot pick out recommendation widgets. Because social cuts across the formats, display itself includes social units. What the series shows is narrower. The format that holds open-web native is losing share to video. ### The budget destinations, on their own terms ![The two large US-listed native platforms are about 1.6% of Meta on a gross basis and 0.6% after paying publishers.](figures/fig-comparators.svg) *Figure fig-comparators. **The two large US-listed native platforms are about 1.6% of Meta on a gross basis and 0.6% after paying publishers.** FY2025 reported revenue as each company defines it. Google Network, the line closest to open-web publisher monetisation, fell 1.9% while Google Search grew 13.4%. Taboola and Teads Holding Co. together reported $3.2 billion of revenue and $1,243.2 million of ex-TAC gross profit, a company-defined measure; revenue less traffic acquisition cost is $1,226.8 million. Ratios to Meta are recomputed in data/derived-checks.json. Units: USD billions, fiscal 2025. Scope: global revenue for every company; US-only comparators were not obtainable from the filings opened. Sources [23], [24], [27], [28], [29]. Caveat: Lighter bars are revenue reported gross of partner payments; darker bars are other reported revenue lines. They are different quantities, shown on one scale only to convey order of magnitude. Data: data/figure-data.json#comparators* The brief also asked how native compares with Meta and other places budgets go. It asked for clearly defined terms. The terms are the problem, so the figure states them. Meta's 2025 ad revenue was $196.2 billion, up 22%. Impressions rose 12% and average price per ad rose 9%. Those compound to about 22% [23]. Google's ad revenue was $294.7 billion. Search grew 13.4% to $224.5 billion, and YouTube grew 11.7%. Google Network is the line that earns money from other people's pages. It fell 1.9% to $29.8 billion, its third annual decline in a row. It is now 10.1% of Google's ad revenue. In 2023, it was 13.2% [24, 25]. Network combines AdSense on the web, AdMob in apps and Ad Manager. The filing says the decline came from AdSense, partly offset by AdMob. So the web piece fell by more than 1.9% [26]. Amazon's advertising services add up to $68.6 billion across the four reported quarters of 2025 [27]. Against those, Taboola's $1,912 million of revenue and Teads Holding Co.'s $1,300.5 million are small. They must be compared with care. Both recognise most revenue gross, as principal. That is, they book what buyers pay as revenue. They record their payments to publishers as traffic acquisition cost. Google also reports its network revenue gross. The share it pays to partners sits in cost of revenues. Meta, by contrast, sells mostly its own ad space [28, 29, 24]. The two platforms reported $3,212.5 million of revenue between them. Their ex-TAC gross profit, as each firm defines it, totals $1,243.2 million. Revenue minus reported traffic acquisition cost totals $1,226.8 million. The gap is Taboola's non-cash add-back. On revenue, they are about 1.6% of Meta's ad revenue. On ex-TAC, they are about 0.6% [28, 29, 23]. The world's largest open-web native firms are a rounding error in the budgets they compete for. Google Network's decline is the most suggestive number here. But the paper does not lean on it. Network is not a native-only line. It includes app and display ads sold through AdSense and Ad Manager. The decline fits with budgets pooling on owned surfaces, but it does not prove that. Like the recommendation platforms, Network depends on third-party page views. Chapter 11 shows those falling. ### Answering the growth question On the evidence above, reported native growth is hard to trace to open-web native. The one native series counts in-feed units on all surfaces, social included. Three-quarters or more of it was social feeds in the vintages that published a split. The industry's own accounts put native inside display. Display includes social units, and it grew more slowly than video. Growth in open-web native itself shows up only in company filings. In 2025, Taboola's reported ex-TAC gross profit grew 6.9%. Teads Holding Co.'s revenue fell 11.7% that year on a pro forma basis. No source reviewed measures the open-web native market as such. So this paper does not quote a size for it. It treats any number that claims to be one as a choice about what to count. That choice still needs to be checked. > **Three numbers, kept apart.** > Company scale benchmark: the two large US-listed platforms reported $3.2 billion of revenue in 2025. That figure is global, and both sell formats beyond native. It is not a market size. No source reviewed publishes one for US open-web native. Addressable opportunity: Taboola claims $55 billion for performance advertising broadly. It cites its own and outside industry data [1]. Management revenue guidance: Taboola's own 2026 range, cut in August to $1.93–1.96 billion [30]. That is a forecast of what one firm expects to book. These are three different questions with three different answers. Chapter 1 explained why the paper refuses to blend them. ## 3. Distribution and platform power *Open-web native is sold mainly through two large US-listed firms. Long, exclusive and guaranteed contracts lock up their supply. The second tier has been folding into fewer owners.* > **In short.** > In 2025, Taboola and Teads Holding Co. reported $3.2 billion of revenue between them. Most of it was booked gross, before paying publishers. Their edge is supply, not a few big advertisers. Taboola's ten largest advertisers are under 10% of its revenue. Two supply partners, Yahoo and Microsoft, are about 34% of its revenue from advertisers on its partners' properties [1]. Teads does not disclose partner concentration. The second tier has been bought, merged or retired since mid-2024. Nativo went to Life360, Sharethrough to Equativ. Nine days before this paper's cutoff, Taboola made a recommended cash offer for Dianomi. It is worth about £19 million upfront, against Dianomi's £27 million of revenue [31]. Programmatic native is native bought through automated systems. In March 2017, its standard was last finalised. No market-wide figure was found for how much native is bought that way. In native ads, the money is decided in distribution. So this chapter starts there, not with formats or targeting. ### Two companies, two supply strategies Taboola.com Ltd reported gross revenue of $1.9 billion for 2025. Its ex-TAC gross profit was $713.5 million [1, 32]. Ex-TAC means revenue after traffic acquisition cost. That cost is the money paid to publishers for the ad space. Ex-TAC is the number the firm manages to. Chapter 4 explains why you cannot compare it across firms. Taboola's supply strategy is written into its contracts. Historically, most of its publisher deals required exclusivity or other preferred-usage incentives. The average term at signing was more than two years [1]. Three deals define its network. The first is Yahoo. The deal was announced on 28 November 2022. It closed on 17 January 2023, and runs for thirty years. It gives Taboola the exclusive right to run native ads across Yahoo's digital properties. Yahoo got 24.99% of Taboola's shares and a board seat [33, 34]. The contract has since been amended at least seven times. Amendments 4 to 7 took effect between February 2025 and March 2026; 5, 6 and 7, in turn, took effect in June 2025, August 2025 and March 2026. They added portal access through November 2027, and a direct insertion-order route for Yahoo. They also rewrote the monthly method for judging performance. The filing redacts the money terms of that method [35]. A thirty-year exclusive needed four amendments in about a year. It is being renegotiated in public, with the numbers hidden. The second is Microsoft. The tie dates from a January 2016 MSN deal. That deal covered 50 markets and 24 languages. In March 2025 Taboola announced ten years of serving MSN, Edge and Windows. It also announced a move into Outlook and Microsoft 365, but gave no term or renewal date. Nor did it say whether the deal is exclusive [36, 37]. Microsoft also sells native itself across Edge, Outlook and Bing [38]. In 2020 it signed MGID for native across its news network [39]. So "exclusive" in this market appears to cover the recommendation unit, not the page. That unit is the block of suggested links a platform runs on the page. The third is Apple. In mid-2024, Taboola became an authorised seller of native ads in Apple News and Apple Stocks. It was still named as Apple's ad partner in February 2026, amid criticism of ad quality [40, 41, 42, 43]. What that deal covers, and where, is known only from press reports. Teads Holding Co. is the other listed platform. It used to be called Outbrain Inc. Outbrain bought Teads from Altice. The deal was announced on 1 August 2024, at about $1 billion. It closed on 3 February 2025 on revised terms worth about $900 million. The parent renamed itself Teads Holding Co. in June 2025. It trades as TEAD [44, 45, 46]. It describes three kinds of publisher contract: revenue share, programmatic bidding and guaranteed minimums. It says they cover about 10,000 media owners. Its top 20 partners, it says, have stayed seven years on average [46]. Its 2025 revenue was $1.3 billion. It rose 46% because the purchase was folded into its accounts. On a like-for-like basis it fell 11.7% [46]. Chapter 4 takes that apart. ![Two supply partners account for more revenue than the whole top-ten advertiser list.](figures/fig-concentration.svg) *Figure fig-concentration. **Two supply partners account for more revenue than the whole top-ten advertiser list.** In 2025 Yahoo and Microsoft together accounted for about 34% of Taboola’s gross revenue generated on digital properties, and its top five properties for about 44%. Its ten largest advertisers were under 10% of revenue, none above 3%. The moat is supply rights, not demand. Units: percent of 2025 revenue. Scope: Taboola, FY2025. Sources [1]. Caveat: The advertiser figure is a disclosed ceiling, not a point estimate; the two supply figures are stated as approximate. Data: data/figure-data.json#concentration* ### Supply is concentrated, and so is one demand channel The clearest fact in Taboola's filing is the gap between its two sides. In 2025 Yahoo and Microsoft were its largest partners. No other partner reached 5% of revenue. Its ten largest advertisers together were under 10%, and none was above 3% [1]. Outside Yahoo, demand is diffuse. Revenue billed through Yahoo was 10.5% of 2025 revenue and 16.6% in the second quarter of 2026, more than the ten largest advertisers combined. In the fourth quarter of 2025 Taboola had about 2,200 "Scaled Advertisers". The term means those that spent more than $100,000 over the last four quarters. In that quarter they made up 84% of revenue. Each brought in about $204,000 on average [1]. That is a mid-market demand base tied to a highly concentrated supply base. The strategic reading follows. Losing a major supply partner could hit revenue much harder than losing any one disclosed advertiser. How much harder would depend on substitution, margins, contract terms and transition periods. So the contracts that matter most are on the supply side. The next chapter covers their cost: the guarantees, and the equity given to Yahoo. Taboola's count of digital property partners was "approximately 14,000" in its 2025 annual report. It was "approximately 12,000" in its second-quarter 2026 filing. Taboola does not define what changed [1, 47]. Revenue grew over the same period. Taboola does not disclose whether the change means lost partners, a new way of counting or something else. Meanwhile it has been opening up ad space that is not native. On 26 February 2025, it launched Realize. It sells display and other ad slots across Taboola's network. Taboola describes that network as about 600 million daily active users. By October 2025 TIME, The Weather Channel, Gannett, Nexstar and Slate had opened display space to Realize advertisers [48, 49]. In August 2026 Realize began running programmatic display on NBCNews.com and TODAY.com. Taboola called this a first [50]. In June 2026 it also opened DeeperDive to ads. DeeperDive is the generative AI answer engine it runs on publisher sites. Now it places ads inside AI results pages [51]. The largest native platform is moving past the recommendation unit, into display and AI answers. It does not disclose how much of its growth each line adds. ### The second tier has fewer owners than it did two years ago ![Five and a half years, one direction: fewer owners.](figures/fig-consolidation-timeline.svg) *Figure fig-consolidation-timeline. **Five and a half years, one direction: fewer owners.** Announcements and completions are drawn as different marks because they were often months apart and on different terms (the Teads deal closed at roughly $900M against about $1B announced). Second-tier exits cluster from mid-2024: Sharethrough into Equativ, Nativo into Life360, and Taboola’s pending offer for Dianomi. Units: dates. Scope: global, entities named in chapter 3. Sources [62], [64], [407], [66], [33], [34], [52], [40], [44], [56], [45], [46], [48], [37], [408], [74], [53], [49], [57], [58], [59], [31]. Caveat: Events are those found in opened sources; private-company events without a public record are absent. Month-only dates are plotted mid-month. Data: data/figure-data.json#timeline* Between June 2024 and September 2026, the independent tier of open-web native fell into fewer hands. - Sharethrough merged with Equativ. The deal was announced on 12 June 2024. It made a firm of 720 or more staff in 18 countries. On 9 June 2025, Sharethrough began to operate under the Equativ brand. But a Sharethrough-branded site was still live at the cutoff [52, 53, 54]. - In 2017, Outbrain bought the native DSP Zemanta. A DSP, or demand-side platform, is the tool buyers use to buy ads. On 6 August 2024 Zemanta was renamed Outbrain DSP [55, 56]. - Life360 bought Nativo. It announced the deal on 10 November 2025, for about $120 million in cash and stock. It completed the purchase on 2 January 2026, as a filing of 5 January reported [57, 58, 59]. Nativo's revenue and publisher count at the time were not disclosed. - Dianomi plc is listed on AIM. It focuses on native for financial content. On 18 September 2026 it announced a cash offer from Taboola Europe Limited. Its board intends to recommend it unanimously. The offer is 64p plus up to 24p contingent per share. That is about £19 million, rising to £27 million. The contingent part is paid only if a set of Dianomi publishers move to Taboola's standard terms. It is scaled to those publishers' revenue. Holders of 75.3% of shares have committed to the offer. The deal is expected to close before the end of 2026. It is subject to shareholder approval. It also depends on the UK competition authority not opening an in-depth review [60, 61, 31]. Dianomi's 2025 revenue was £27.4 million. Since March 2021, TripleLift has been majority-owned by Vista Equity Partners. It changed CEO twice between July 2024 and February 2025. In July 2025 it made layoffs. Anonymous sources put the cuts in the mid-to-high double digits. The base was over 400 staff [62, 63, 64]. MGID, Revcontent, Readpeak and Kargo remain private. MGID claimed 185 billion monthly impressions in 2025. In 2020, it used the same figure. Its legal owner could not be established from any opened source [65, 39]. Two research leads for this chapter were wrong. They are recorded here as such. Sharethrough was not bought in 2023. The Equativ deal was announced in June 2024. Adyoulike was not bought by Opera. OpenWeb bought it in April 2022. The reported price was $100 million [66]. > **What the Dianomi price says.** > Taboola's offer is a 68% premium to Dianomi's share price. The upfront price of about £19 million is roughly 0.69 times Dianomi's £27.4 million of 2025 revenue. Up to £8 million more is contingent. That ratio uses equity value. It is not enterprise value to revenue, because Dianomi held about £5.8 million of cash and no debt. On its own, the ratio says nothing about distress. The offer was not complete at the cutoff. It needs shareholder, court and competition approvals [60, 31]. The buyer is the market leader. ### Buying routes: the major platforms checked support native, and the market split is unmeasured Programmatic native runs on IAB Tech Lab's OpenRTB Native Ads specification. Native entered the bid stream with OpenRTB 2.3 in January 2015. In March 2017, the spec was last finalised as version 1.2. That version added third-party creative, event trackers and a privacy link [4, 67]. Prebid.js uses it for header bidding. There, the OpenRTB approach replaces the legacy native format [68]. The major platforms checked support the format. They include DV360, Google Ad Manager, AdSense and Microsoft Advertising. The Trade Desk has supported it since March 2016. Its first native supply partner was Sharethrough. Amazon Publisher Services supports it too. It announced APS Native ads for Amazon advertiser demand on 21 May 2025. It did not say when or where they would be available, or whether other DSPs get access [69, 70, 71, 72, 73, 38, 74]. None of the platforms above says how much the format is used. No IAB figure for OpenRTB Native uptake was found. Prebid publishes none either. A Sharethrough executive said in 2016 that uptake of the protocol had been slower than expected [71]. Support is not the same as spend. Some vendors disclose fragments, such as Dianomi's programmatic revenue. But no market-wide measure was found of what share of open-web native is bought through open programmatic pipes. The documented routes are the platforms' own. Taboola sells through Realize, both self-serve and managed service. Teads sells through the self-serve platform from its Outbrain side, its own DSP and managed service. MGID, Revcontent and Readpeak sell through self-serve platforms. Dianomi and Kargo sell mainly direct. A buyer who wants open-web recommendation inventory at scale must still go to a platform. It has to be the one that owns the contract with the publisher. That is what supply rights mean in practice. It is why the next chapter follows the money through those contracts. ## 4. The economics of an impression and a click *About six in ten dollars of reported revenue are booked as traffic acquisition cost. The rest pays for the business. At the two large listed platforms, that business is moving in opposite directions.* > **In short.** > In 2025 Taboola booked 63.5% of revenue as traffic acquisition cost, or TAC. Its adjusted EBITDA was 30% of its ex-TAC gross profit. For Teads Holding Co. the two figures were 59.3% and 17.6%. Each firm defines these measures its own way. They do not compare line for line. Taboola's cost of guarantees to publishers is what it pays above what revenue share alone would owe. That cost was about 18% of TAC in 2024, up from 9% to 10% in 2021 and 2022 and 13% in 2020; the 2023 value was not captured. It was about 15% in 2025 and 13% in the second quarter of 2026. In 2025, Teads Holding Co.'s revenue fell 11.7% on a pro forma basis. It wrote down $352 million of goodwill. It funded its purchase with $637.5 million of 10% notes. In August 2026 it suspended guidance. Dianomi is the one small listed vendor whose cost of sales is almost all publisher payments. It pays publishers 73% of revenue. Its reported revenue per click on open-web publishers was about 4.9 times that inside Apple News. Native platforms sell a click to an advertiser and buy a slot from a publisher. Everything about their economics follows from two things. One is the gap between those two prices. The other is what they must promise publishers to keep the slot. ![Roughly six in ten dollars of reported revenue are recorded as traffic acquisition cost.](figures/fig-platform-economics.svg) *Figure fig-platform-economics. **Roughly six in ten dollars of reported revenue are recorded as traffic acquisition cost.** FY2025 revenue split into traffic acquisition cost (the accounting expense for publisher payments and impressions bought on exchanges, including non-cash amortisation at Taboola), other cost of revenue and GAAP gross profit. This is an expense split, not a cash-flow waterfall. Both companies recognise most revenue gross, as principal, so the bar is the advertiser dollar. Ex-TAC gross profit is each company’s own non-GAAP measure and is not comparable line for line: Taboola adds back non-cash amortisation of its Yahoo agreement inside TAC; Teads adds back only other cost of revenue. Units: share of gross revenue, USD millions in labels. Scope: FY2025, company-wide (Teads Holding Co. includes video, CTV and display; Taboola includes display sold through Realize). Sources [1], [32], [46], [75]. Caveat: Teads Holding Co. FY2025 consolidates the legacy Teads business only from 3 February 2025; its pro forma revenue fell 11.7% year on year. Data: data/figure-data.json#platform_annual* ### Where a gross dollar goes Both listed platforms book most revenue gross, as principal. Payments to publishers are booked as traffic acquisition cost, or TAC, inside cost of revenues [1]. So the reported revenue line is roughly the advertiser dollar. TAC is an accounting expense. It is not a line of cash paid out in the same period. Taboola's TAC includes non-cash amortisation of the Yahoo agreement asset. It also includes payments for impressions bought on exchanges. In 2025, Taboola took in $1,912.0 million. It paid $1,214.9 million in TAC and spent $127.6 million on other cost of revenue. It reported GAAP gross profit of $569.5 million [1, 32]. Its ex-TAC gross profit was $713.5 million, or 37.3% of revenue. Ex-TAC gross profit is, roughly, revenue minus TAC. It is the non-GAAP measure Taboola manages to. Adjusted EBITDA was $215.5 million, 30.2% of ex-TAC. GAAP net income was $42.3 million, after a loss of $3.8 million in 2024. Teads Holding Co. took in $1,300.5 million. It paid $770.8 million in TAC and $100.6 million in other cost of revenue. It reported gross profit of $429.1 million. Its ex-TAC gross profit was $529.7 million and adjusted EBITDA $93.4 million, 17.6% of ex-TAC. Its net loss was $517.1 million. That included a $352.1 million goodwill impairment and $28.9 million of acquisition costs. It also included $15.3 million of restructuring charges [75, 46]. Three cautions before comparing the two. First, each firm defines ex-TAC its own way. Taboola adds back non-cash amortisation of its Yahoo agreement asset, which sits inside TAC. That was $16.4 million in 2025. From the second quarter of 2026 it also adds back a $12.2 million write-off of publisher prepayments. Reported ex-TAC growth that quarter was 11.8%. Under the old method, with the write-off left in, it would have been about 4.7%. That gap shows how much the add-back choice moves the number. It does not measure underlying growth, because the write-off also hit the reported accounting result [47, 76]. Teads adds back only other cost of revenue. Second, neither firm reports net revenue. Gross revenue less TAC is a derived proxy. For Taboola it differs from ex-TAC by the add-backs. Third, Teads Holding Co. includes video, connected TV and display. Taboola includes display sold through Realize. So neither figure is a total for the native market. ![Taboola’s TAC share rose from 59% to 64% of revenue between 2022 and 2025; the Outbrain lineage’s fell, then merged.](figures/fig-tac-trend.svg) *Figure fig-tac-trend. **Taboola’s TAC share rose from 59% to 64% of revenue between 2022 and 2025; the Outbrain lineage’s fell, then merged.** Taboola’s traffic acquisition cost was 59.3% of revenue in 2022, 62.8% in 2023, 62.4% in 2024 and 63.5% in 2025, not a rise every year; the period includes the Yahoo supply deal. The Outbrain business ran at 73.5–76.3% before the Teads acquisition; the 2025 point (59.3%) is the combined Teads Holding Co., whose mix includes video and CTV and is not comparable with the prior three years. Units: percent of gross revenue. Scope: fiscal years 2022–2025, company-wide. Sources [77], [1], [32], [78], [46], [75]. Caveat: The 2025 Teads Holding Co. point mixes two businesses and consolidates one of them for eleven months; do not read it as a like-for-like improvement. Data: data/figure-data.json#platform_annual* Taboola's TAC has risen as a share of revenue in each year but one. It was 59.3% in 2022, 62.8% in 2023, 62.4% in 2024 and 63.5% in 2025 [1, 77]. It is not a pure publisher payout ratio. It also covers ad space bought on exchanges, up-front and incentive payments, and non-cash amortisation. Since 2024, some amounts tied to Yahoo have been shown another way. That affects the comparison. The Outbrain business paid out more. The share it paid out fell from 76.3% to 73.5% between 2022 and 2024. Over the same years its revenue shrank from $992 million to $890 million [78]. A shrinking recommendation business can still improve its unit margin. Outbrain did that in the three years before it bought Teads. ### The price of exclusivity Chapter 3 showed that supply rights are the moat. The filings show what they cost. Taboola discloses two ways it pays publishers. The most common is a revenue share. The other is a minimum guarantee. Under it, Taboola pays whichever is more: the revenue share, or an agreed amount per thousand page views. It settles each month. Guarantee contracts usually run two to five years. Large property contracts "in general, contain minimum guarantee requirements" [1, 47]. ![Guarantee costs rose from about a tenth of supply cost in 2021–2022 to about 18% in 2024, then eased.](figures/fig-guarantee-share.svg) *Figure fig-guarantee-share. **Guarantee costs rose from about a tenth of supply cost in 2021–2022 to about 18% in 2024, then eased.** Taboola’s cost of publisher guarantees (payments in excess of what revenue share alone would have required) was about 13% of traffic acquisition cost in 2020, about 9–10% in 2021–2022, about 18% in 2024, about 15% in 2025 and about 13% in the second quarter of 2026. The 2023 value was not captured in the filings opened, so the series cannot show whether 2024 was the peak. Trade reporting in 2018 described guarantees as fading; the filings say they came back. The rise followed the Yahoo agreement of January 2023, but the filings do not attribute it to that deal. Units: percent of TAC; USD millions indicative. Scope: Taboola, fiscal years 2020–2025 (2023 not captured in the opened filings). Sources [77], [1]. Caveat: Percentages are as stated in filings ("approximately"); the dollar labels are the paper’s recomputation and carry the same approximation. Data: data/figure-data.json#guarantee_share* Taboola also discloses what those guarantees cost. The cost is what it owes under guarantees above what revenue-share terms would have required. That was about 9% of TAC in 2021 and 10% in 2022, about 18% in 2024 and about 15% in 2025. In the second quarter of 2026 it was about 13% [1, 77, 47]. Applied to reported TAC, the 2025 share works out to something on the order of $180 million. Publishers got that money because a guarantee, not a revenue share, set the price. In January 2018 trade reporting said publishers were giving up $1–2 million annual guarantees for revenue shares, to gain flexibility [79]. The filings say guarantees came back. Their share of supply cost reached about 18% in 2024, roughly twice the 2021 level, then eased; with 2023 missing, the series cannot show whether 2024 was the peak. In January 2023, the Yahoo agreement closed. The rise followed that, but the filings do not say the deal caused it. The guarantee described is a rate per thousand page views. So it guarantees a price, not a total annual income. Teads Holding Co. warns in its own filing that guarantees "require us to pay our media partner for the ad impressions we receive, regardless of whether the consumer engages with the ad or we are paid by the advertiser". It also warns that TAC "may not correlate with fluctuations in revenue" [46, 78]. That is the risk transfer in one sentence. Under a guarantee, the platform bears the gap between what advertisers pay and what the slot costs. The publisher does not. ### What Yahoo costs, and what it returns The Yahoo agreement was paid for in equity. Taboola books it as an asset and amortises it inside TAC. At the end of 2025, that asset stood at $270.2 million. In the same year Taboola bought back 76.9 million of its own shares for about $255.4 million [80, 1, 47, 32]. It discloses two related-party lines. The first is revenue billed through Yahoo, where Yahoo acts as the advertiser. That was $201.6 million in 2025, 10.5% of revenue, down from $233.6 million. In the first half of 2026 it rose to $148.9 million, 15.8% of revenue. The second is TAC paid to Yahoo for placements. That rose to $349.0 million in 2025 from $275.5 million. In the first half of 2026 it rose to $201.1 million [1, 47, 80]. Those two lines do not add up to a profit test. This paper does not run one. Revenue billed by Yahoo is not all that is earned on Yahoo supply. Some advertisers buy direct through Taboola and run on Yahoo pages. They are not in that line. What can be said is narrower. In its 2022 announcement, Taboola forecast about $1 billion of annual revenue from the deal. It also set out a path to $1 billion of ex-TAC gross profit by 2025 [80]. Total ex-TAC in 2025 was $713.5 million. The forecast was not met on its announced timeline. The amendments described in chapter 3 rewrote the terms for judging performance. The money terms were redacted. In May 2026 Taboola's CEO said it paid over $1.5 billion to publishers in 2025 [81]. Audited 2025 TAC was $1,214.9 million. This paper cannot reconcile the two figures. It uses only the audited one. ### Post-acquisition losses, impairment and financing pressure Teads Holding Co.'s 2025 headline growth of 46% is acquisition accounting. Pro forma figures treat the deal as if it had closed on 1 January 2024. On that basis revenue fell 11.7%, from $1,507.2 million to $1,330.9 million. The pro forma net loss was $545.5 million. Between 3 February and 31 December 2025, legacy Teads added $517.2 million of revenue and a $350.8 million net loss [46]. The purchase was funded with $637.5 million of 10.000% senior secured notes due 2030, issued at a discount in February 2025. At the end of 2025, $628.2 million of principal was still owed. About $62.8 million of interest is due each year through 2029. At 30 June 2026 the balance sheet showed $614.5 million of debt: $607.4 million long-term and $7.1 million short-term. Against that the firm held $88.0 million of cash. Stockholders' equity was $7.4 million, down from $95.4 million at year end [46, 78, 82]. The 2026 quarters got worse. In the first quarter, revenue was $266.0 million, down 7%. Adjusted EBITDA was $0.8 million. In the second quarter, revenue was $284.6 million, down 17%. Ex-TAC fell 14%. Adjusted EBITDA was $7.0 million and the net loss $42.5 million. On 6 August 2026 the firm suspended guidance. That included its target of about $100 million of adjusted EBITDA for the year. It cited "the volatility of the Direct Response and SME business" facing "open-web headwinds" [83, 82, 75]. That $7.0 million second-quarter figure had missed the firm's May guidance of $14 million to $22 million [83, 82]. The notes carry about $31.4 million of interest every six months. Half-year adjusted EBITDA of $7.7 million was roughly a quarter of that. Connected TV, the line the company highlights, grew 67% year on year. It made up 13% of revenue in the second quarter [83, 82, 46]. In the second quarter, operating cash flow was positive and the firm held cash. But adjusted EBITDA is not the same as cash to pay interest. Judging liquidity needs data on working capital, debt service, capital spending and covenant terms. The results release does not set those out. The filings neither show a liquidity event nor rule one out. In August 2026 the company also received a second Nasdaq notice for trading below the $1.00 minimum bid, with until 8 February 2027 to comply, and sued Google over its ad-tech conduct [84, 85]. They do show that the Outbrain–Teads deal took a goodwill impairment within a year of closing. That impairment was larger than a third of the price. Financing pressure has followed. ### The one clean set of books Dianomi is small and listed in London. It focuses on financial and business content. Its accounts give the clearest public view of the share of revenue a native vendor pays publishers. That is because its cost of sales is almost all publisher revenue share. In 2025 it paid £19.97 million to publishers on revenue of £27.4 million. That is a payout of 72.9%, leaving a gross margin of 27.1%. Adjusted EBITDA was a loss of £0.3 million. It earned 78% of revenue in the United States [86]. Two of its operating figures matter beyond its size. By its own total, impressions fell 14.1% in 2025. The company said this was partly due to "AI-generated and so-called zero-click summaries". Revenue per click rose 7.4% to 58p [86, 87]. The same release gives rounded Apple and non-Apple impression figures. They add up to a slightly different total and imply a fall nearer 15%. This paper keeps Dianomi's stated figure and notes the gap. In Apple News, Dianomi earned 22.5p per click, against 110.5p on its open-web publishers. That is a ratio of about 4.9. The ratio is descriptive. Advertiser, audience, geography, creative and buying mix all differ between the two placements. So it does not show that the placement alone causes the gap. ### The arbitrage loop, and who pays for it Recommendation widgets are the blocks of sponsored links that sit on publisher pages. They are documented as one channel that funds made-for-advertising sites. These are pages built to host ads rather than to be read. The joint trade-body definition of September 2023 makes "a high percentage of paid traffic sourcing" a core trait. The ANA's report names content recommendation platforms, along with social networks, as the channels that source those visits [88, 89]. It works as a loop. An arbitrage operator buys clicks from a widget. It sends readers to an ad-heavy page. Programmatic brand budgets pay for the ads there. The operator earns more from them than the clicks cost. > **Illustrative arbitrage economics, not a measurement.** > A media auditor's op-ed put the loop in round numbers. Pay $0.01 for a click, and earn $0.02 from the ads on the landing page [90]. The same piece counted about $2.4 billion of 2022 Taboola and Outbrain revenue as click spend. Taboola's share of that includes Connexity's e-commerce revenue. The piece also counted about $1.6 billion paid to publishers as TAC. It set these against an ANA estimate of $10 billion a year flowing to made-for-advertising sites. Those totals cover all advertisers on the two platforms, not spend by arbitrage operators. So they size the channel, not the abuse. No opened source measures how much widget demand is arbitrage. And the head of Jounce Media has said the vast majority of made-for-advertising traffic comes from paid Facebook [91]. The best-documented case is Forbes. In March 2024, Adalytics showed three things about the www3.forbes.com subdomain. Per Similarweb estimates, it drew more than 70% of its readers from paid display ads on Taboola, Outbrain and other paid traffic sources. In one observed session, a 52-slide slideshow served more than 200 ads, against three to ten on a normal main-site article. And it had run ads since at least May 2017. Bid requests routed through a Prebid server cut the URL short, so buyers saw www.forbes.com. The verification tags present on the main site were not seen on the subdomain. Forbes shut it on 2 April 2024 after a Wall Street Journal inquiry. It called the subdomain an insignificant part of its business [92, 93]. Who pays the cost can now be answered. The reader pays in ad load: 201 impressions in a session is the product. The brand advertiser on the programmatic side pays in wasted spend. The ANA's 2023 study covered 21 marketers. It found that made-for-advertising sites took 21% of their impressions and 15% of their spend. Their CPMs, the price per thousand impressions, were 25% below other sites. And they passed standard quality screens for viewability, invalid traffic and brand safety [89]. The widget platform is paid on the way in. The exchange is paid on the way out. The publisher whose masthead lends the page its credibility is paid too. Each party was paid, which may explain how the subdomain ran for seven years. The record does not show why Forbes kept it. Whether that loop is shrinking is taken up in chapter 10, because the answer depends on who is measuring. Some marketers opted into the ANA's log-level benchmarks. Among them, the median share of spend on made-for-advertising sites fell from 10% in 2023 to 0.39% in the third quarter of 2025. On the ANA's separate cost-waterfall measure it was 1.1% in the first quarter of 2026, a quarter in which the panel grew from 54 to 86 marketers and the ANA named "AI slop" as an emerging sub-type [94, 95, 96, 97, 98]. That result covers only the marketers who took part. The top quartile of that same group still had up to 27% of web spend on such sites in the third quarter of 2025 [97]. No market-wide figure after mid-2023 could be accessed. > **The economics in one paragraph.** > A native platform's gross margin is set by what it must promise publishers. The largest platforms promise a lot. They promise exclusivity and multi-year terms. They offer guarantees that ran at 15% of supply cost in 2025. In Yahoo's case, the promise was a quarter of the company. The demand that pays for those promises is mid-market and fragmented. Part of it is arbitrage that lands readers on pages built to be monetised rather than read. Its size is unmeasured. The platform earns on the click either way. That is the incentive structure the rest of this paper keeps coming back to. ## 5. The reader and the creative *Several studies found that most readers did not know certain native formats were ads. The studies ran between 2014 and 2018. This review found no representative US replication since, and two recent leads arrived too late to screen.* > **In short.** > This paper opened peer-reviewed and regulator studies. In three studies of sponsored articles and in-feed ads, recognition as ads ran from 7% to 37%. The rate depends on the measure and the ads shown. In one of them, conventional ads got 81%. In the FTC's lab, about 47% recognised the ads, and 68% with its recommended labels; only that study tested a recommendation widget. Better labels and other conditions gave higher rates. In the studies reviewed, better labels helped but did not close the gap with conventional ads. Disclosure helps readers spot the ad but lowers credibility. A meta-analysis pools the results of many studies. One, covering 278,791 people, finds no significant average link to what they say they plan to do. That is not evidence about actual purchases. This project could open just one randomised field test of click quality, native against display. It found in-feed native earns more clicks than display on the same page. But it draws less attention and lowers trust in the site. This paper's hypothesis is that the click edge and the recognition gap share a cause: looking like editorial content. That is not a tested finding. Native advertising is defined by one trait. It looks like the content around it. This chapter asks what that does to the reader. It uses original studies and systematic reviews. Work paid for by a vendor counts as what the vendor reports. ![Recognition was low in several studies, and the measure decided how low.](figures/fig-recognition.svg) *Figure fig-recognition. **Recognition was low in several studies, and the measure decided how low.** Recognition of native ads as advertising in the peer-reviewed and regulator studies opened for this paper, by study, task and stimulus. The 7–9% figures come from open-ended coding on article-style sponsored content; the 37% figure is a forced-choice question across 16 real native ads, one of them on Facebook; the FTC’s 47% to 68% combines its search and native lab conditions before and after its recommended disclosure fixes. These are dated study results, not an estimate of how many 2026 readers recognise native ads; no representative US replication after 2017 fieldwork was found. Units: percent of respondents. Scope: US samples, 2014–2018 fieldwork, article-style and widget native units. Sources [99], [100], [102], [103]. Caveat: Different measures, stimuli and samples; the points are not comparable as a series and are not pooled. Independent studies in olive; the FTC lab study in amber. Data: data/figure-data.json#recognition* ### How many readers know it is an ad The anchor study is Wojdynski and Evans. It was published in 2016 with fieldwork in 2015. Readers saw a sponsored article on a page styled like a newspaper. The study tested twelve disclosures. Readers then got open-ended questions, with no answers to pick from. Across all of them, 17 of 242 adult readers knew it was an ad. That is 7%. A follow-up eye-tracking study used 60 students. With the clearest wording, "sponsored by Dell", 18% knew it was an ad [99]. In a YouGov sample of 800 US adults, 9% spotted the ad. The sample was matched to the population. It was fielded in early 2017 and published in 2020 [100, 101]. Hyman and colleagues used a different measure. They showed 896 panel members 16 real native ads. The ads came from nine to ten publisher and social platforms. They asked a forced-choice question: paid, unpaid or don't know. Recognition was 37% for native, against 81% for conventional ads. Per ad, it ran from 21% to 72%, and 15 of 16 were below half. The 72% item was a Facebook ad, outside this paper's core scope [102]. The measure explains most of the gap between 7% and 37%. A forced choice gives credit more easily than open-ended coding. So any headline recognition rate must say which one it used. Labels matter, within limits. Shown on their own, labels with "paid" were read as ads by 83–89% of those asked. Labels with "sponsored" got 76–79%. Industry coinages such as "Brand Voice" or "Presented By" got 57–64%. In context, "Brand Voice" content on Forbes was recognised by 21–38%. Adding a prominent "Paid Ad" bar to two real native ads raised recognition from 40% to 56%. On one of them, the bar replaced the small "Sponsor Content" label. A third of readers still believed the content was unpaid [102]. Wording drove recognition in Wojdynski and Evans too. The labels "advertisement" and "sponsored content" got 12–13%, against 2–3% for "brand voice" and "presented by" [99]. Where the label goes is disputed. Wojdynski and Evans found a mid-article disclosure raised the odds of recognition five times. That was against the top-of-page spot the Federal Trade Commission recommends. And 90% of readers' eyes landed on a mid-article label, against 40% for a top one [99]. The FTC ran its own lab study, with 48 people. The fixes it recommends raised recognition by 21 percentage points across its search and native versions combined. That took it from about 47% to 68%. For the native versions alone, the gain was 23 points. A recommendation widget is the box of suggested links often shown under a story. The FTC noted that labels in the top-right corner of a widget were rarely looked at. It also noted that a significant share still did not know the ads were ads [103]. So advice on where labels go rests on a few small, old studies. And they disagree. > **What is missing from this evidence.** > Take open-web native units. Some run in feeds, some in recommendation widgets. For the years after the 2017 fieldwork, this paper found no representative US study of how often readers spot them. Smaller tests go on. A 2022 study of 600 US adults looked at disclosure. A 2024 study of 567 students looked at recognition. Neither is a representative replication [104, 105]. The first draft of this paper said no post-2022 work existed. The fact-check found these, and the paper says so. Two more leads came up after drafting and were not screened. One is a 2025 journal paper on disclosure and detection. The other is a March 2026 conference experiment. Its listing reports that fewer people spotted sponsored content in AI search overviews, despite labels [106, 107]. Ad literacy, the skill of spotting ads, may have risen. Eisend and colleagues report that the hit to brand ratings from disclosure has grown less negative over time. Jung and Heo say readers spot ads because they know platform ad tactics, not because of labels [108, 109]. The sources reviewed give no population estimate of ad literacy for the open-web native formats considered here. Two systematic reviews published in 2026 sort the studies but pool no results [110, 111]. ### What recognition does to the brand and the publisher Spotting the ad goes with harsher judgements. But these particular comparisons show links, not proof of cause. Chance did not decide who spotted the ad. So the gaps may partly reflect who spots ads. In Wojdynski and Evans, readers who knew the sponsored piece was an ad found it less credible. They liked the sponsor less, rated the story lower and were less keen to share it. In the second study, only the credibility penalty held up [99]. The publisher is the site that runs the ad. In the YouGov sample, spotting the ad lowered how credible readers found the publisher. It also lowered how they felt about it. The result was the same for legacy titles and digital-only ones [100]. The largest pooled evidence is a 2020 meta-analysis by Eisend, van Reijmersdal, Boerman and Tarrahi. It covers 61 papers, 57 datasets, 473 effect sizes and 278,791 people. The formats run from product placement to blogs to native. Disclosing sponsorship helps people spot the content as an ad (r = .255). It helps them see that the content aims to persuade (r = .257). It lowers brand attitude, or how much they like the brand (r = −.108). It lowers credibility (r = −.132). It shows no significant average link to what they say they plan to do (r = −.023 across 137,601 people). Results vary a lot from study to study [108]. The word "advertising" in a label raised both recognition and credibility. Disclosures placed after the content did the most harm to the brand [108]. Two things follow. First, disclosure goes with lower brand attitude and credibility. Its effect on actual behaviour is not proven either way. A stated plan is not a purchase. And a non-significant average is not proof of no effect. Second, there is no pooled estimate for native alone, because the meta-analysis mixes formats. The mechanism runs both ways. In the meta-analysis's path model, grasping the intent to persuade drives the attitude penalty. A 2023 experiment adds the opposite path. Disclosure can make the ad seem more transparent. That lowers persuasion knowledge, the reader's sense of being sold to. It also raises brand attitude and purchase intent [108, 112, 109]. A companion banner for the same brand raised recognition as much as a text disclosure did. Bad reactions were muted when the sponsorship seemed transparent [113]. "Disclosure hurts performance" is not a general law. Which path wins depends on how the ad is made. ### Clicks, attention and trust are three different things Two randomised studies matter for buyers. Randomised means chance decided who saw which version. Sahni and Nair ran field experiments on native ads in mobile search, with more than 200,000 users. They found no evidence that typical disclosure formats fooled users. But that is search, not the open-web widget [114]. Aribarg and Schwartz ran the only randomised test located that pits in-feed native against display on a news page. They held page position fixed on a news site. An in-feed native unit got a higher click-through rate than a display ad, because it looked like editorial content. The display ad drew more visual attention and higher brand recognition. With it, readers also rated the website more trustworthy [115]. The authors put the click edge down to looking like editorial content. The same look may also lower recognition. That is a reasonable hypothesis. If true, that edge would be the FTC's concern and the vendor's selling point at once. But the study does not test that path. Its effect sizes also sit behind a paywall this project could not open. Attention evidence points the same way in social feeds. A 2025 German eye-tracking study used a mock Instagram feed. A fixation is a pause of the eyes on one spot. Organic, unpaid posts drew 25% more fixations than sponsored ones, and about 250 milliseconds more dwell. Once readers fixated on a disclosure or a call-to-action, dwell time usually fell. They were then asked how they told an ad from a post. They cited product visibility most often (178 mentions) and the disclosure label second (95). Then came logos (81) and buttons (77) [116]. The first draft of this paper said readers ignored the label. The source says it was the second most-used cue. The FTC's lab found that better disclosures cut looking time by 21%. It warned that looking time is independent of understanding. So attention metrics cannot stand in for recognition [103]. Vendors have paid for attention and trust claims too. Funding is a conflict that lowers their weight. But the deciding problem with the two found here is that their methods are not disclosed. Outbrain paid for a Lumen study of 900 consumers in three European countries. It reported that native ads on premium news were "+31%" more likely to be trusted than social ads, and "+16%" more likely to be clicked. It disclosed no baselines, definitions or method. The report itself could not be retrieved [117]. It compares surfaces, news against social, not formats on the same page. So it neither confirms nor contradicts the field experiment. A 2015 vendor survey by Contently found 48% of readers felt deceived on realising content was sponsored. And 62% said a news site loses credibility when it runs native ads. The survey also found higher trust in the sponsor when readers judged the content high quality [118, 102]. ### What the units actually carry For recommendation widgets, consumer harm is not an abstract question. Researchers have measured what the ads contain. In January 2020, a crawl covered 6,498 mainstream news sites and 1,055 misinformation sites. The team coded 2,419 ads on 300 sampled sites. By the authors' codebook, 44.6% were "problematic". That category spans content farms and supplements. It also spans insurance and mortgage advertorials, investment pitches, sponsored search and misleading polls. Problematic in this sense is not the same as fraudulent. The sample is also a dated snapshot, not a full count. Of native ads, 87% were problematic, against 20% of display ads. Taboola served 61.1% of all problematic ads. Of Taboola's coded ads, 85.7% were problematic. For ads Google served, the figure was 16.9% [119]. On this measure, mainstream news and misinformation sites showed no significant difference. In a 2020 survey, raters judged 500 real web ads. Most raters called 20.6% of the ads clickbait and 11.2% deceptive. The four lowest-rated groups of ads were 43–72% native and recommendation ads [120]. Political clickbait rides the same units. Researchers crawled 1.4 million ads on 745 US news sites around the 2020 election. Of the political ads, 52% promoted "political news and media" articles. The recommendation network Zergnet served 79% of those. Taboola came next at 10%, then Revcontent at 5.7% and Content.ad at 1.8%. A smaller follow-up crawl of the same 745 sites ran in October and November 2024, over five weeks. It covered 15,110 ads. Again, it found clickbait political news ads common [121, 122]. Regulators met the pattern even earlier. In April 2011 the FTC sued ten operations that ran fake news sites. The sites used made-up reporters to sell weight-loss products. The sites drew traffic with "attention-grabbing ads on search engines and high volume websites". Together they had paid more than $10 million to advertise [123]. Two caveats carry weight. First, the 2020 crawl is a snapshot. It comes from before the industry's made-for-advertising crackdown. Chapters 4 and 10 describe it. A February 2026 preprint crawled 450 news sites from three locations over five months. It pins problematic advertorial ads mainly on Taboola and Outbrain. That suggests the pattern went on. But it is a preprint, and its coding differs [124]. Second, "problematic" is a broad codebook. The evidence shows a narrower point than a verdict on the category. Compare the recommendation unit with display on the same pages. The unit has carried a materially higher share of low-quality demand. That was true on one platform above all. One independent 2026 preprint, with a different codebook, suggests it still is. This review found no peer-reviewed replication in the sources screened by the cutoff. ### What the rules require The FTC's December 2015 guidance sets a performance standard, not a formula. The test is whether consumers know the native ad is an ad. It recommends "Ad", "Advertisement", "Paid Advertisement" or "Sponsored Advertising Content". It calls "Promoted" and "Promoted Stories" at best unclear. It wants the disclosure in front of or above the headline. It also rejects one disclosure for a mixed group of paid and unpaid items in a recommendation widget [2]. A policy statement on enforcement came with it. It cites consumers reaching fake news sites "by clicking on ads presented as attention-getting news headlines, which frequently appeared on legitimate news websites". It also makes ad agencies and affiliate networks liable [125, 2]. Chapter 10 sets out the European and UK rules. It also covers the 2023–2024 changes to rules on endorsements and reviews. Two points close the chapter. The FTC's advice is the top of the item. Wojdynski and Evans found that spot least effective. And the FTC's standard turns on whether consumers recognise the ad. But it sets no numeric pass mark. Compliance depends on each disclosure's wording, placement and overall net impression. The studies above show that few readers spotted the ads under several common disclosure practices of their time. They cannot settle whether any one unit running in 2026 is legal. ## 6. What the machines optimise *In the per-click campaigns the help centres document, the commercial unit is still a click. The objective has moved to conversions. That shift hands the platform, not the advertiser, the decision about which clicks to buy.* > **In short.** > Taboola's annual report says plainly how it ranks ads. It combines three inputs into a "relative value". Two are predictions: the chance of an interaction and the chance of a conversion. The third is the bid. Bids priced per click (CPC) and per thousand impressions (CPM) run in one auction. The system handles 500,000 recommendation requests a second. Papers by Taboola, Outbrain's Zemanta and Yahoo describe the same six-stage machine. They report lifts of 0.5–3% for model changes and 5–10% for exploration. The vendor measured each one itself. Taboola, Outbrain and MGID publish rules for bidding toward conversions. The rules differ in kind. Taboola needs 50 conversions in seven days before you can set a target cost per acquisition. Outbrain recommends seven to ten a day for its target mode. MGID expects ten to thirty over a learning period of one to two weeks. Recommended budgets run from five to fifteen times the target. The rules hold back low-volume campaigns. They do not lock them out of conversion optimisation. Chapters 3 and 4 followed the contracts and the money. This chapter opens the box between them. Inside is the software that picks which ad a reader sees and what the advertiser pays for it. ![The commercial unit stays a click; the machine chases conversions.](figures/fig-delivery-architecture.svg) *Figure fig-delivery-architecture. **The commercial unit stays a click; the machine chases conversions.** The six stages the opened documents describe. The platform decides which clicks to buy; the advertiser still pays for each click bought, so conversion risk stays with the advertiser unless a contract guarantees the outcome. Cells in olive are established by a current filing; cells in grey are established by vendor-authored papers as of their publication dates (2018–2022) and are inferred to persist. The loop underneath is the point: conversions feed the models, so the platform, not the advertiser, decides which clicks to buy toward a CPA target. Units: none (schematic). Scope: Taboola, Zemanta/Outbrain and Yahoo native systems as documented 2018–2026. Sources [1], [127], [128], [129], [130], [131], [133], [134], [136]. Caveat: A paper establishes deployment at its publication date, not in 2026. Stage boundaries are the paper’s inference from the documents, marked as such.* ### Six stages, documented by the vendors themselves The recommendation engine behind a native widget does the same job as any large ad system. The vendors have published enough to describe it. Taboola's 2025 annual report says how its system ranks each candidate. It combines three inputs into a relative value. The first is the predicted chance that a user interacts. The second is the predicted chance of a conversion after a click or view. A conversion is the action the advertiser actually wants. The third input is the bid. Its auctions run CPC and CPM pricing models at the same time [1]. The same filing reports 500,000 recommendation-related requests a second. That figure has stayed the same in each annual report since fiscal 2022 [77]. The filing also reports over 600 million people reached a day. Its peak figure is up to 1.2 trillion recommendations a month. And it cites four data centres that process over 170 terabytes a day to train its models [1, 126]. The papers fill in the stages. In 2022 Taboola said it served more than a billion requests a day. Each day it deployed hundreds of models that predict clicks, on separate traffic segments. Each teacher model trained from scratch for about two and a half hours. It used around 100 million impressions from the prior fortnight. Student models were updated in small steps. Each was fine-tuned on about 12 million fresh samples in twelve minutes. They were redeployed every four hours [127]. Taboola's 2018 paper was on deep density networks. It set the response budget at under 50 milliseconds. It defined the goal as revenue per thousand impressions at a fixed CPC. It split the algorithm into two modules. Exploitation uses what the model already knows. Exploration tests new options. The split lets it handle tens of thousands of new candidates a day [128]. Zemanta is Outbrain's demand-side platform (DSP), the software that bids for ad space for buyers. It reported more than a million bid requests a second. Its maximum response time was 100 milliseconds. Its click models computed over 600 million predictions a second on CPUs [129, 130]. Yahoo's native marketplace used a feature-enhanced collaborative-filtering model. It predicted the chance of both a click and a conversion for its auctions [131, 132]. Read together, the documents describe six stages. The first three are candidate retrieval, a click model and a conversion model. The fourth ranks candidates by expected value, with the bid. The fifth picks the creative after the auction. The sixth is an exploration module fed by incremental training. For Taboola, a current filing plus its papers show that design. For Zemanta and Yahoo, papers dated 2019 to 2022 show it [1, 127, 128, 129, 130, 131, 133, 134]. A paper shows what was deployed when it was published, not in 2026. The figure marks which is which. ### Every lift number in this chapter was measured by the vendor The technical papers are specific about what each part is worth. They are also all self-reported. Distillation is the teacher and student setup above. Taboola's pipeline lifted revenue per thousand impressions by 0.53% to 0.85%. That was across four traffic segments in a six-month test. A naive warm-start on fresh data cut revenue. So fresher is not always better [127]. Taboola's exploration module found 6.5% more new targets and 2.1% more advertisers. The cost was 0.05% of revenue. The deep model lifted revenue 2.9% over a tuned regression [128]. Zemanta documented a feedback loop. A DSP only sees outcomes for impressions it wins. So it under-buys what it under-observes. Zemanta countered it by exploring where the model was least sure. That raised revenue 5.0% and click-through 9.6% against control. Random exploration also raised revenue, by 3.2%, but it made the model worse. So part of the gain was simply bidding more [130]. Yahoo's post-auction stages report the largest numbers. Carousel asset optimisation gave +8.6% click-through and +4.3% revenue. Dynamic creative optimisation based on conversions gave a 53.5% lift in conversion rate. But the control served title-and-image pairs at random. So the added value over click-based selection is unknown [133, 134]. The same authors credit a later near-40% rise in carousel revenue to more traffic. That traffic came from extra advertiser demand. They take that as a sign advertisers were happy with the product. It is not a per-impression lift from the algorithm. It is a useful warning not to read product growth as optimisation. Clicks themselves are also noisy. Verizon Media studied accidental clicks with a short dwell time. It found they mislabel training data. Re-weighting them lifted revenue 1.18% [135]. The pattern is the point. Each performance claim for these systems comes from a live test. The vendor's staff ran it, and the vendor's paper reported it. No independent party has replicated any of them. None has an audited measurement. That does not make the numbers wrong. It makes them what the source hierarchy says they are: evidence of what the vendor reports. ### From clicks to conversions, and who holds the risk The commercial shift of the last three years is in the goal, not the unit. On Taboola, Outbrain, MGID and Yahoo, the advertiser still pays per click. But the platform now offers to turn your goal into per-click bids. The goal can be a cost per acquisition (CPA) or a return on ad spend (ROAS). The platform sets the bids with its own conversion model. Taboola's Realize offers three bid strategies. Maximize Conversions is automated and is the one Taboola recommends. A target CPA is optional. Enhanced CPC raises a baseline bid on high-value impressions and lowers it on low-value ones. Fixed Bid is the third. Taboola's documentation says the entered bid "will always be the price paid in auctions" [136]. Outbrain's Conversion Bid Strategy offers Max Conversions, Target CPA, Target ROAS and a semi-manual mode. Each is described as adjusting the campaign CPC [137, 138, 139]. Its Engagement Bid Strategy aims at on-site metrics from Google Analytics. The options are clicks, pages per session, session duration or bounce rate. It may raise the CPC to three times the entered bid [140, 141]. In June 2025, MGID announced CPA Tune. It keeps CPC pricing. But its algorithm "decides which clicks to buy based on predicted conversion probability" [142, 143, 144]. That last phrase describes a shift in who decides, not in who carries the financial risk. Under fixed CPC, you chose the bid. Under a CPA goal on a CPC unit, the platform's model picks which clicks are worth buying. But you still pay for each click bought, even those that never convert. MGID says so directly. The target is "a directional goal rather than a guaranteed cost per conversion", and "you continue to pay per click" [143]. Financial risk would move to the platform only under a different contract. It would have to guarantee the outcome or charge per conversion. The self-serve product documentation reviewed offers neither. But both large platforms' filings describe campaigns priced per acquisition, and Teads books some revenue on cost per incremental action. Neither discloses what share of revenue these carry or how incrementality is measured [46, 1]. You gain convenience and give up two things. One is control over what is bought. The other is the data that would show whether the platform's model is right. Fixed Bid and manual CPC are still on offer, so the choice is real. But the vendor guides steer buyers toward automation. The learning rules published with these products are the practical limit. They also converge. Realize's Maximize Conversions has a learning phase of two to five days. Before you can set a target CPA, it needs at least 50 conversions. They must come in seven days in a row. It recommends a daily cap of ten to fifteen times the expected CPA. The 50-conversion rule applies only to setting a target. You can use Maximize Conversions without meeting it. Some campaigns lack the Taboola pixel or server-to-server tracking. For them, the strategy aims at page views and impressions instead [145]. Outbrain's Target CPA aims for seven to ten conversions a day. It wants daily spend of at least seven times the target. It takes up to a week to learn [137, 139]. MGID's learning takes seven to fourteen days or ten to thirty conversions. It advises daily budgets of five to seven times the target. Its exploration phase bids higher until the first conversion, or until three times the target has been spent [142, 143, 144]. The rules come in three kinds. Taboola's threshold for a target is a hard gate. Outbrain's daily volumes and the budget multiples are advice. MGID's period is learning guidance. Together they mean a low-volume campaign learns slowly. It may never qualify for target-based bidding. Take an advertiser with three conversions a day. It gets 21 in a week, below Taboola's 50-in-seven-days threshold. It can run Maximize Conversions, but it cannot set a target cost per acquisition. How many advertisers sit below these lines is not published. The documented workaround is to aim at an upper-funnel event. That is a step that comes before the conversion. It eases the limit by optimising to a proxy. | Vendor and product | Learning phase | Conversion volume: gate or guidance | Budget rule | Sources | |---|---|---|---|---| | Taboola Realize, Maximize Conversions | 2–5 days | 50 conversions in 7 days before a target CPA (hard gate) | daily cap 10–15× expected CPA | [145] | | Outbrain, Target CPA / Target ROAS | up to 1 week | 7–10 conversions a day (recommended) | daily spend ≥7× target CPA | [137, 139] | | Outbrain, Max Conversions | wait 48 hours | ≥5 conversions a day (guidance) | not stated | [137, 139] | | Outbrain DSP (Zemanta), Conversions | not stated | 20–50 hard conversions, window not stated (recommended) | spend may exceed daily budget by 20% | [146] | | MGID CPA Tune, Target CPA | 7–14 days | 10–30 conversions over the learning period (learning guidance) | 5–7× target CPA daily | [142, 143] | | MGID CPA Tune, MaxConversions | 7–14 days | 10–30 conversions over the learning period (learning guidance) | 5–7× target; may exceed target by 2× | [144] | Two details from the documentation deserve a buyer's attention. First, take a "Maximize Conversions" campaign with no conversion tracking. It is optimising to page views. So the label and the objective can differ. Second, Realize's help centre states the price rule by strategy and route. Enhanced CPC cuts a winning bid to the second-highest bid. Creative Expedition charges the second price, even when it raises bids up to seven times. Fixed Bid charges the entered bid. Realize also sells ad space programmatically. That space clears at first price in private marketplaces and at second price on the open exchange [147, 148, 149, 136]. No primary source reviewed states one rule for all on-network inventory. For programmatic supply, the OpenRTB protocol signals the rule on each request [150]. The mechanism may differ by product, route and date. The public documentation does not say. That silence is itself the finding. A buyer paying per click cannot see from it how the price was set. ### Generative creative, and what it has been shown to do Creative automation is the part of the stack vendors talk about most and document least. Taboola's GenAI Ad Maker makes images and titles. It uses outside models plus its own layer trained on network trends. Taboola announced Realize+ on 23 April 2026 after a beta. It was entering a second expansion phase in the second quarter. It adds a decision engine that moves budget in real time. It also adds an element generator that makes and revises ads and targeting. The same release opened the platform to a first Claude Skill for campaign setup and optimisation [151, 152, 1, 153]. MGID has offered image generation, from text or from other images, since February 2024. It has also offered generated titles and descriptions in 29 languages since then [154]. MGID also scores a creative from poor to excellent before launch. The score uses its title, image and targeting. No accuracy figures are published [155]. Making ads, scoring them before launch and setting bids are different jobs. The vendors document each on its own. Teads markets a predictive pre-test built on eye-tracking and EEG data. Its product page lists "4 billion signals a minute" [156, 157, 158]. The one outcome number comes from a Taboola press release. It covers a university study of Realize ads. The release's own headline is that AI ads match human creative. In the raw data, AI-made ads had a 0.76% click-through rate. Human-made ads from the same advertiser, campaign and day had 0.65%. The data spanned hundreds of thousands of ads and 500 million impressions. Under the study's tightest statistical controls, the two did about the same. The release gives no number for conversions. It says only that AI visuals did not reduce downstream conversion performance [159]. The working paper behind it, by Exner, Hartmann, Netzer and Zhang, can be identified. The version this project opened, dated January 2025, has three points of note. First, ads with AI-generated images beat human-made ones on click-through only when they do not look AI-made. Second, it finds no significant effect of AI images on conversion rate for ads with a purchase objective. Third, it cautions that its data cannot isolate conversion effects [160]. Its sample totals differ from those in the SSRN abstract and in Taboola's release. So each figure is quoted with its version. The first draft of this paper read the raw gap as the result. The fact-check pass corrected it. Clicks in native units include accidental ones. So a click-through gap would not be a value gap anyway. Here is the summary that survives. Generative creative is shipped and widely used. A study on the vendor's platform found it no worse than human creative on clicks. Under some conditions it did better. Its effect on conversions is not established. ### Speed, and what "real time" means here The sources give latency figures for different stages. Latency here means response time. Taboola's engine responded inside 50 milliseconds in 2018. Its 2025 engineering blog puts the whole widget response at "hundreds of milliseconds". The blog also notes strict limits on database queries. Zemanta's budget for a bid reply was 100 milliseconds [128, 161, 129]. Taboola's rendering engine cut how long its scripts block the page's main thread. The cut was 485 milliseconds, about 70%. It also improved interaction-to-next-paint on publisher pages. That metric tracks how fast a page reacts to input. The gain was 6–36% at the 75th percentile, across four publishers. It did so without hurting click-through or revenue [161, 126]. No vendor publishes a current end-to-end budget. Model training happens offline. But in the systems the papers describe, some work still happens when the ad is served. That work is retrieval, feature computation and running the models. So the sources do not support treating the request as a simple lookup. For a buyer, the learning-phase rules above limit a campaign more directly. The millisecond figures do so less directly. > **What this chapter changes about how to read a vendor pitch.** > Here is what "AI-powered optimisation" means. A conversion model picks which clicks to buy toward your target. It uses the vendor's data. You still pay per click. And it runs under learning rules that slow low-volume campaigns and keep them out of target-CPA bidding. The published lifts for the core techniques are small. They are single-digit percentages, measured by the vendor. The larger claims include a 53% conversion lift for creative selection and 2.4× conversion efficiency. These are measured against random serving or against nothing at all. Chapter 7 asks what would count as proof. ## 7. What performance evidence proves *No public incrementality study of open-web native, with its design and results, was found. One practitioner account says a Taboola geo test was run and used to calibrate a mix model. It does not publish the test's own result. What the vendors publish instead uses designs that randomised tests show can overstate.* > **In short.** > Look at 663 treatment-control pairs from 563 large Facebook tests. Observational methods skip random assignment. They put median lift at roughly three to thirteen times the randomised median. The gap depends on method and funnel stage. It held even with 5,000 user features. Randomised tests are the standard. But even they are not precise. Across 25 large tests, the median confidence interval on return was over 100 percentage points wide. What native vendors offer as evidence is brand-lift surveys. Some compare those who saw an ad with those who did not. Others compare clickers with non-clickers. By this paper's reading, such designs fall in the weak, proxy class of the IAB's November 2025 guidance. In January 2026 Meta dropped its 7-day and 28-day view windows from its reporting API. Shorter reported windows cut attributed conversions. It does not change what the ads caused. Any efficiency claim should say which window it uses. Chapter 6 showed the machines chasing conversions. This chapter asks what evidence would show they caused any. That is the question of incrementality: what an ad caused, beyond what would have happened anyway. The standard test uses a holdout, a group kept from seeing the ads to compare against. For open-web native, no such evidence has been published in a form a reader can check. ### Two facts every incrementality claim has to survive First, ad effects are small next to the noise in outcomes. So even large randomised tests are not precise. One study looked at 25 field tests run on Yahoo for US retailers and brokerages. They covered millions of customers and $2.8 million of digital spend. The median confidence interval on return on investment was more than 100 percentage points wide. The authors calculate that a test that tells you something can need more than ten million person-weeks [162]. "Experiment-grade" does not mean precise. Randomising supports an unbiased estimate when the test gets five things right. They are assignment, exposure, outcome measurement, attrition and spillover. A geo test runs ads in some areas and not in others. It needs similar areas too, and enough power to detect an effect. The label alone ensures neither. Second, observational methods do not match the randomised answer. That holds even with great data. These methods skip random assignment. Gordon and colleagues used 15 Facebook tests. The tests held 500 million user-experiment observations. They found such methods "often fail to produce the same effects as the randomized experiments, even after conditioning on extensive demographic and behavioral variables" [163]. Their 2023 follow-up covered 663 treatment-control pairs from 563 Facebook experiments. It used more than 5,000 user-level features. Its result is the figure that matters most in this chapter. ![On Facebook, observational methods put median lift at three to thirteen times the randomised median.](figures/fig-rct-vs-observational.svg) *Figure fig-rct-vs-observational. **On Facebook, observational methods put median lift at three to thirteen times the randomised median.** Across 663 large Facebook experiments, the median lift estimated by double machine learning was 83%, 58% and 24% for upper, middle and lower funnel outcomes against 29%, 18% and 5% from the randomised experiments; stratified propensity matching was worse. The ratio labels are ratios of medians recomputed from the paper, not medians of experiment-level ratios, and they are not a correction factor for any other platform. Every native vendor brand-lift study opened for this paper uses an exposed-versus-unexposed or clicker-versus-non-clicker comparison, designs this evidence shows can overstate. Units: median lift, percent. Scope: Facebook feed advertising, 663 US experiments; the method finding, not the magnitudes, transfers to native. Sources [164]. Caveat: Feed ads on one platform; the paper’s claim is about method bias, which the authors show is not fixed by richer covariates. Data: data/figure-data.json#rct_vs_observational* Take one method, double machine learning. For upper, middle and lower funnel outcomes, it gave median lifts of 83%, 58% and 24%. The randomised lifts were 29%, 18% and 5%. So the ratios of medians were 2.9, 3.2 and 4.8. Stratified propensity matching was worse still, at 6.0, 9.8 and 12.8. A ratio of medians is not the median of the per-test ratios. These are also Facebook results. They are not a correction factor for any other platform. The authors conclude they "are unable to reliably estimate an ad campaign's causal effect" without experiments [164]. The reason got its name a decade earlier: activity bias. Those who see an ad on a given day do more of everything online that day. So comparing those who saw an ad with those who did not overstates the effect. So does comparing before and after. In one Yahoo test, comparing before and after would have overstated the effect by 350%. The true effect was small and insignificant [165]. Neither fact is about native. Neither gives a bias multiplier for Taboola, Teads or MGID. What carries over is the problem of telling cause from selection. The selection that creates the bias is an algorithm choosing who sees the ad. That is what chapter 6 describes. ### What native vendors publish This project could open three sources on whether open-web native works. The first is Taboola's 2023 meta-analysis of Kantar brand-lift studies. They span Europe, Latin America and the United States. These surveys test whether ads shifted views of a brand. Taboola reports an average lift of 15% in favourability, against a 4.2% Kantar norm. The average lift in consideration was 6.7%, against 4.6%. For message association it was 87%, against 10.1%. Pixel and cookie tracking marked who saw the ads and who did not. Taboola gives no study count, sample sizes or intervals. The units are not stated [166]. Most of the studies included video. The same page says the favourability lift was nearly the same as Kantar's video-only norm, at about 15%. The fourfold headline compares against an all-digital norm. The second is Outbrain's 2018 "True Engagement" case. Outbrain commissioned Nielsen to survey 1,573 users, for an auto client in Germany. It reports +23% brand lift. The groups compared were "those who clicked on the page and those who did not" [167]. That is selection on the outcome. The third is Taboola's September 2026 Realize ID release. It claims "up to 2.4x increase in conversion efficiency". It does not define the metric. It gives no baseline, period, advertiser count or method. It uses the word incremental. But no holdout, test or measurement partner stands behind the figure [168, 169, 170]. These are evidence of what the vendors report. The IAB and IAB Europe issued guidance in November 2025. It defines incrementality as the causal impact of marketing against what would have happened without it. It states plainly that attribution and return on ad spend "show what happened, not whether marketing caused the result". Then it grades the methods. Experiments are strong. Model-based counterfactuals, which model what would have happened, are strong to moderate. Marketing mix models, which link sales to spend over time, are moderate to weak. Hybrid proxies are weak. These include baseline versus exposed analysis and new-to-brand share. So does incrementality as the platform reports it [171]. The native vendor designs above fall in the weak class. Measured, a measurement vendor, markets a way to run incrementality tests on Taboola. Realize's own explainer describes geo-holdout and audience-split tests. But no randomised control was found in any Taboola or Teads product, study or case [169, 170]. Randomised evidence on native formats outside the open web points the other way from the vendors' silence. Facebook's in-feed ads are native by format, and the same large experiments found positive median lifts of 29%, 18% and 5% by funnel stage. Sahni and Nair's randomised test of native ads in mobile restaurant search, with more than 200,000 users, found that the ads benefited advertisers, and that most of the extra conversions came from exposure, not from clicks on the ad [114]. That second point matters for chapter 6: per-click billing prices the click, while the effect in that test came from being seen. Neither study covers the open-web widget. > **The absence, stated precisely.** > This project found no published incrementality study for open-web native with design and results. That covers Taboola, the Outbrain and Teads lineage, and MGID. It covers randomised holdout, ghost-ad and geo studies. It covers recommendation widgets and in-feed units. A recommendation widget is the box of suggested links often shown under a story. The closest item is a growth consultancy's account. It says a Taboola geo test ran in February 2024 and was used to calibrate a marketing mix model. The client then raised its Taboola spend by nearly 300%, with what the consultancy calls stable blended performance. The account gives a last-click cost per acquisition of about £43 and a model-based one of about £13. It does not publish the geo test's own lift, its uncertainty, its design or the formats that ran, and the client is unnamed. That makes it limited practitioner evidence. It shows such tests are run and used. It is not a reproducible estimate of what native adds, and its model-based cost is not an experimental one [172]. Vendors may hold private tests. The finding is about the public record, not the products. It means a buyer who wants causal evidence has to produce it or get it under contract. Chapter 10 sets out how. ### What the experiments that exist say about the mechanisms native relies on Retargeting follows the visitor. It shows ads to users who have been to a site. Recommendation widgets carry it at scale. There are randomised tests of it. One ran with an online home-improvement store. Retargeting caused 14.6% more users to return within four weeks. A third of the first week's effect came on the first day. Second-week ads worked better when first-week ads had been shown [173]. Against that, a field test at an online travel firm looked at dynamic ads. These show products a user viewed before. On average they worked less well than generic brand ads. They stopped doing worse only when browsing showed a reader's preferences had narrowed [174]. In those settings, retargeting raised return visits. Showing the products a user had viewed was not better by default. Neither study measured incremental profit. Neither result extends to all retargeting. Differences between users matter more than averages. Tests at eBay in 2012 found brand-keyword search had no short-term benefit they could measure. Non-brand keywords had negative average returns. New and infrequent users responded. Frequent users, who drove most of the spend, did not [175]. A second study covered 288 US consumer-goods brands. Their TV elasticities were far below those in published studies. These measure how much sales move with ad spend. Marginal returns, the return on an extra dollar, were negative for more than 80% of brands. The authors list publication bias in earlier estimates as one possible source of managers' optimistic priors [176]. The lesson that carries over is the direction, not the size. Published and vendor benchmark lifts are biased upward. And the users a system finds easiest to reach are often the ones who would have bought anyway. Ghost ads show what a platform can do when it wants to measure cause and effect. Google's method finds the control-group users who would have seen the ad. For one retargeting campaign it measured +17.2% site visits and +10.5% purchases. It gives the same precision at a tenth of the cost of public service ad controls. It logs over 100 million predicted ghost ads a day [177]. The platform has to build the method. This review found no version built by a native platform. ### Windows, lag and the reported number Attribution windows decide what a platform reports. They moved in 2026. A window is the time after a click or view in which a conversion counts for the ad. Google Ads' default click-through window is 30 days. Its default view-through window is one day. A view-through conversion comes from someone who saw the ad but did not click. These count only where at least half the ad was on screen for a second. They exclude users who interacted with any other ad. They sit outside the main conversions column. And they cannot be reported where cross-site cookies are blocked [178, 179]. Meta deprecated the seven-day and 28-day view-through windows in its Ads Insights reporting API. The change took effect on 12 January 2026. The sources opened do not say campaign attribution settings changed that day [180]. When a platform shortens its windows, attributed conversions fall. What the ads caused does not change. Meta's change says nothing about Taboola's reporting. Taboola's published defaults are a 30-day click window and a 24-hour view window [181]. The point is general. Take the September 2026 Realize ID release, which states no window or baseline. A claim like that cannot be checked against either kind of change. Conversion lag is real. It differs by channel. Lag is the delay between the ad and the conversion. Retargeting gave a third of its first-week effect on day one [173]. Haus pooled hundreds of geo tests from the 2025 holiday period. It found 41% of media's incremental value came in the post-treatment window. That is the period after the test ended. But Haus is a measurement vendor, and the studies are its clients' [182]. The MRC's Outcomes and Data Quality Standards, backed by the ANA, the 4A's and the ACA, cover outcome measurement across media. They require lookback windows to be disclosed before a campaign runs, and measurement providers to set limits on window length that data supports [183]. The IAB and MRC retail-media guidelines set the same kind of rule for retail media. The windows must be backed by data and disclosed before a campaign runs. Day-level lag data must be on hand [184]. Several native vendors publish their attribution windows. Among them, Taboola's click window runs 1–30 days and its view window 1–24 hours. Revcontent uses a 30-day pixel window. Nativo's windows can be adjusted [181, 185, 186]. None of the sources opened for this paper shows the realised lag distribution behind reported conversions. ### Marketing mix models do not settle it either Buyers who cannot run tests turn to mix models. Both large platforms now publish tools to calibrate those models with tests. Calibrating means using test results to correct the model. Google's Meridian documentation states "there is no single formula to translate an experiment result into a prior". It warns that results from a different period may not transfer. It notes that a mix model's counterfactual is zero spend. Many tests instead measure reduced spend [187, 188]. Meta's Robyn treats agreement with tests as a third optimisation goal. Its documentation cites an Analytic Edge whitepaper. The paper puts the average gap from ground truth at 25% for models that were not calibrated [189]. Meta's October 2025 guidance cites two sources. In a study, calibrating with tests changed measured return by 29%. In a survey, 55% of those asked often get clashing results from different measurement tools [190]. The IAB's December 2025 guide tells marketers to use tests to calibrate models. It says to match the same outcome, region and window. It also says to repeat a test before treating its result as an anchor [191]. Take a mix model built on sales measured outside the ad platforms. It does not depend on any platform's window. Window choices matter only where attributed conversions feed the model. They can feed it as inputs or as calibration targets. The industry itself is not happy. Three in four marketers say their attribution, incrementality and mix-model approaches lack the speed, accuracy or trust they need. That is the finding of the IAB's State of Data 2026 report [192]. An often repeated figure says 71% of advertisers rank incrementality as the top retail-media metric. It could not be traced to the ANA release it is credited to. That release instead reports 71% naming sales conversion as their biggest goal [193]. The figure is not used here. ### What this means for the questions the paper set out The incrementality question asked when native creates new demand. It also asked when native harvests intent, or takes credit for conversions that would have happened anyway. On the public record, the answer is unknown outside the platforms and their clients. That is because no test that would answer it has been published for this channel with its design and results. The question on AI and agentic buying asked what truly autonomous buying can do beyond today's optimisation. Whatever it can do, the evidence for it will have the same problem until it includes a holdout. An algorithm that picks the readers most likely to convert is what selection means. The Facebook tests show how badly selection can inflate observational lift. They do not tell us by how much on any other platform. In practice, this leads to the measurement decision tool in chapter 10. The buyer's question decides the test. "Did this campaign cause sales" needs a randomised or geo holdout. It also needs enough volume to see a small effect. "Which of two creatives performed better" needs only a split. "Is the platform's reported CPA true" needs the window, the conversion definition and the lag data. The MRC's outcomes standard and the retail-media guidelines both call for the window to be disclosed before the campaign, and the lag data to back it. Native vendors document the first two in varying detail. None of the sources opened discloses the third. ## 8. Commerce and new discovery surfaces *Each native vendor points to commerce. None of them reports it as a line of its own. The systems next door are better documented. They show attributed sales, not incremental ones. That is, sales an ad gets credit for, not sales the ad caused.* > **In short.** > Taboola's annual report recasts the firm. The old frame was "a leader in native advertising". The new one is a performance platform that "addresses the limitations of native advertising alone". It cites 500 million product listings through Connexity. It gives no figure for e-commerce revenue. The last hard number is from 2020, the year before Taboola bought Connexity. Its revenue then was $176 million. Teads Holding Co.'s annual report names retail media only as a place it plans to grow. It names no revenue, product or partner in retail media or commerce. Its retail-media tie-up with Pentaleap exists as a July 2025 press release. No live retailer is named. Publisher commerce lines are set by affiliate terms, traffic and search-platform policy, not by ad tech. Google's spam policies exempt native-advertising pages that share content directly with readers, but treat paid articles with ranking-passing links, and third-party sections hosted to exploit a site's ranking, as violations [194]. BuzzFeed's affiliate revenue fell 6.9% in 2025 as partners changed their bonus terms. And 28% of its revenue depended on Amazon. In AI tools, OpenAI announced ChatGPT ads in January 2026 and began testing them on 9 February. By September they had a reported $1 billion annualised run rate. Perplexity retired its sponsored follow-up questions. And Taboola opened DeeperDive as a place to sell ads inside answers. DeeperDive is an AI answer engine that runs on publisher sites. This paper grew out of the Performance & Native playbook. The playbook argues that native moves in steps. First comes placement, then recommendation, then commerce, then outcome infrastructure. This chapter tests the commerce step. It checks it against filings, product docs and publisher accounts. Then it looks at the newest discovery surface, the AI assistant. ### What the native vendors sell as commerce Taboola's commerce story is a data story. Its 2025 annual report points to Connexity. Through it, publishers get access to over 500 million structured product listings. These earn money per click or per action. The report pitches the pairing of reader data with purchase data as an asset [1]. In 2021, Connexity cost about $800 million. It had itself bought Skimlinks the year before. At the time of the deal it had $176 million of revenue. It had 1,600 direct merchants. Its ex-TAC gross profit was $78 million [195]. Ex-TAC means gross profit after traffic acquisition cost. That is the money paid out to publishers for their traffic. Turnkey Commerce launched with TIME in February 2023. In March 2024, it was extended to the Associated Press. For a publisher's commerce section, it supplies the editors and the tech. It also supplies the affiliate links that make money. The terms are not disclosed [196, 197]. Skimlinks says it turns links into affiliate links across 48,500 merchants. It says it serves 23 of the top 25 publishers worldwide [198]. Its public FAQ says it shares revenue and takes "a small percentage" [199]. Its support documentation states a typical 75/25 revenue share: the publisher receives 75% of the commission the merchant pays, and Skimlinks keeps 25%. Individual terms may differ [200]. None of this shows up as a revenue line. In the readable parts of the 2025 annual report, e-commerce and Connexity revenue are not broken out [201, 1]. If Connexity had not grown at all since 2020, it would be about 9% of Taboola's 2025 revenue. That number illustrates what the gap hides. It is not a measurement. The filing is clear on one thing: where the firm now stands. Realize launched in February 2025. The filing calls it a performance platform that "addresses the limitations of native advertising alone". Realize supports formats that reuse display and social creative. It claims a $55 billion opportunity, based on internal and external industry data [1]. The largest native firm's 10-K no longer leads with native. Teads Holding Co. says even less. Its 2025 annual report names retail media as a planned growth focus. It also names it as a risk to its growth plan. Connected TV, in-app and AI answer environments are named alongside it. But the report discloses no retail-media or commerce revenue, product or partner. Pentaleap is not mentioned. The first-quarter 2026 report does not mention retail media or commerce at all. It says Teads makes money on CPC, CPM, CPV and CPA, with no commerce line [46, 202]. The commerce aim lives in a press release. On 24 July 2025, Pentaleap and Teads announced a real-time-bidding integration. It lets advertisers buy retailers' onsite sponsored-product ads through Teads Ad Manager. The release named no launch retailer. By the cutoff, no live retailer, volume or result had been published [203, 204]. ### What the adjacent systems show Native vendors point to retail media as the budget next door. They usually attach a forecast. Criteo, which calls its program the world's largest independent retail-media API program, tells a flatter story. Its Retail Media segment made $263.9 million of revenue in 2025, up 2%. That was about 13.6% of the group total. It made $259.7 million of contribution ex-TAC, about 22.1% of the group's. The fourth quarter was down 17% on scope changes with two clients. For 2026, the firm guides flat to up 2%. Criteo serves about 235 retailers. It defines offsite as ads "across the open internet". It does not disclose the split between onsite and offsite [205, 206]. Its Commerce Grid supply platform lists native among the formats it supports. It claims 60 or more DSP integrations [207, 206]. Among the sources retrieved, Amazon's docs are the clearest. They show how retail purchase signals reach publisher pages on the open web. Amazon Publisher Cloud is built on AWS Clean Rooms. It lets publishers plan packages of ad space and enrich their signals. The goal is to unlock Amazon DSP demand. The data never leaves Amazon. Dotdash Meredith is quoted mapping "intent-to-buy signals" across more than 1.5 million articles [208, 209]. In August 2023, Amazon said Sponsored Products would run on Pinterest, BuzzFeed, Hearst Newspapers, Raptive and Ziff Davis sites. They would run automatically, on advertisers' current cost-per-click bids [210]. They are served "Beyond Amazon" on third-party sites and priced per click. In mid-2025, two Amazon ad-tech vendors reported a new setting for off-Amazon campaigns. They said its default was "Maximize reach" [211, 212, 213]. Suppose the vendors have the default right. Then a rise in ads shown off Amazon, in native-like slots, may not mean more advertiser demand. It may reflect a supply-side default. No Amazon docs on the 2025 setting were found. Amazon Marketing Cloud is free for advertisers who qualify. It lets them query the platform's data in aggregate. The rules say plainly that uploaded signals cannot be exported [214]. A clean room is a shared space where firms can match data without handing it over. Access to one does not make an analysis causal. Take an exposed-versus-unexposed comparison: those who saw the ad against those who did not. Run it inside a clean room and it is still just that. What a clean room does is allow a proper design, where the platform supports one. None of the nine native vendors profiled in chapter 9 documents a clean room in the public material this review opened. ### Content-to-commerce at the publisher Publisher filings show what commerce is worth to the firms whose pages carry native units. The answer varies by publisher. The reasons have nothing to do with ad tech. | Publisher | Line | Period | Value | Change | Share of base | Sources | |---|---|---|---|---|---|---| | People Inc. | performance marketing revenue | Q2 2026 | $68.8M | +13%; affiliate commerce +11% | about 24% of $289.9M digital revenue | [215, 216] | | BuzzFeed | affiliate commerce | FY2025 | $55.5M | −6.9%; Q4 −22.7% | about 30% of $185.3M total revenue | [217, 218] | | Future plc | B2C eCommerce affiliates | FY2025 (to Sept 2025) | £76.7M | −9% reported; H1 +10%, H2 −22% | 16% of B2C revenue | [219] | People Inc. was renamed from IAC in June 2026. It grew affiliate commerce 11% in the second quarter of 2026. Meanwhile its core sessions fell 22%, "due primarily to the impact of the growing prominence of Google AI Overviews" [215, 216]. BuzzFeed's affiliate commerce fell. It tied the fourth-quarter drop to "changes in supplemental bonus structures from our partners". About 28% of its 2025 revenue came from Amazon [217, 218]. Future's affiliate revenue rose 10% in the first half. It fell 22% in the second. For the full year, Future cites a 13% fall in unique page views, weaker consumer confidence, currency moves and closures. Future reports that Google Search's share of its audience fell from 67% in 2019 to 27%. It also reports AI Overviews reaching half of its key terms [219]. Three publishers, three directions, on similar traffic shocks. What they share is how the money is booked. Each of these lines is booked as attributed commissions. One partner can change its commission or bonus. That can move the line more than any change in how the placement performs. ### Conversion data in, aggregates out The playbook's commerce thesis needs data to flow both ways. Conversion data, the record of who bought or signed up, must flow into the native platform. Learning must flow back out to the advertiser. The docs show the first path is narrow. For the second, the public documentation reviewed describes aggregate reporting. It does not establish an advertiser right to export the platform's learned models or reusable audience data. Contracts may differ. The docs that come from Outbrain list the ways in. One is a browser pixel with 24-hour first-party matching. Another is offline conversion files. They are emailed to a service account. They are keyed to the click ID, the tag the platform gives each click. Only conversions within 72 hours of the click are used to optimise campaigns. Since November 2025, there is also a free Shopify app [220, 221, 222]. The system also takes a server-to-server postback, keyed to its click ID. A postback is a message sent straight from the advertiser's server to the platform. Order value, currency and order ID are optional fields [223]. These are legacy Outbrain settings. They are not a window for every current Teads product. Teads also publishes a server-side conversion API template for Google Tag Manager. It supports consent mode. That page does not document its endpoint, fields or deduplication [224]. Taboola's server-to-server path is a postback keyed to its click ID. It carries an event name, plus optional revenue, currency, quantity and order ID [225]. Taboola's own docs say server-side tracking covers attributed conversions only. That limits audience building and optimisation, compared with the pixel [226, 227, 228]. Meta's Conversions API is the reference design. It has server events, event deduplication, hashed customer IDs and offline events. These work for both measurement and audiences. By contrast, the native vendors' public docs mainly show a postback with a click ID and an event [229, 230, 226, 220]. On the public docs, the gap with a conversions API is identity. The native postbacks carry a click ID and value fields. They carry no hashed customer IDs. Parity is not shown. What integrations offer under contract may differ. What comes back out is aggregate reporting. Taboola's Backstage API returns campaign summaries by day, campaign, site, country, platform and item. It takes uploads of first-party audiences, by email or device ID. Outbrain's Amplify API exposes campaign and performance data. It is "available for a select number of partners by request". None of the public docs read offer a way to export audiences, click-level logs or model learnings [231, 232, 233, 234, 235]. That is not proof that no customer can get them. An advertiser keeps its own click IDs and conversion records. Contracts can grant more than help centres describe. On the public record, though, the model's learning stays with the platform. That is the dependence the research brief asked about. The advertiser's conversions train the platform's model. What the docs show coming back is a summary report. ### Affiliate incentives and the rules that bind them Content-to-commerce runs on affiliate commissions. The disclosure rules tightened in 2023 and 2024. The revised FTC Endorsement Guides took effect on 26 July 2023. Under them, a disclosure must be hard to miss and easy to understand. In interactive media it must be unavoidable. The FTC struck the word "small" from its affiliate-link example. So any affiliate pay needs disclosure. Staff guidance says "affiliate link" alone is not enough [236, 237]. The FTC's Trade Regulation Rule on the Use of Consumer Reviews and Testimonials came next. Adopted 5–0 in August 2024, it took effect on 21 October 2024. It bans fake or false reviews and testimonials. It bans buying reviews that express a particular sentiment. It also bans certain insider reviews that are not disclosed. It bans review sites a firm controls but presents as independent. And it bans suppressing reviews. Knowing breaches carry civil penalties. The rule does not ban AI-assisted writing as such, or all incentives. Incentivised reviews stay lawful if the reward does not hinge on sentiment [238, 239]. A publisher commerce section that ranks products by commission, not merit, is now exposed on two fronts. ### The newest surface: ads inside the answer Paid placement inside AI assistants fits the native definition most clearly. That is because the ad sits beside or inside the answer. In 2026, several outcomes arrived. They do not point one way. OpenAI announced plans for ads in ChatGPT on 16 January 2026 [240]. It began testing them on 9 February. The test covered logged-in US adults on the Free and Go tiers. Plus, Pro, Business, Enterprise and Education were excluded from the initial test. The ads were labelled and sat below the answer, apart from it. OpenAI said that ads do not affect answers [241, 240]. They gained a self-serve Ads Manager in May. By September they were reported in more than 40 countries. An OpenAI statement was reported on 1 September 2026. It claimed a $1 billion annualised revenue run rate, reached in under 200 days. It also claimed tens of thousands of advertisers. It listed product feeds, location targeting, custom audiences and outcome-optimised bidding [242, 243]. A run rate scales recent revenue up to a full year. It is not what was earned over a year. On 16 September OpenAI said it was testing "Sponsored Agents" with select US advertisers. After clicking an ad, a user can start a clearly labelled chat with an agent the business sponsors. OpenAI says that chat is kept apart from ChatGPT's own answers and from the user's original conversation [244]. Its help centre calls this a limited alpha and was not taking requests for early access. No technical documentation for advertisers was published [245]. Trade press named Wayfair as a pilot advertiser [246]. The revenue and advertiser figures are the seller's own, and they reach this paper through reporting. OpenAI's own pages were opened in the second review round, after an earlier access error; they describe the products but say nothing about how well they work. Perplexity had shown sponsored follow-up questions in the United States from November 2024. They were labelled and sat beside answers [247]. The firm phased the programme out in 2025. In February 2026 it said the programme would not return. In its words: "A user needs to believe this is the best possible answer, to keep using the product and be willing to pay for it" [248, 249]. One assistant found an ad business. Another concluded that ads and trust do not mix. No third party has measured either outcome. Research on how well users spot ads inside AI answers is thin. A 2024 study of users found they often missed product placements in generated search answers [250]. Work on machine detection shows sentence-transformer classifiers catch generated native ads with precision and recall above 0.9. The language models themselves struggle at the task [251]. No representative study was found. The native vendors are moving into the same surface from the publisher side. In June 2026 Taboola opened DeeperDive as an ad engine. DeeperDive is the generative AI answer engine Taboola runs on publisher sites. It puts ads into AI results pages. Taboola also offered the same ad model to other AI firms. The scale figures are Taboola's own. It claims tens of millions of answers a month, for more than 7 million users [51]. Teads has had an SDK in public beta since November 2025. It places native units inside chat interfaces. An MCP integration is listed as coming soon [252]. In March 2026 Dianomi announced a tie-up with Dappier. The plan is to embed a chat-style answer engine, with brand-agent ad units, on publisher sites. By mid-2026 it was still in talks with publishers [253]. Microsoft says ads in Copilot drew 73% higher click-through than traditional search between February and May 2025. That is a first-party figure from the vendor [254]. Amazon made sponsored-product prompts generally available in the United States on 25 March 2026. Clicking one can open a chat in its shopping assistant, Rufus. It was renamed Alexa for Shopping in May 2026 [255, 256]. None of this means all assistant answers carry ads. It means the economics of the recommendation unit are being rebuilt inside answer pages. That unit is the block of suggested links under an article. Google is moving in small steps. After May 2025, ads in AI Overviews reached desktop and entered AI Mode. In May 2026 the firm announced formats "instantly tailored to a person's unique query". It also expanded the Direct Offers pilot it had introduced in January 2026. It gave no uptake or revenue figures [257, 258]. The commerce rail is separate from the ads. OpenAI and Stripe co-maintain the Agentic Commerce Protocol, which is labelled beta. It powers Instant Checkout. That launched in September 2025. It came with the statement that results "are not sponsored, and ranked on relevance alone" [259, 260, 261, 262]. That statement came before ChatGPT Ads. OpenAI's 2026 commerce docs refer to separate rules for ads and checkout. Whether the two stay apart is open at the cutoff. > **Where the commerce argument lands.** > The playbook claims native is becoming commerce infrastructure. That is a strategy, not yet a market fact. The Taboola and Teads Holding Co. filings reviewed report no separate, current commerce-revenue line. Dianomi is a small exception: it reported £0.5 million of affiliate revenue for FY2025, which is affiliate commission, not retailer advertising [263]. Commerce media grew about 18% in 2025 by the IAB/PwC count. Criteo's retail-media segment grew 2%, held back by scope changes with two clients. The two scaled native platforms report no retail-media revenue. Publisher commerce moves with affiliate terms and traffic. And on the public docs, conversion data flows into the platform and aggregate reports flow back; the platform's learning does not. The AI assistant is where discovery and answers merge. In the same year it produced a reported success, a retirement and the native vendors' first answer-page products. Chapter 11 returns to it as an emerging surface, not a market. ## 9. The vendor field: who can document what *24 vendors scored on 12 dimensions. One summary index, built the way the author's Contextual Quotient is built. Nine deep profiles, each with a SWOT. And the wider universe they came from. The matrix measures what a vendor can document in public. It does not measure how well its ads perform.* > **In short.** > 24 vendors were scored on 12 dimensions. 58% of cells could be scored. Measurement has the lowest ceiling in the field: no vendor scores above 3 on it. No vendor documents holdout tooling of its own, and the one independent-partner route found, a measurement firm's page for Taboola, gives no design. Holdout tests keep a control group from the ads to show what the ads caused. On average, commerce and optimisation score lower. The Native Quotient, NQ, bands the 9 vendors with at least six scored dimensions: 1 in the front rank, 6 in Strong and 2 in Mid. The other 15 get a score but no band, because their records are too thin. Under four other weightings, positions move by up to 5 places, and 4 of the 9 banded vendors keep their band. Missing evidence is not neutral either: if MGID's commerce capabilities score of 1 were unknown, its NQ would rise from 62 to 66. The front-rank vendor keeps its band under all four tested weightings, though Teads Holding Co. joins it under one. The order is not a finding. ### How to read the matrix Each cell scores one capability from 0 to 5, capped by the evidence. A 5 means a filing, an official document or an independent observer shows in detail how the capability works. A 3 means it is real but thinly evidenced, narrow or in beta. A 0 appears only where a source documents that it is absent. If the record gives no basis to score, the cell says n/e. That is not a zero. If the dimension does not fit the business, the cell says n/a. One case is conversion data at a pure exchange, where it sits in the buyer's own buying tool. Conversion data is the record of sales, sign-ups and other actions that follow an ad. Each scored cell carries the reason for its score, an availability label, an evidence grade, a date and its sources. The web edition opens them from the cell. The long-format CSV in the appendix lists them all. Vendor marketing can earn at most a 3, and vendor documentation at most a 4. So a large help centre does not buy a high score. The coverage column shows how many of the twelve dimensions could be scored. That keeps a thin record in view. **The capability matrix** (0–5 evidence-bounded scores; n/e = insufficient evidence; n/a = does not apply; availability and evidence grade per cell are in data/capability-matrix-long.csv; rows grouped by peer group; banded vendors ordered by NQ, the Native Quotient, and thin records listed A to Z without a score) | Vendor | Group | NQ (band) | Supply | Formats | Workflow | Creative | Optimise | Conv. data | Measure | Transp. | Pub ctrl | Commerce | Interop | Geo | |---|---|---|---|---|---|---|---|---|---|---|---|---|---|---| | Taboola | core | 71 (Front rank) | 5 | 4 | 4 | 4 (beta) | 4 | 4 | 3 | 3 | 3 | 4 | 4 | 5 | | Teads Holding Co. | core | 65 (Strong) | 4 | 4 | 4 | 3 | 4 | 4 | 3 | 3 | 3 (beta) | 2 (unconfirmed) | 3 | 5 | | MGID | core | 62 (Strong) | 4 | 4 | 4 | 4 | 4 | 4 | 3 | 3 | 2 | 1 | 3 | 3 | | Dianomi | core | 57 (Strong) | 4 | 3 | 4 | 1 (unconfirmed) | 1 | 4 | 3 | 3 | 2 | 2 | 4 | 5 | | Nativo (Life360) | core | 56 (Strong) | 3 | 4 | 4 | 3 | 1 | 2 | 3 | 2 | 4 | n/e | 4 | 2 | | Readpeak | core | 56 (Strong) | 3 | 4 | 3 | 3 | 3 | 3 | 3 | 4 | 2 | 1 | 2 | 3 | | Revcontent | core | 49 (Mid) | 3 | 3 | 3 | 3 | 1 | 3 | 1 | 2 | 2 | 2 | 4 | 2 | | AdNow | core | thin record | 2 | 1 | 3 | n/e | n/e | n/e | n/e | n/e | n/e | 2 | n/e | 3 | | Adyoulike | core | thin record | 3 | 3 | n/e | n/e | n/e | n/e | n/e | n/e | n/e | 2 | 3 | 3 | | Dable | core | thin record | 3 | 1 | 2 | n/e | n/e | n/e | n/e | n/e | n/e | 2 | n/e | 2 | | Ezoic | core | thin record | 2 | 1 | 2 | n/e | n/e | n/e | n/e | n/e | 3 | n/e | n/e | n/e | | Media.net | core | thin record | 3 | 1 | n/e | n/e | n/e | n/e | n/e | n/e | n/e | n/e | 3 | 2 | | Microsoft Advertising native (audience ads) | core | thin record | 3 | 1 | 3 | n/e | n/e | n/e | n/e | n/e | n/e | n/e | n/e | 2 | | NewsBreak | core | thin record | 2 | 4 | 3 | n/e | 3 | n/e | n/e | n/a | n/a | n/e | n/e | 1 | | popIn Discovery | core | thin record | 2 | 1 | n/e | n/e | 2 | n/e | n/e | n/e | n/a | n/e | 3 | 2 | | TripleLift | adjacent | 62 (Strong) | 4 | 3 | 4 | 3 | n/a | n/a | 3 | 2 | 2 | 4 | 4 | 3 | | Sharethrough (Equativ) | adjacent | 53 (Mid) | 4 | 1 | 3 | 2 | n/a | n/a | 2 | 3 | 3 | 2 | 4 | 3 | | Amazon Publisher Services Native ads | adjacent | thin record | 3 | 1 | n/a | n/a | n/a | n/a | n/a | n/e | n/e | 3 | n/e | n/e | | AppLovin MAX native | adjacent | thin record | n/e | 1 | n/a | n/a | n/a | n/a | n/a | n/e | 3 | n/a | 3 | 2 | | Google Ad Manager native styles | adjacent | thin record | n/a | 3 | n/a | n/e | n/a | n/a | n/a | n/a | 3 | n/e | 3 | 2 | | Google AdSense native formats | adjacent | thin record | 3 | 4 | n/a | n/a | n/a | n/a | n/a | n/e | 3 | n/e | n/a | 2 | | GumGum | adjacent | thin record | 3 | n/e | 2 | n/e | n/e | n/e | n/e | 2 | n/e | n/e | 3 | 3 | | Meta Audience Network native | adjacent | thin record | n/e | 3 | n/a | n/a | n/a | n/a | n/a | n/e | 3 | n/a | 2 | 2 | | StackAdapt (native channel) | adjacent | thin record | n/e | 3 | 3 | n/e | n/e | n/e | n/e | n/e | n/a | n/e | 3 | 3 | ### A quotient, and why not a sum The matrix has one summary column, the Native Quotient. It uses the method the author built for the Contextual Quotient in his contextual paper, so the two indices share a construction and a way of reading them. Their scores and bands are relative to different capabilities and samples and are not interchangeable. Sharing a construction does not validate the method. Adding up the twelve scores would be wrong. A sum counts each n/e cell as zero. So it punishes a company for a thin public record, not a weak product. It would also reward breadth over depth. Averaging only the scored cells is the opposite mistake. Take a vendor documented on four dimensions, all of them strong. It would beat a vendor documented on all twelve. Missing evidence would become an advantage. NQ does three things instead. 1. **It groups the dimensions.** Depth is what the engine does with an advertiser's goal. It covers optimisation, conversion data, creative tools, commerce and workflow. Reach is where it can act and through which routes. It covers supply, formats, interoperability and geography. Standing is whether buyers and publishers can check and trust it. It covers measurement, supply transparency and publisher controls. Each group is the mean of the scored cells inside it. So an n/e cell is skipped, not counted as zero. 2. **It weights the groups 45/25/30.** Depth counts most. That is because the buyer's question in this paper is whether a platform can pursue an outcome. Standing comes next, because the paper's central gap is proof. The weights are the Contextual Quotient's. They are a design choice, kept so the two indices read the same way. They were not re-derived for native. 3. **It shrinks the result toward the field mean.** A vendor with few scored dimensions is pulled toward the field average of 55. The pull grows as the record thins. This is a rule the author chose, not a statistical estimate: the twelve dimensions are different capabilities, not repeated measures of one thing. The strength is four extra field-average observations, a third of the twelve dimensions. A vendor with all twelve scored keeps three quarters of the weight on its own record; one with four scored keeps half. Below six scored dimensions the record is too thin to band at all, so those rows get a score but no band. The bands are set from the field mean. The field mean is the score of a vendor with exactly average evidence. Because every score is pulled toward it, even a perfect record does not reach 100: at the current field mean, a vendor scored 5 on all twelve dimensions would get NQ 89. Front rank starts at 67, Strong at 56 and Mid at 47. Scores below that are Narrow. The three bars under each score in the matrix are Depth, Reach and Standing, in that order. Shrinkage damps a sparse record. It does not make missing evidence harmless. The build tests this directly: it turns each of the 168 scored cells into n/e, one at a time, and recomputes. Leaving out a low score can lift a vendor by more than the extra shrinkage takes away. If MGID's commerce capabilities score of 1 were unknown, its NQ would rise from 62 to 66. Across the field, one missing cell moves a score by up to 6 points up or 6 down, and 13 vendors have at least one cell that moves them by three or more. That is why n/e stays n/e and is never treated as zero, and why the coverage behind each score is shown next to it. ![Read the band, not the rank: one vendor clears the front-rank line, and most records are too thin to band.](figures/fig-nq.svg) *Figure fig-nq. **Read the band, not the rank: one vendor clears the front-rank line, and most records are too thin to band.** The Native Quotient for the vendors with enough evidence to band, grouped by peer group, built like the author’s Contextual Quotient: Depth, Reach and Standing group means combined 45/25/30 and shrunk toward the field mean by how little evidence backs each vendor. Vendors with fewer than six scored dimensions are listed under the chart, not plotted. Across four alternative weightings, positions move by up to 5 places and by 1.4 on average, and 4 of the 9 banded vendors keep their band under every weighting. NQ measures documented capability, not performance. Units: NQ points on a 0–100 scale, pulled toward the field mean; at the current field mean a vendor scored 5 on all twelve dimensions would get NQ 89. Scope: 9 banded vendors of 24 in the capability matrix, scores as of 27 September 2026. Caveat: Single scorer. The index inherits every limit of the dimension scores and is sensitive to the weights; the bands, not the positions, are the finding. Data: data/nq.csv* > **The honest limit of this number.** > Four other weightings were tested: equal thirds, Depth-led, Reach-led and Standing-led. Positions move by up to 5 places, and by 1.4 on average. The front-rank vendor keeps its band under all four, and Teads Holding Co. joins it under one. 4 of the 9 banded vendors keep their band, and the rest cross a band line under at least one weighting. The matrix footer names them. Two other shrinkage strengths, two and eight extra observations, move positions by up to 2 and 2 places. The missing-evidence test above moves single scores further than either. These tests cover the weights, the shrinkage and single cells. They do not test the evidence grades, the scorer or which vendors were sampled. So the matrix shows banded vendors by peer group and lists thin records apart, unordered. It does not claim that one banded vendor is better than the next. NQ is not calibrated to ad performance or buyer outcomes. All inputs are in the appendix CSVs. Anyone can rerun NQ with other weights. ### What survives every weighting **The top capability score goes to the platform moving fastest beyond the recommendation unit.** That unit, or widget, is the box of suggested links often shown under a story. Taboola has the highest Depth and Reach in the field. Most of those scores come from Realize. It brings conversion bidding, server-side conversion feeds, and native video and carousel formats. Its display formats are not counted as native. Chapter 3 shows the company branching out from the widget. It is moving into display, and into AI answers through DeeperDive. Chapter 11 shows its revenue guidance falling while profit guidance rose. So the strongest documented native capability belongs to the company building most beyond the widget. Its filings do not say how much of its growth those new lines carry. **Standing is where the field is thin.** Measurement support never scores above 3. It is unscored for two-thirds of vendors. Supply transparency is scored for fewer than half. Revcontent has the lowest Standing of the profiled vendors. It names no third-party verifier. The platform measures its own results. Dianomi shows how much a thin reading costs. Its first scores were low. They rose once the verifier read its help centre in full. There the verifier found conversion tracking, DoubleVerify and IAS integrations and publisher-level reporting. Readpeak scores well on Standing for a small company. That is because it lists all its publishers before a campaign runs. The two large platforms do not. **The exchanges hand the outcome to the buyer.** TripleLift and Sharethrough score on supply, interoperability and transparency. But conversion data and outcome bidding sit in the buyer's DSP, the tool the buyer uses to buy ads. So those cells are unscored or low. TripleLift's Depth comes mainly from commerce units and its deal tool. That describes their model. It is not a defect in it. **The large platforms' native products are thinly documented as native.** Google, Microsoft, Meta, Amazon and AppLovin each have three or four scored dimensions, too few to band. That is because their public material describes native as a format inside a much larger product. Their NQ sits near the field mean because of how the score is built. It says little about them. ### Deep profiles and SWOT Nine companies were profiled in depth. Two are the listed, scaled platforms: Taboola and Teads Holding Co. Two are the independent networks MGID and Revcontent. Dianomi is the finance specialist under offer from Taboola. Readpeak is a European regional platform. Nativo is now part of Life360. The last two are the exchanges with native roots: TripleLift and Sharethrough. Sharethrough is now part of Equativ. Each profile covers what the company sells, who buys it and how it earns. It covers what the company depends on. It lists each capability with its availability and evidence grade. It gives the performance evidence and who paid for it. It also sets out limits, risks and fit. Each profile opens with a SWOT. Each line cites its sources and carries a tag. **Record** means a filing, contract, regulator document, study or independent report. **Vendor documentation** means the company's own help centre, specifications or product pages. **Vendor claim** means a performance or scale figure the company states. **Inference** means the author's reading of dated facts. Most opportunities and threats are inferences by nature. A gap in the public record shows up as a weakness only in the form "not documented publicly as of September 2026". It never appears as "cannot". The web edition lets a reader pick one quadrant and compare it across all nine. ### Taboola (active) *Taboola.com Ltd.* The largest open-web native business: it sells display and video through Realize, locks supply with long exclusive contracts and guarantees, and describes itself as more than native. **SWOT (2026-09-27).** The largest listed open-web native platform, selling display and video through Realize, with Yahoo and Microsoft its largest supply partners and 2026 revenue guidance cut in August. - **Strengths.** Filings record about 12,000 publisher partners at Q2 2026 and a 30-year exclusive agreement to power native advertising across Yahoo's properties. (record) [47, 1] FY2025 revenue was $1,912.0M, net income $42.3M including a $20.1M one-time tax benefit, and company-defined free cash flow $163.4M. (record) [1, 348] Conversion-goal bidding is documented in detail, including a target cost-per-acquisition mode, fed by a site pixel, server-to-server feeds, or AppsFlyer and Adjust app data. (vendor documentation) [409, 410] The ten largest advertisers were under 10% of 2025 revenue, and about 2,100 advertisers spending over $100,000 across four quarters produced 86% of Q2 2026 revenue. (record) [1, 47] - **Weaknesses.** Supply is concentrated: Yahoo and Microsoft were about 34% of 2025 gross revenues from advertisers on digital properties, and the top five properties about 44%. (record) [1] This review found no public study with design and results showing what Taboola campaigns cause; its 2023 Kantar brand-lift meta-analysis compares exposed and unexposed groups. (record) [166, 169, 172] Site controls work by exclusion only in the API, and a list of sites shown before a campaign runs is not documented publicly as of September 2026. (vendor documentation) [411, 296] A January 2020 crawl coded 85.7% of sampled Taboola ads as problematic, a broad category, and a February 2026 preprint linked problematic advertorials mainly to Taboola and Outbrain. (record) [119, 124] - **Opportunities.** If the Dianomi deal completes by end-2026, finance and business native supply joins Taboola, with part of the price contingent on Dianomi publishers accepting Taboola's standard terms. (inference) [31] If Realize keeps adding display supply such as NBCNews.com, added in August 2026, growth could come from outside the recommendation widget, a split the company does not disclose. (inference) [50, 49] If Taboola publishes a held-back control test with design and results, it could reach buyers who need causal proof, using region-based holdout methods it already describes. (inference) [169, 170] - **Threats.** If revenue at existing publishers keeps falling, down about $32.4M year on year in Q2 2026, more publisher-prepayment write-offs like that quarter's $12.2M could follow. (inference) [47] If Realize fails to attract enough advertisers to Yahoo's properties, the 10-K warns the 30-year exclusive deal may add less to long-term profit than expected. (inference) [1] If AI search answers keep cutting publisher visits, as Teads says they already have, Taboola's widget impressions could shrink, though its own filings do not discuss it. (inference) [46, 353] - **What is sold.** A self-serve and managed performance-advertising platform (Realize) that places sponsored native recommendations (primarily below-article feeds), display, vertical video, carousel, motion and app-promotion ads on publisher sites, apps and OEM device surfaces outside search and social. Publisher-side products: monetization widgets, audience-development (Newsroom, push notifications, Taboola News content distribution) and e-commerce content/monetization via Connexity. Advertisers buy on CPC or CPM with automated bid strategies aimed at conversions (MAX_CONVERSIONS, TARGET_CPA), enhanced CPC (SMART) or fixed CPC. - **Who buys it.** Performance advertisers (lead generation, online purchases, app installs, subscriptions) buying directly or via agencies; approximately 2,100 Scaled Advertisers (>$100,000 trailing-four-quarter gross spend) in Q2 2026 contributed 86% of revenue; the 10-K states most advertisers buy on insertion orders cancellable on 24 hours' notice and that a significant majority of revenue comes from Scaled Advertisers working directly rather than via agencies. Demand also arrives via DSP integrations (unspecified) and via Yahoo, which sells and bills its own advertisers on Taboola's network. - **How it earns.** Advertisers pay per click (and in some cases per impression or per conversion/product-listing action); Taboola records gross revenue and pays digital properties traffic acquisition cost (TAC) under either a straight revenue share or a guarantee model (greater of a revenue-share percentage and a guaranteed amount per thousand page views or fixed amount). TAC also includes impressions bought from real-time advertising exchanges and third parties, up-front payments and bonuses to publishers, and non-cash amortization of the Yahoo Commercial agreement asset (about $4.1M per quarter). FY2025: revenue $1,912.0M, TAC $1,214.9M (63.5% of revenue), ex-TAC Gross Profit $713.5M (37.3% of revenue), Adjusted EBITDA 30.2% of ex-TAC. - **Supply and data dependencies.** Supply concentration: Yahoo and Microsoft together about 34% of 2025 gross revenues generated from advertisers on digital properties; top five properties about 44%. Yahoo is both a supply partner (TAC to Yahoo $349.0M in 2025) and a billing entity for advertiser revenue ($201.6M, 10.5% of 2025 revenue; 16.6% of Q2 2026 revenue). Approximately 12,000 digital property partners as of Q2 2026 (approximately 14,000 stated for Q4 2025, approximately 11,000 for Q3 2025 and Q4 2024). Multi-year exclusive contracts with guarantees (guarantee cost about 15% of TAC in 2025). Data: code-on-page integration gives first-party contextual and engagement signals and first-party cookies; the 10-K states some cookies are third-party and that browser/OS changes (Safari, Firefox, Chrome, Apple) could restrict signals. Generative creative tools rely on third-party software (10-K). Cash and facility: $120.9M cash and short-term investments and $102.3M drawn on a $270M revolver at 31 December 2025. - **Differentiation.** Direct code-on-page integrations with publishers give first-party contextual and engagement signals and first-party cookies (10-K). Exclusive, multi-year publisher contracts (average term over two years at inception) including the 30-year exclusive Yahoo native agreement and a ten-year Microsoft relationship. Scale claim of over 600 million daily users and about 12,000 digital property partners, plus OEM device distribution via Taboola News (Samsung, Xiaomi). Realize extends the recommendation engine to display, vertical video, carousel and app formats with conversion-based bid strategies, positioning against DSPs and social rather than only native rivals. Commerce data via Connexity (over 500 million product listings; CPC/CPA product monetization in editorial). Financial capacity: positive net income and $163.4M FY2025 free cash flow, an aggressive buyback (share count reduced 18% in 2025 per CEO), and a pending vertical acquisition (Dianomi, finance). - **Risks.** Yahoo accretion risk is a stated risk factor: the partnership 'grants exclusive rights to native advertising on their platform' and may not deliver expected long-term financial benefits if Realize 'does not attract enough advertisers to Yahoo's platforms'; TAC paid to Yahoo ($349.0M in 2025) exceeded revenue billed through Yahoo ($201.6M). Revenue guidance for FY2026 was cut in August 2026 by roughly $76-106M versus May while profitability guidance rose; existing digital property partners' revenue fell about $32.4M YoY in Q2 2026; a $12.2M publisher-prepayment impairment was booked 'due to a decline in the expected future economic benefits of the underlying arrangements'. Press headlines (not opened in full) attribute the Q2 2026 revenue miss to a 'Google policy' headwind and report a roughly 30% one-day share-price fall; the nature and dollar impact of that policy could not be verified from filings or the release. Realize platform risk is a stated risk factor ('inherently risky'; success depends on market acceptance and user experience). Exclusivity with publishers may not be renewable on similar terms (10-K); loss of exclusivity would let publishers add competing platforms. Guarantee exposure: 'A significant amount of our revenue comes from contracts ... where we are obligated to pay a specified minimum guaranteed amount per thousand impressions'. Brand-safety and quality reputation: Adalytics (Jan 2024) observed Taboola and Outbrain ads sourcing traffic to made-for-advertising sites; Apple News ad-quality complaints appeared in headlines in Feb 2026 (headline-level only). Signal dependence: reliance on data signals from sites Taboola does not control and on cookies; browser/OS changes by Apple and Google are cited as adverse (10-K). Workforce reduction of about 6% in April 2026 and reliance on founder-CEO (key-person risk factor). Israel-based R&D and incorporation expose the company to regional conflict and Israeli-law governance features (exclusive Tel Aviv forum) noted in the 10-K. Non-GAAP definition drift: ex-TAC Gross Profit definition changed in Q2 2026 to add back publisher-prepayment write-offs; comparisons across periods require adjustment. - **Best fit.** Direct-response advertisers (lead generation, subscriptions, e-commerce purchases, app installs) with pixel or server-to-server conversion feeds who can run MAX_CONVERSIONS or TARGET_CPA at sufficient conversion volume. Buyers seeking reach on Yahoo, Microsoft (MSN, Outlook), Apple News/Stocks and OEM news surfaces that are not available through most open-web DSP paths. US-weighted campaigns: about 48% of 2025 revenue and 54% of Q2 2026 revenue came from US-billed advertisers. Advertisers wanting site-level reporting and campaign or account-level site exclusion without a separate verification stack. Finance and business advertisers if the Dianomi acquisition completes (endemic finance/business publisher inventory to be brought onto Taboola terms). - **Poor fit.** Brand advertisers requiring documented third-party viewability or verification partners and incrementality measurement (not documented in vendor pages fetched). Buyers who need include-list (allow-list) targeting of specific publishers via API, or publisher-network blocking via API (not supported). Programmatic buyers expecting standard OpenRTB Native documentation, deal IDs or AdCP-based agentic buying (not documented; Taboola not listed as an AdCP member). Advertisers with low conversion volume who cannot feed automated bidding (Budget Allocator needs 'a few conversions per campaign per day'; Max Conversions requires the pixel). Buyers sensitive to appearing alongside content-arbitrage or low-quality landing pages, absent an explicit MFA/arbitrage policy in the vendor's public policy pages. - **Limits.** Supply is concentrated: Yahoo and Microsoft about 34% of 2025 gross revenues from advertisers on digital properties; top five about 44%. Publisher-count disclosures move materially between filings (approximately 11,000, 14,000 then 12,000 within three quarters) and are 'approximately' figures without a definition of a counted partner. Q2 2026 ex-TAC Gross Profit relies on a revised non-GAAP definition adding back a $12.2M publisher-prepayment write-off; underlying growth excluding it was about 4.7%. Publisher blocking via API is exclude-only and cannot block publisher networks; network-level blocking requires an account manager. No third-party verification partners, viewability standards or incrementality tooling are documented in the vendor pages fetched. OpenRTB Native / DSP integration details are not documented in the developer center; the 10-K statement is generic. Abby and Realize+ (Budget Allocator, Element Generator) are labelled beta in vendor documentation as of Feb-Jul 2026. Generative creative tooling depends on third-party software and Taboola does not verify IP status of generated content (10-K risk factor). Most advertisers buy on campaign-specific insertion orders cancellable on 24 hours' notice (10-K), so budgets are volatile. Minimum guarantees to publishers (about 15% of TAC in 2025; 'approximately 19% or less' in 2024-2025) create fixed-cost exposure if yields fall. | Dimension | Availability | Evidence | Observed | |---|---|---|---| | Supply access | live | A | Approximately 12,000 digital property partners (Q1 and Q2 2026 10-Qs; approximately 14,000 stated for Q4 2025 in the 10-K; approximately 11,000 for Q3 2025 and Q4 2024). Company claims reach of 'over 600 million people a day'. Largest partners Yahoo and Microsoft (each 5%+ of 2025 revenues; together about 34% of gross revenues from advertisers on digital properties). Apple is named among partners in the 10-K; Marketing Brew (Jul 2024) reported Taboola selling native ads in Apple News and Apple Stocks. Q2 2026 release cites 'the addition of Fox News'. OEM device surfaces via Taboola News (Samsung, Xiaomi named). TAC definition also includes 'advertising impressions purchased from real-time advertising exchanges and other third parties', and Realize is expected to shift more business to 'inventory where we bid for ad placements'. [1, 47, 540, 405, 541, 30, 542] | | Format support | live | C | Native recommendations (historically 'primarily in below-article placements'), plus Realize formats named at launch: display, native, vertical video, carousel, motion ads, app promotion; realize.com lists native, display, performance video, vertical, carousel and app-promotion ads. Yahoo Amendment No. 4 records that a 'Vertical Ads' format replacing 'Moments' had not reached general availability as of 1 Feb 2025. [543, 386, 1, 405] | | Advertiser workflow | live | C | Self-serve Realize platform with sign-up, campaign objectives and automated bidding; Backstage API for campaign management, reporting and automation; Abby conversational assistant ('opening Abby beta to all eligible advertisers', free, no enrollment, per help article updated 2 Feb 2026); Claude Skills 'conversational front door' announced Apr 2026; account-manager route for network-level blocking. 10-K: most advertisers place insertion orders per campaign and can typically terminate on 24 hours' notice. [386, 544, 545, 546, 1] | | Creative tools | live | A | 10-K: 'Generative AI tools to easily adapt existing creative or to create new ones from scratch'; risk factor states the generative ad-creation tool 'uses software tools provided by third parties' and Taboola does not verify IP status of generated content. Abby PR (Oct 2024): generative editing of images, backgrounds, CTAs and captions. Realize+ 'Element Generator' (beta) 'creates and continuously improves ads and targeting'. Backstage API documents an 'ai_disclosure' item field. [1, 547, 546, 231] | | Optimisation | live | C | Documented bid strategies: MAX_CONVERSIONS ('Recommended for new campaigns'; 'Requires that Taboola Pixel be installed'; cpc and publisher bid modifiers cannot be set), TARGET_CPA (auto-adjusts baseline bid to cpa_goal), SMART (Enhanced CPC) and FIXED (fixed CPC). Budget Allocator (Realize+) reallocates a campaign-group budget across 2+ campaigns; help article dated 28 Jul 2026 shows a beta badge, requires Max Conversions on all campaigns, pixel or S2S tracking, 'a few conversions per campaign per day' and a 2-3 day learning period. Minimum conversion thresholds and learning-phase rules for Max Conversions are not stated in the API overview. [409, 548, 231] | | Conversion-data access | live | C | Taboola Pixel (codeless conversions, Google Tag Manager, Shopify app, WooCommerce, WordPress, manual); server-to-server via postback URL or bulk endpoint; MMPs AppsFlyer and Adjust (Kochava, Branch, Tune, Singular 'with vendor assistance'); third-party trackers Voluum, ClickBank, RedTrack, Ringba, Finance Ads, Triple Whale; help center has 7 S2S articles, 3 Shopify articles and a DMP tracking collection; API conversion-rule and custom-audience references. Attribution windows, required click-id fields and rate limits are not stated on the overview page fetched. [410, 549, 231] | | Measurement support | live | C | Platform reporting by campaign, site, country, platform and day; real-time reports, change history, custom reports, pacing health score, report scheduling. Trust Center cites a 'dedicated global Anti-Fraud team' monitoring invalid traffic. No named third-party verification partner (IAS, DoubleVerify, MOAT) was found and TAG certification status was not checked in the registry; no viewability standard and no incrementality/holdout product were found in the vendor pages fetched. [550, 549, 551] | | Supply transparency | live | A | 10-K: 'All of our supply partners are directly connected meaning advertisers do not have to trace complex supply chains' and advertisers 'can easily select which properties they wish to appear on'. Site-level reporting is available in campaign reporting. However the same 10-K defines TAC as including 'advertising impressions purchased from real-time advertising exchanges and other third parties', so not all delivered supply is direct. [1, 550] | | Publisher controls | live | C | Advertiser side: campaign-level 'publisher_targeting' supports EXCLUDE only (INCLUDE not supported via API); account-level blocking available; blocking of publisher networks 'not supported via the API' (contact account manager). Publisher side: 'acceptance profiles' to determine advertiser types shown. Content review team 'reviewed over ten million items in the fourth quarter of 2025'; campaign quality and safety ratings; AI-assisted compliance; landing-page policy bans countdown timers, fake testimonials and requires company details, privacy policy and advertorial disclosures. No explicit made-for-advertising or traffic-arbitrage policy text was located in the fetched policy articles. [411, 1, 552, 549] | | Commerce capabilities | live | A | Connexity subsidiary (acquired 1 Sep 2021): 'over 500 million product listings' syndicated to publishers; CPC or CPA monetization of product listings and links to retailers 'directly within editorial content'; search-engine API and editorial tools for publishers. Connexity holdback compensation completed in 2025. 'Skimlinks' and 'Turnkey Commerce' are not named in the FY2025 10-K or Q2 2026 10-Q; their current product status is unconfirmed. Dianomi acquisition (pending) targets finance/business verticals rather than commerce. [1, 553, 31] | | Interoperability | unconfirmed | A | 10-K: 'We support the leading programmatic channels via integrations with leading demand side platforms, or DSPs' (no DSP names, no OpenRTB Native specifics). Developer center offers Backstage API, pixel/S2S, dynamic creative, mobile SDK, web integrations and Newsroom; an official Realize MCP server offers read and write campaign management (documented July 2026); no OpenRTB/DSP documentation found. Realize+ release: platform 'opening itself to Claude Skills'. Taboola is not among members named on the AdCP homepage (members directory did not render) and no Taboola AdCP announcement was found. [1, 544, 546, 375, 402, 403] | | Geographic reach | live | A | FY2025 revenue by advertiser billing address: United States $910.4M (47.6%), Germany $159.9M (8.4%), Israel $101.8M (5.3%), United Kingdom $77.8M (4.1%), Rest of World $662.1M (34.6%). Q2 2026: US $259.1M (54.3%, +17.5% YoY) while Israel, Germany, UK and Rest of World all declined YoY (non-US -11.1%). Sales hubs in Bangkok, London, Los Angeles, New York and Sao Paulo; data centers in the US, Israel, Hong Kong, Singapore and the Netherlands; localized Realize press releases in French, Italian, Spanish, German and Chinese. [1, 47] | Performance evidence: Campaigns launched with Abby go live 75% faster than manually built campaigns. (vendor assertion, sponsor Taboola) [547]; 80% of marketers would increase open-web investment if automated AI solutions were available; 86% would allocate up to 25% of performance budgets to the open web. (commissioned measurement, sponsor Taboola) [546]; Realize+ Budget Allocator delivers 'stable performance at the campaign group level' and time savings. (vendor assertion, sponsor Taboola) [548, 546]; Realize 'is working': 2025 revenue growth of 8.3%, Scaled Advertisers +6.2% and Average Revenue per Scaled Advertiser +2.4%. (direct record, sponsor Taboola (filing)) [1, 348]; No independent measurement of Realize or Taboola campaign performance (lift, incrementality, CPA vs alternatives) was found. (synthesis, sponsor none) [1, 386] ### Teads Holding Co. (active) *Teads Holding Co. (formerly Outbrain Inc.; Delaware; SEC CIK 0001454938)* The Outbrain lineage combined with Teads' video and CTV business; revenue fell in 2026 and guidance was suspended, and the filings do not isolate the recommendation business. **SWOT (2026-09-27).** Outbrain's recommendation business plus Teads' video and connected-TV sales under one listed company, with revenue shrinking in 2026, 10% secured notes and an unfinished product integration. - **Strengths.** Supply spans about 10,000 media owners reaching over 2 billion consumers a month, per the FY2025 annual report, and the top 20 partners average seven years' tenure. (record) [46, 46] 93% of FY2025 revenue came directly from advertisers, and about 50 joint business partnerships, signed or in process, averaged $4 million of annual spend. (record) [46, 46] Company-reported connected-TV revenue exceeded $100 million in 2025 and grew 67% year on year in Q2 2026, reaching 13% of quarterly revenue against 7% a year earlier. (record) [271, 29] Conversion bidding rules are published: help pages say three automated modes adjust the per-click bid and recommend 7 to 10 conversions a day for target cost-per-acquisition bidding. (vendor documentation) [138, 139, 412] - **Weaknesses.** Reported 2025 revenue growth of 46% reflects the Teads combination, while revenue on a combined-company basis fell 11.7% and Q2 2026 revenue fell 17% year on year. (record) [46, 271] FY2025 interest expense was $75.1 million against adjusted EBITDA of $93.4 million, and total debt had a carrying value of $614.5 million at 30 June 2026. (record) [413, 29, 271] This review found no public incrementality study with published design and results for any product, as of September 2026, and Outbrain's 2018 brand-lift case compared clickers with non-clickers. (vendor documentation) [167, 414, 415] Help pages say publisher and section reports appear once a campaign has gathered data, and no pre-campaign inventory list is documented publicly as of September 2026. (vendor documentation) [292, 416] - **Opportunities.** If buyers accept attribution without control groups, CTV Performance, declared generally available in July 2026, could sustain connected-TV revenue growth, 67% year on year in Q2 2026. (inference) [415, 271] If publishers adopt EngageOS, announced in June 2026 to put editorial recommendations and ads in one auction with Magnite demand, Teads could gain a role beyond the widget. (inference) [417, 271] If Walmart Connect access, which Teads says went live in June 2026, draws retail budgets, commerce could become a revenue line its filings do not yet mention. (inference) [418, 46, 202] - **Threats.** If AI answers in search keep cutting clicks to publisher sites, which the FY2025 annual report says has already affected revenue, sellable impressions could keep falling. (inference) [46] If unifying its performance technology within Teads Ad Manager, stated in May 2026 without a timetable, disrupts self-serve advertisers, volatile direct-response and small-business demand could weaken further. (inference) [346, 271, 419] If guidance stays suspended and the share price does not regain Nasdaq's $1.00 minimum by 8 February 2027, buyers that score vendor stability may delay commitments. (inference) [84, 271] - **What is sold.** An open-internet advertising platform (two-sided marketplace) selling advertisers placements on approximately 10,000 media owners' properties (publisher web/app, CTV homescreens and instream via OEM integrations such as LG, VIDAA, Samsung, TiVo), plus retail-media access (Pentaleap RTB to sponsored product ads; Walmart Connect) through Teads Ad Manager. Legacy Teads brought branding video (InRead), CTV and Teads Ad Manager; the Outbrain lineage brought self-serve recommendation/direct-response buying (Amplify, Smartlogic feeds, Conversion Bid Strategy) now marketed as 'Outbrain Direct Response (ODR), a subsidiary of the new Teads', and the Zemanta/Outbrain DSP which the vendor says is 'now part of the Teads offering'. A publisher-side product, EngageOS (unified feed OS with Magnite demand), was announced 2026-06-11. - **Who buys it.** Enterprise brands and agencies (93% of FY2025 revenue directly from advertisers; ~50 joint business partnerships averaging ~$4M annual spend; ~2,000 agency relationships; named examples Apple, LVMH, Stellantis, Warner Brothers, Dyson) and a Direct Response / SME segment (self-serve Amplify, CPC). Media owners (publishers, OEMs) are the supply-side customers under revenue share. - **How it earns.** Advertisers pay by CPC or CPM (and outcome-optimised buying via bid strategies); the company pays media owners a revenue share recorded as traffic acquisition costs (TAC). FY2025: revenue $1,300.5M, TAC $770.8M (59.3% of revenue, derived), other cost of revenue $100.6M, GAAP gross profit $429.1M, non-GAAP Ex-TAC gross profit $529.7M. Interest expense on 10% senior secured notes due 2030 ($75.1M in FY2025) is a material cash claim. - **Supply and data dependencies.** Depends on publisher/OEM partnerships described as 'exclusive, long-term and strategic' (no exclusivity share quantified), CTV OEM homescreen integrations (LG, VIDAA, Samsung Ads SEA, TiVo), third-party DSP demand (DV360, The Trade Desk, Yahoo for PMPs), Magnite demand for EngageOS, retail networks via Pentaleap and Walmart Connect, and its own Universal Pixel / household graph for CTV attribution. Revenue is EMEA-heavy (58.7% FY2025) and exposed to open-web headwinds in the Direct Response/SME business (cited when guidance was suspended). Litigation against Google (filed 2026-08-03) carries disclosed retaliation risk. - **Differentiation.** Scale on the open internet: FY2025 revenue $1.30B, ~10,000 media owners, >2B consumers/month, 20,000+ advertisers, 93% of revenue direct from advertisers (10-K). CTV homescreen supply via OEM integrations (LG, VIDAA, Samsung Ads, TiVo) combined with performance attribution (Universal Pixel + household graph) marketed as 'CTV Performance'; CTV revenue >$100M in 2025 and growing 50-67% YoY through Q2 2026. Two lineages in one platform: branding video/CTV (Legacy Teads) plus self-serve recommendation/direct-response with conversion bid strategies (Outbrain). Retail-media access from a non-retailer platform: Pentaleap RTB sponsored-product integration (2025) and Walmart Connect access (2026) inside Teads Ad Manager. EngageOS: an attempt to move up the publisher stack (feed OS + auction) with Magnite demand, distinct from a recommendation widget. - **Risks.** Integration and migration risk: unifying Outbrain performance tech into Teads Ad Manager while the DR/SME business faces 'open-web headwinds' (Q2 2026). Financial risk: leverage (10% notes due 2030), negative net income, guidance suspended, goodwill impairment of $352.1M within the first year post-close. Listing risk: recurring Nasdaq bid-price deficiencies; dilution or reverse split. Concentration in EMEA (58.7% of FY2025 revenue) exposes results to European macro and FX. Google litigation (2026-08-03): company itself discloses possible retaliatory actions, disruption and cost. Product-marketing claims (attention rankings, MMM rankings, 'first ever' retail RTB) lack visible methodology. - **Best fit.** Enterprise brands and agencies seeking omnichannel video + CTV homescreen reach on the open web with attention-based PMPs, especially in EMEA and APAC. Advertisers wanting one workflow (Teads Ad Manager) spanning open-web video, CTV, retail-media sponsored products and direct-response recommendation units. Publishers evaluating a feed-level yield system (EngageOS) rather than a widget bolt-on, if willing to join an early-access programme. - **Poor fit.** Performance buyers needing a stable, documented server-side conversion API and publisher-level transparency controls today (documentation found is thin). US-only buyers expecting US-scale supply parity with EMEA (U.S. is 27% of FY2025 revenue). Buyers requiring vendor financial stability as a procurement criterion during the current guidance suspension and listing deficiency. - **Limits.** Consolidated FY2025 results include Legacy Teads only from 2025-02-03; all reported YoY growth for FY2025 and Q4 2025 is inflated by the acquisition and is not like-for-like. Q2 2026 revenue fell 17% YoY and Adj. EBITDA missed the $14-22M guide at $7.0M; all forward guidance was suspended on 2026-08-06. Balance sheet: ~$628M principal of 10% senior secured notes due 2030 against ~$91M cash and short-term investments at 2026-06-30; FY2025 interest expense $75.1M vs FY2025 Adj. EBITDA $93.4M. Second Nasdaq minimum-bid-price deficiency (notice 2026-08-11); reverse split contemplated. Outbrain lineage products (Amplify, Smartlogic feeds, Zemanta DSP, Onyx) are in flux: Zemanta page is a redirect notice, Onyx URL 301s to teads.com, ODR is a 'subsidiary' brand; consolidation timetable undisclosed. No independent (non-vendor, non-partner) performance evidence was found for any product; all outcome claims are vendor-reported. A branded 'Conversion API' and any 'attention guarantee' with Adelaide could not be documented from opened sources. | Dimension | Availability | Evidence | Observed | |---|---|---|---| | Supply access | live | A | 10-K: ~10,000 media owners 'ranging from premium publishers to original equipment manufacturers', >2B consumers/month, publishers in >30 countries, 'exclusive, long-term and strategic relationships' (share of exclusive supply not quantified); CTV homescreen via LG, VIDAA, Samsung Ads (HK/TW), TiVo (5.3M households). Outbrain lineage adds recommendation placements plus push notifications, OEM and newsletter 'Platforms'. [46, 271, 419] | | Format support | live | D | InRead outstream video, full-page, AR/360 video, shoppable formats, CTV HomeScreen (incl. '3D CTV homescreen takeovers') and InStream, recommendation/feed units (Amplify/Smartlogic lineage), sponsored product ads via retail networks. [416, 554, 419, 555] | | Advertiser workflow | live | C | Teads Ad Manager positioned as 'end-to-end platform' with self-serve and managed service (stated for CTV Performance); Outbrain Amplify remains a separate self-serve CPC dashboard (my.outbrain.com/amplify) under 'Outbrain Direct Response'; CEO says performance technology is being unified within Teads Ad Manager (timing unstated). [415, 556, 419, 346] | | Creative tools | live | C | Teads Studio (creative transformation, AR/360, CTV overlays/interactive); DCO and generative-AI creative suggestions cited for Teads Conversions and CTV Ensemble; 10-K cites generative AI to broaden CTV participation. [414, 557, 558, 46] | | Optimisation | live | C | Outbrain Conversion Bid Strategy: Target ROAS, Target CPA, Max Conversions, Semi Manual; requires >=7 conversions/day (tROAS/tCPA) or >=5/day (Max Conversions). CTV Performance optimises toward 'qualified visits' with impressions served 'only when attribution is possible'. Attention-optimised PMPs (Adelaide AU) since 2023. [412, 415, 559] | | Conversion-data access | live | C | Outbrain pixel; 'Imported Conversions' via CSV emailed to a teads.com service account keyed on Outbrain click ID (dicbo), or 'server-to-server' (no endpoint documented in opened article); 72-hour post-click window for optimisation use; MMP integrations listed. Teads Universal Pixel + household graph for CTV. A branded 'Conversion API' was not found in opened docs. [220, 415] | | Measurement support | live | D | 30+ partner integrations claimed (Kantar, IAS, DoubleVerify, Lumen, Samba TV); Lumen CTV HomeScreen attention prediction model (exclusive access claimed) in US/EMEA/APAC/LATAM; Adelaide AU-based PMPs; TVision CTV attention; deterministic CTV attribution claimed via Universal Pixel. No independent validation of any measurement claim opened. [416, 560, 559, 561] | | Supply transparency | unconfirmed | D | Enterprise page cites Jounce Media 2024 '#1 Most direct supply path'; EngageOS promises 'buyer-level performance reporting with third-party verification' for publishers. No advertiser-facing publisher/site-level reporting or blocklist documentation was found in opened pages (help-centre programmatic and FAQ category URLs returned 404). [416, 417, 562] | | Publisher controls | beta | C | Media owners paid by revenue share (10-K). EngageOS (announced 2026-06-11) offers publishers a single auction combining editorial recommendations and ads, direct-sold and third-party demand via Magnite Demand Server / Prebid Server-compatible connections, session-yield optimisation; early access only. [417, 46] | | Commerce capabilities | live | C | Pentaleap RTB integration for Sponsored Product Ads across multiple retail networks via Teads Ad Manager (2025-07-24; retailers in the Teads integration not named); Walmart Connect Sponsored Products/Brands/Videos and Onsite Display in Teads Ad Manager ('now live', 2026-06-16); shoppable formats. [555, 418, 416] | | Interoperability | live | C | Inventory/PMPs buyable through Google DV360, The Trade Desk and Yahoo (2023 Adelaide PMP release); Zemanta/Outbrain DSP 'now part of the Teads offering' (standalone future unclear); Havas audience API integration listed on press page (not opened). OpenRTB Native support not documented in any opened page. [559, 563, 564] | | Geographic reach | live | A | FY2025 revenue: EMEA 58.7%, Americas 30.8% (U.S. 27.0%), Asia 10.5% (derived from 10-K); operations in >30 countries; marketing claims '50 markets'. CTV Performance beta covered US, EU5, APAC; GA is global. [565, 46, 414, 415] | Performance evidence: CTV Performance beta (automotive): cost per qualified visit $1-$2, average onsite engagement 2m21s across ~40 campaigns in 14 countries; Audi/Re-Mind PHD: 34,000+ engaged visits, 2m52s, 60% lower cost per quality visit vs target. (vendor assertion, sponsor Teads (vendor)) [415]; Travel Texas CTV Performance: 2,800+ incremental site engagements, 1m47s average duration, 4.4% increase in visits per viewable household over 30 days post exposure. (vendor assertion, sponsor Teads (vendor)) [414]; Citroen UK: 'over 5,000 qualified conversions directly from CTV'. (vendor assertion, sponsor Teads (vendor), advertiser quote) [557]; High-attention PMPs carry attention scores averaging 31% higher than Adelaide's Teads AU benchmarks. (vendor assertion, sponsor Teads and Adelaide (commercial partners)) [559]; Teads CTV HomeScreen averages ~5,300 Attention Per Mille, 173% above outstream video and 114% above YouTube. (vendor assertion, sponsor Teads (internal data) with Lumen model) [560]; Outbrain Direct Response: '+35% Time Spent vs. Social', '-90% CPA vs. Social', '-15% CPL vs. Social'. (vendor assertion, sponsor Teads/Outbrain (vendor)) [556]; Teads 'Ranked 1st for attention score' and '2nd Most effective MMM platform' across '132 MMM studies'. (vendor assertion, sponsor Teads (vendor) citing unnamed third parties) [554, 416] ### MGID (active) *MGID, Inc. (contracting entity in Advertisers TOS; data controller in Privacy Notice), 1149 3rd Street, Suite 210, Santa Monica, CA 90403, USA* A private, multi-region recommendation network with deep direct-response plumbing, unverifiable scale claims and undisclosed ownership. **SWOT (2026-09-27).** A private, self-serve, per-click open-web native network with detailed conversion tracking and a performance-marketing tilt, whose scale and results are not independently verified and whose ownership is undisclosed. - **Strengths.** Conversion data can flow back five ways, the help centre says, including direct server-to-server reports tied to each click's ID, webhooks and tracker integrations. (vendor documentation) [420, 421] Advertisers see traffic sources by domain or app, with page-category detail for MSN and Xiaomi, and can block up to 1,000 per upload, the help centre says. (vendor documentation) [422] Every advertiser must pass identity or business checks before funding, and invalid-traffic credits apply above 5% of billed clicks and from USD 5, the help centre says. (vendor documentation) [423, 424] Publishers get eight widget types plus Google auction demand inside MGID widgets, which MGID claims should lift revenue by at least 10%. (vendor claim) [425, 426] - **Weaknesses.** Ownership and revenue are not documented publicly as of September 2026, and its audience claims range from 850 million to 1 billion across its own pages. (vendor documentation) [65, 427, 428] Its public sellers.json lists 2,571 sellers across 2,522 domains, about 8% of the 32,000-plus publishers it claims, and the gap is not explained publicly. (vendor documentation) [429, 430] Independent research found misleading MGID-served ads on a Brazilian news site in 2024 and named MGID among organisers of conflict clickbait campaigns in a 2023 peer-reviewed study. (record) [431, 432] Ad-category exclusions for publishers go through an account manager, the help centre says, though MGID marketing promises 'full control of placements and ad content'. (vendor documentation) [433, 425] - **Opportunities.** If US news publishers keep signing, as Patch.com did in July 2025 and Mediaite in September 2026, MGID could widen its US news supply while rival independents consolidate. (inference) [434, 435, 58] If MGID publishes IAS viewability and brand-safety data from its April 2025 partnership, brand buyers would gain third-party evidence not public as of September 2026. (inference) [436] If more networks run on MGID technology, as Strossle does under the Azerion Native brand 'Powered by MGID', growth could come through partners, not only direct publisher deals. (inference) [437] - **Threats.** If Microsoft favours Taboola, which announced an Outlook and Microsoft 365 expansion in March 2025, MGID's MSN supply, signed in 2020 with no public term, could shrink. (inference) [39, 37, 422] If Google changes its Multiple Customer Management programme, the Google demand MGID brings into publisher widgets is at risk; Google is one of six intermediaries it lists. (inference) [426, 429] If brand buyers screen vendors on independent ad-quality findings, the 2023 and 2024 studies naming MGID-served clickbait could cost it budgets despite its TAG brand-safety seal. (inference) [431, 432, 438] - **What is sold.** A native advertising / content-recommendation network on the open web. To advertisers: self-serve (and managed) native placements sold on a CPC auction (CPM/vCPM/CPA defined in TOS but self-serve billing doc says CPC only; 'MGID does not offer a CPA model'), with automated CPA-targeted bidding (CPA Tune), push-ad and search-arbitrage campaign types, AI creative tools and an AI assistant (MIA). To publishers: native widgets (8 unit types), in-article video, IAB banners, an 'Exit Story' unit, MGID+ publisher tools, Google MCM Managed Inventory access, and content recirculation. - **Who buys it.** Self-serve performance and affiliate advertisers (help centre documents ClickBank, MaxBounty, AdCombo integrations; Brazilian checkout platforms; Voluum/Binom/Keitaro trackers; search-arbitrage feeds Tonic/Domainactive/System1), agencies via 'Super Client Account', and brand/direct advertisers via managed service (case studies: Bedrop, Beach Park, Checkfox; rich-media Ubisoft mention). Publishers ranging from news brands (MSN/Microsoft news suite 2020, RealClear, Abrams Media, Patch.com 2025, Mediaite 2026, Al Jazeera and detik.com in sellers.json) to long-tail sites meeting a 5,000-unique-daily-visitor threshold. - **How it earns.** Advertiser prepayments billed per click (CPC auction; minimum deposit $100, recommended $650; minimum campaign budget $50) with the publisher share paid out at thresholds ($100 PayPal/Tipalti, $1,000 bank). Revenue share percentages are not disclosed. Additional publisher-side revenue via Google MCM Managed Inventory (Open Bidding demand inside MGID widgets) and programmatic resale (sellers.json lists 2,571 sellers across 2,522 domains; OpenRTB 2.5 DSP guide). No audited revenue figure exists in the public record. - **Supply and data dependencies.** Depends on publisher widget placements (claims 32,000+ publishers; sellers.json lists 2,571 sellers across 2,522 domains) and on Google Ad Manager/Open Bidding for MCM demand. Fraud/quality depends on third parties: Pixalate (pre-bid sampled, post-impression), Anura, GeoEdge (2019 partnership), IAS (2025 partnership replacing Oracle/Moat), TAG Certified Against Fraud seal, iDenfy for advertiser KYC/KYB. Audience data: first-party click IDs, retargeting audiences, and Digiseg household-level cookieless segments in seven markets. Contextual targeting is marketed but not technically documented. - **Differentiation.** Broad multi-region office footprint outside the US (Ukraine, South and Southeast Asia, Southern Europe, Brazil) with local-language dashboards and regional policy pages. Deep direct-response plumbing: click-ID postbacks, webhooks, server-side GTM template, 14 named tracker integrations, affiliate-network integrations and Shopify native integration. Publisher stack beyond the widget: Exit Story, MGID+, Google MCM Managed Inventory (Open Bidding inside widgets), Marfeel analytics tie-in. Named-domain source transparency and blocklists for advertisers, including page-category breakdown for MSN and Xiaomi inventory. Documented IVT programme with TAG CAF seal, Pixalate pre-bid sampling, and a written credit rule (IVT >5% of billed clicks and >= USD 5). - **Risks.** Leadership transition (new CEO Oct 2025 from a CPA-network background) may sharpen the performance/affiliate tilt. Headline audience figures vary across its own pages and years (850M, 900M, over 1 billion), undermining the reliability of scale claims. Founding-year and HQ statements vary across MGID's own releases (2007 vs 2008; Los Angeles vs Santa Monica vs Beverly Hills). Ukrainian operational base (Kyiv office; CEO's military service) carries wartime continuity risk; no disclosure of mitigation located. Reliance on Google (MCM and Open Bidding; Google is one of six intermediaries in its sellers.json) and on third-party verification vendors for quality assurances. OpenRTB DSP guide contains a stray reference to 'Sovrn' in its tmax description, suggesting copied text and raising doubt about documentation currency. - **Best fit.** Performance and affiliate advertisers running CPC-bought, conversion-tracked campaigns across LATAM, South/Southeast Asia, Southern Europe, MENA and Ukraine-adjacent markets. Publishers in those regions or US mid-tier news/politics sites seeking a bundled native widget plus MCM/Open Bidding and recirculation tools. Buyers needing named-domain transparency and blocklists at the source level within a native network. - **Poor fit.** Brand advertisers requiring audited independent outcome measurement, incrementality testing or value-based (ROAS) bidding objectives. Publishers needing self-serve granular category/advertiser exclusion controls rather than account-manager requests. Buyers requiring CPA- or guaranteed-outcome pricing rather than CPC risk. - **Limits.** Self-serve billing is CPC only ('MGID does not offer a CPA model'); CPA Tune automates CPC bids toward a CPA goal but does not shift pricing risk to MGID. Publisher category controls are account-manager mediated, not self-serve, per the help centre. Conversion-value passing, ROAS objectives, attribution windows and incrementality tools are not documented in the pages read. sellers.json lists 2,571 sellers across 2,522 domains against a claimed 32,000+ publishers, about 8%; the programmatic-resale footprint is a small fraction of the claimed network or the file is incomplete. No audited financials, ownership disclosure or independent performance measurement exists in the public record. Supported campaign types include search arbitrage and push, and integrations skew to affiliate trackers and networks, which shapes the advertiser mix and brand-suitability perception. | Dimension | Availability | Evidence | Observed | |---|---|---|---| | Supply access | live | D | Vendor claims 32,000+ publisher partners / '32K+ content websites', 850M+ unique monthly visitors and 185B+ monthly recommendations (a separate product page says 900M users over 25,000 publisher assets). Named supply: MSN/Microsoft news suite (2020), RealClear Media, Abrams Media, Patch.com (2025), Mediaite (2026), Al Jazeera and detik.com (sellers.json). MGID's sellers.json lists 2,571 seller entries (506 PUBLISHER, 1 INTERMEDIARY = Google LLC, 1 BOTH), none confidential. Publisher onboarding threshold 5,000 unique daily visitors, 30+ articles of 500+ words. Inventory also reachable via OpenRTB (DSP guide) and a DV360 instruction article. [427, 430, 429, 39, 434, 566, 567, 421] | | Format support | live | C | Publisher units: Smart, Under-article, In-article (auto-placement, supports video), Sidebar, Header, Mobile, IAB Banners, In-Content Impact; plus Exit Story and recirculation. Advertiser formats: native widget ads, In-Content Impact, native video, rich media, Motion Ads, push-ad campaigns, search-arbitrage campaigns. OpenRTB guide covers banner, native 1.1/1.2 and video objects. Prebid adapters: 'mgid' banner+native; 'mgidX' display+video+native. [425, 267, 421, 568, 569, 570, 567] | | Advertiser workflow | live | C | Self-serve MGID Ads dashboard with mandatory iDenfy KYC (individuals) / KYB (companies) before funding; prepayment (card, auto top-up, wire); minimum deposit $100 (recommended $650), minimum campaign budget $50; creative moderation with Compliance Risk Levels (High/Medium/Low) and Ad Safety Rankings (High/Medium) gating premium/Tier 1-2 publisher rotation; agency 'Super Client Account'; API section in help centre; managed service implied by case studies. Final billing report within 10 business days of period end. [423, 269, 270, 571, 572, 573, 574] | | Creative tools | live | C | Documented: AI Image Generator (prompt-based text-to-image and image-to-image; 4 images per generation; model/provider not named), AI Editor for Creative Images, AI Text Generator for title and description, Motion Ads (static-to-motion), Perfect Crop, RSS-based feed creative import. Policy: AI content 'must not include misleading, deceptive, absurd, or unrealistic representations; deepfakes... are not allowed'. [421, 575, 576] | | Optimisation | live | C | Manual CPC bidding with CPC Multiplier, rules-based optimisation, source blocklists, CTR Guard, Performance Prediction, Reachmeter. CPA Tune (launched 2025-06-26; upgraded 2026-09-16) offers Target CPA and MaxConversions strategies that auto-adjust CPC bids using device, location and behaviour signals; requires active conversion tracking, $650 balance, daily budget 5-7x target CPA, 7-14 day learning phase. MIA AI assistant (2026-07-16) diagnoses campaigns and advises. Objectives documented: clicks (CPC), conversions/target CPA; no ROAS-, value- or incrementality-based objective documented. [421, 577, 578, 579, 580, 394] | | Conversion-data access | live | C | Five tracking methods: MGID Pixel (JS), Pixel via Google Tag Manager, Postback S2S using a per-click {click_id}, Webhooks, and Native Integrations (Shopify, Voluum); server-side GTM template; integrations with Voluum, BeMob, RedTrack, Binom, Adform, Adjust, Branch, ClickFlare, AppsFlyer, ClickMagick, Keitaro, Kochava, and affiliate networks ClickBank, AdCombo, MaxBounty; Brazilian checkout platforms (Braip, Kiwify, Perfect Pay, Cartpanda, Yampi) via webhooks. Passing of conversion value/revenue and attribution window: not documented in pages read (unconfirmed). [420, 421] | | Measurement support | live | C | Third-party impression pixels at creative level; IAS partnership (2025-04-25) for viewability, brand safety/suitability, IVT, attention, reach/frequency (pre-/post-bid not stated); IVT stack: proprietary filters on 100% of monetisable transactions, Pixalate (MRC-accredited) pre-bid sampled 1-100% by supply-partner risk, Pixalate/Anura post-impression on 100% of click traffic (widget only, not XML/API); TAG Certified Against Fraud seal (since Dec 2018, renewed annually, TAG ID d4c29acad76ce94f); IVT credits when IVT exceeds 5% of billed clicks and at least USD 5, 30-day lookback. 2020: billing only 100%-viewable native impressions (MRC guideline referenced); TOS vCPM = 50%+ visible. No independent measurement of MGID outcomes located. [421, 436, 424, 581, 428, 582, 573] | | Supply transparency | live | C | Advertisers: Sources Optimization shows traffic sources by domain/app name (replacing widget UIDs), with page-category/placement breakdown for MSN and Xiaomi OEM inventory; per-source statistics; campaign- and account-level blocklists (1,000 sources per upload). Public sellers.json with 2,571 named, non-confidential sellers across 2,522 domains (contact client.support@mgid.com). Publisher ads.txt configuration guide. Revenue-share percentages and take rate not disclosed. [422, 429, 583] | | Publisher controls | live | C | Category exclusions are available only 'by contacting your account manager' (no self-serve control documented); help centre asserts all ads are 'family-friendly and do not contain any explicit content'; Ad Safety Rankings (High/Medium) 'directly influence traffic filtering and inventory eligibility'; publishers can be blocked without notice for IVT with appeal. Marketing claims 'Full control of placements and ad content' and '100% transparency'. Payout thresholds $100 (PayPal; Tipalti with $26 fee) and $1,000 (bank); a NET-30 article exists (not opened). [433, 572, 424, 425, 584, 583] | | Commerce capabilities | unconfirmed | C | No native commerce/product-listing ad unit, product feed, or retail-media integration documented. Commerce touchpoints are tracking-side: Shopify native integration, webhook integrations with Brazilian checkout platforms, affiliate-network integrations, RSS feed creative import, and search-arbitrage campaigns (feed providers Tonic, Domainactive, System1). The native-advertising page lists 'Search & Promoted Listings' as a format without documentation. [421, 585, 567] | | Interoperability | live | C | Prebid.js 'mgid' adapter (supportedMediaTypes [BANNER, NATIVE]; endpoint prebid.mgid.com; native requires title, image, sponsoredBy; GVL ID 358; CCPA yes; schain 'check with bidder') and 'mgidX' adapter (display, video, native; Prebid Server yes; schain yes; GPP usstate_all; regions us-east/eu). OpenRTB 2.5 DSP integration guide supporting banner, Native 1.1/1.2 and video, second-price default, pchain field; cookie-matching guides for DSPs and SSPs; DV360 campaign instruction; Google MCM Managed Inventory with Open Bidding. GVL ID 358 is confirmed as MGID Inc. in the global vendor list (v178). [569, 570, 586, 568, 587, 588, 426, 589] | | Geographic reach | live | D | Claims 200+ countries/territories and 70+ languages. Twelve offices: Kyiv, Beverly Hills, Singapore, Sao Paulo, Gurugram, Mumbai, Milan, Rome, Berlin, Barcelona, Bangkok, Ho Chi Minh City. Regional signals: Spanish dashboard for LATAM (2017), Qubz exclusive MENA representative (2020), India publisher deals (Jagran 2019, Network18 2018), Brazilian checkout integrations, 'Taiwan Financial Vertical' policy, Digiseg segments in US/India/Germany/Greece/Poland/Turkey/Ukraine, MSN deal spanning US/APAC/Europe/MENA. US presence via Santa Monica HQ and US publisher deals (RealClear, Abrams, Patch, Mediaite). [430, 590, 591, 421, 592, 39] | Performance evidence: CPA Tune early adopters saw a '155% higher average conversion rate', with 'up to 300%' CVR growth in some verticals/markets, strongest in the US. (vendor assertion, sponsor MGID) [579]; Upgraded CPA Tune delivers 'up to 237%' higher conversion rate over manual optimisation, 39% lower cost per conversion; Target CPA +15% CVR / +20% revenue; MaxConversions +13% CVR / +25% revenue; based on 30,000+ tracked conversions across 13 verticals and 19 markets. (vendor assertion, sponsor MGID) [580]; Bedrop (German health and wellness) cut CPA by up to 76% and raised ROAS by 297%; RECORD publisher revenue up 335% via Exit Story. (vendor assertion, sponsor MGID) [593]; Google MCM Managed Inventory adds 'at least 10%' revenue uplift to publisher widget earnings. (vendor assertion, sponsor MGID) [426]; MGID delivers '96% real human traffic' (Dec 2018). (vendor assertion, sponsor MGID) [428] ### Revcontent (active) *RevContent, LLC* A US recommendation marketplace under private-equity majority ownership since 2021, still operating, with little public documentation of controls or scale. **SWOT (2026-09-27).** Private-equity-owned, US-centred recommendation network selling CPC and viewable-CPM widget placements, with tooling suited to hands-on direct-response buyers and scale and performance figures that only the vendor states. - **Strengths.** Advertisers can see widget IDs, adjust bids per widget, upload bid changes by CSV and keep campaign- or account-level blocklists, per the help centre. (vendor documentation) [439, 440] DSPs can bid on its inventory and SSPs can sell supply into it over OpenRTB 2.5, and Prebid.org lists a Revcontent adapter. (vendor documentation) [441, 442, 443] Documented placements reach beyond article pages into Apple News, email newsletters and in-app units, alongside below-article, mid-article, sidebar and homepage widgets. (vendor documentation) [444, 445, 446] Landing-page rules require a top advertisement label and a footer saying the page is not a news article, and ban fake-news formats and fake reviews. (vendor documentation) [447, 448] - **Weaknesses.** Conversions arrive by pixel or server postback, but no conversion, target-CPA or ROAS bidding objective is documented publicly as of September 2026, only CPC and viewable CPM. (vendor documentation) [449, 185, 450] Public help documentation names no third-party viewability, invalid-traffic or brand-safety verifier and claims no MRC accreditation; it is TAG Registered, but no current TAG certification is confirmed. (record) [451, 452, 453] Undated marketing pages claim 9,000+ publisher sites, while its public sellers.json listed 436 seller entries on 26 September 2026, a gap not explained publicly. (vendor claim) [454, 455] The 13 November 2024 advertiser agreement refunds only as ad credit at the vendor's discretion, requires invalid-click claims within 7 days and caps its liability at $300. (record) [456] - **Opportunities.** If the commerce, vertical-video and connected-TV phases announced with Mula on 9 March 2026 ship, the offer would reach beyond page widgets. (inference) [457, 458] If independent tests confirmed the '70% faster page loads than competitors' claimed on 9 March 2026, speed-focused publishers would have a reason to switch. (inference) [457] If publishers seek a widget supplier beyond Taboola and Teads Holding Co., formed from Outbrain's February 2025 combination with Teads, this operating private network is an option. (inference) [46, 367] - **Threats.** If the direct-response volatility and open-web headwinds Teads Holding Co. cited on 6 August 2026 are market-wide, a network whose documented tools target direct-response buyers faces them too. (inference) [271, 459, 450] If generative-AI search keeps cutting clicks to publisher sites, as Teads Holding Co.'s FY2025 10-K says it already has, the widget inventory it sells would shrink. (inference) [46] - **What is sold.** A content-recommendation / native advertising network: CPC (or vCPM) native ad placements in RevContent widgets on third-party publisher pages (below-article, mid-article, sidebar, homepage/section fronts, banner/IAB sizes, Apple News, email newsletters, in-app), bought self-serve or with account management; publishers get the widget as a monetisation product with a 'lightweight' page-speed positioning. Also sells its inventory to DSPs/SSPs over OpenRTB 2.5 and Prebid. - **Who buys it.** Advertisers: predominantly direct-response / performance and affiliate-style buyers (the documented workflow centres on CPC bids, widget-level bid CSVs, S2S postbacks to trackers such as Voluum, Maximus and Brax integrations, and a landing-page policy aimed at advertorial funnels); DSPs via OpenRTB. Publishers: US-centred mix from major news brands (Hearst, Tribune, CNN/CBS, NPR, The Guardian, Forbes, Washington Post appear in sellers.json or logos) to long-tail sites including politically partisan domains; local news (McClatchy, Boone, Nexstar cited) and newsletters. - **How it earns.** Takes a spread between advertiser CPC/vCPM spend and publisher payouts; publisher agreement defines Net Revenue as gross revenue net of taxes less adjustments of up to 10% for processing fees, with the revenue-share percentage not published. Programmatic resale (RTB via PubMatic/BidSwitch redirects) adds demand competition on widgets. Financials undisclosed (private). - **Supply and data dependencies.** Supply: direct publisher widget integrations (JS, GAM/AMP units, Apple News via GAM 360 + Apple Workbench, mobile SDKs incl. a React Native package, email CNAME setup); sellers.json lists 436 seller entries across 399 domains as of 2026-09-26. Data: first-party click/impression data via cookies, web beacons, web storage and device identifiers (privacy policy), served-impression counting, IP/UA; no documented third-party audience data, retargeting or lookalike products; targeting documented is geo/device/OS/browser/zip/metro/language only. Demand: vendor claims 90%+ first-party (direct) demand; remainder programmatic via DSP connections. - **Differentiation.** Page-speed / lightweight-widget positioning aimed at publishers (Core Web Vitals content, 'no feed units', repeated speed comparatives) rather than audience data. Granular, transparent widget-level bidding and blocking with CSV bulk tools — a direct-response buyer's toolset — plus S2S postbacks and tracker integrations (Voluum, Maximus, Brax). Multi-surface widget catalogue including Apple News (via GAM 360/Apple Workbench), email newsletters (CPC) and in-app, in addition to open-web placements. Two-way OpenRTB 2.5 / Native 1.2 connectivity and Prebid.js + Prebid Server adapters, so the network operates as both a buyer-facing self-serve platform and a programmatic supply source. Two-tier Brand Safety Seal that gates premium inventory (e.g., Apple News) by creative quality, and an unusually explicit landing-page policy (mandatory 'THIS IS AN ADVERTISEMENT…' footer, two-click rule, WCAG 2.2 contrast, auto-renewal disclosures). - **Risks.** Supply quality mix: sellers.json places premium news brands beside partisan and low-tier domains (breitbart.com, youngcons.com, secretsofthefed.com), and a 2018 headline records the company 'enlisting fact checkers' — brand-safety scrutiny risk for advertisers; the platform's own policy corpus (fake-news-format bans, 'THIS IS AN ADVERTISEMENT' footer) signals a history of advertorial-funnel demand. Advertiser terms are one-sided: refunds only as ad credit at RevContent's discretion, funds forfeited after one year of inactivity, invalid-click claims must be filed within 7 days with the burden on the advertiser, $300 aggregate liability cap, personal guarantee for credit terms. Publisher terms allow termination 'at any time for any reason, without notice'; revenue share is not published and net revenue may be reduced by up to 10% processing fees. Scale claims (9,000+ sites, 1B+ daily views, 130M+ daily users) are undated and unverifiable, and sit uneasily with a 436-entry sellers.json. PE ownership since 2021 with no disclosed exit path; leadership churn in the CRO seat (Welch 2021 → Hoy 2024); board addition (Waldshan, Feb 2025) not reflected on the team page at cutoff. AI positioning ('agentic OS', 'AI-driven creative optimization') outruns documented product (RevGPT headline variants, CPC/vCPM manual bidding) — execution risk and expectation gap. - **Best fit.** Direct-response and affiliate-style advertisers who optimise manually by widget/placement and run their own trackers via S2S postbacks. Publishers prioritising page speed and Core Web Vitals who want a lighter alternative or complement to Taboola/Outbrain-Teads widgets, including local news, newsletters and Apple News channels. Buyers wanting cheap open-web reach on US news inventory at CPC with no minimum commitment beyond $50/day. - **Poor fit.** Brand advertisers requiring third-party viewability/IVT verification, category-level inclusion lists by domain, or documented brand-safety vendor integrations. Advertisers needing audience, retargeting or lookalike targeting, or automated CPA/ROAS optimisation. Buyers needing video via the programmatic SSP path (not supported) or first-price auction mechanics. Non-US or non-Brazil markets where local support and inventory depth are undocumented. - **Limits.** No independent measurement of reach, performance, brand safety or IVT was located; every quantitative claim in this profile is vendor-originated except the sellers.json count, the Prebid adapter listing and competitor mentions in Taboola/Teads/Eva Live 10-Ks. Financials are undisclosed (private, PE-owned); the only money figure is a vendor statement of >$500M distributed to publishers in 2020 with no basis. Targeting documentation covers geo/device/OS/browser/zip/metro/language only; no audience, interest, retargeting or lookalike products are documented, and no conversion-optimised (CPA/ROAS) bidding objective is documented despite marketing language about ROAS and AI optimisation. Programmatic sell-side (SSP) path excludes video and does not support user sync or first-price auctions per vendor doc; TCF-EU and USP flags are 'No' on the Prebid adapter page although the privacy policy claims IAB TCF v2 vendor status (GVL ID 203). Thresholds conflict across vendor documents: 'no minimum spending requirements' (marketing) vs $50/day minimum budget (help doc); $250 payment threshold (publisher agreement) vs $100 ACH/PayPal and $1,000 wire minimums (help doc). Web search budget was exhausted before this profile; leads came from Google News RSS, PR Newswire company page, EDGAR full-text search and direct URLs. Business Insider (2021), Adweek (2021), MediaPost (2018), Sarasota Herald-Tribune (2018), CourtListener, Justia, FTC search, Google Scholar and several search engines were blocked or returned no usable results. | Dimension | Availability | Evidence | Observed | |---|---|---|---| | Supply access | live | D | Vendor claims 9,000+ publisher sites, 1B+ daily ad views and 130M+ daily users (undated marketing page) and 'hundreds of millions of unique users' (homepage). Structured record: sellers.json lists 436 seller entries (432 PUBLISHER, 4 BOTH, 0 INTERMEDIARY, none confidential, 399 domains) updated 2026-09-26, including washingtonpost.com, forbes.com, cbsnews.com, npr.org, theguardian.com, buzzfeed.com alongside breitbart.com, youngcons.com and secretsofthefed.com. Named exclusive/partner publishers in releases: Barstool Sports, Minute Media, Salon, McClatchy, Boone Newspapers, Money Times (Brazil), On3, Us Weekly, Brit & Co, Hearst Newspapers, CBS Interactive. [454, 367, 455, 594, 457] | | Format support | live (web/app/email/Apple News widgets, native video); announced (vertical video automation, CTV, smart commerce) | C | Widget catalog: Below Article (fixed height / infinite scroll / masonry / recirculation), Mid Article, Sidebar (desktop), Homepage & Section Fronts, RevMore (mobile, raised above below-article), Banner (IAB standard and custom sizes), Apple News (claims Apple Preferred Vendor Program approval), Email Newsletter, In-App (customisable JS). Labs widgets: Mobile RevToaster, RevExit, RevFlicker, RevShifter. Native video marketplace launched Jan 2020 (10–120s creatives). Campaign content accepts headline (80 chars), brand name (30 chars), images (JPG/JPEG/GIF/PNG/WebP) and video. Vertical Video, CTV and Smart Commerce formats described as future phases in the March 2026 Mula release. [444, 595, 596, 449, 457, 445] | | Advertiser workflow | live | C | Self-serve web platform: campaign name, bid type (CPC or vCPM), daily budget (minimum $50/day), pacing, conversion tracking, brand logo, scheduling/dayparting, targeting, UTM tracking, content; creative/landing-page approval 1–2 business days. Widget-level bid adjustments (per-widget or percentage) and CSV bid upload; widget blocklists at campaign or account level with a 'Widget Blocklists' tab (Jan 2024). Account API (OAuth 2.0 client-credentials, 24h tokens, api.revcontent.io) and third-party tool integrations (Maximus, Brax.io, Voluum, Google Analytics guides). Dedicated account managers offered. Credit card, auto-funding; advertisers on credit terms may be asked for a personal guarantee. [449, 439, 440, 597, 459, 598, 454] | | Creative tools | live | C | RevGPT: generative-AI headline variants from an approved creative ('Ask RevGPT' button), paired with the original image; expanded to 100% of advertisers (blog updated 2025-03-18, first announced late 2023); not available via API; model/provider and cost not stated; vendor says it 'consistently outperformed user-created alternatives' in testing without metrics. Native video upload (2020). No documented AI image generation. [599, 600, 596] | | Optimisation | live (manual CPC/vCPM); unconfirmed (conversion/CPA/ROAS auto-optimisation) | C | Documented bid types are CPC and vCPM only; the campaign-creation guide documents no auto-optimisation, conversion-goal, target-CPA or ROAS setting; the vCPM guide says the algorithm prioritises creatives by viewable CTR. Optimisation is largely manual (widget bids, blocklists, scheduling, one-device/one-geo/one-language campaign structure). Marketing pages claim 'AI-driven algorithms' and 'AI-driven creative optimization that tests headlines, images, and formats in real-time' and objectives of sales/leads/conversions/ROAS, but no product documentation for conversion-based bidding was found. [449, 601, 439, 454, 602] | | Conversion-data access | live | C | Pixel conversion tracking (HTML pixel on conversion page; 30-day attribution window; multiple pixels for funnel events, each with a value) and server-to-server postback (GET/POST to trends.revcontent.com/api/v1/conversion.php with api_key, rc_uuid click ID and integer 'amount'; optional user_ip/user_agent). Pixel and S2S must not be combined (duplicates). Conversion data appears in reporting; no documented use for automated bidding. [185, 450] | | Measurement support | live (first-party reporting); unconfirmed (third-party verification) | C | Impressions counted as 'served ads' (ad sent by ad server and downloaded/rendered/viewed), citing IAB/MRC guidance; viewability percentage, Ad RPM and Ad vRPM in publisher reporting; real-time dashboards and API stats for advertisers; Google Analytics discrepancy guidance. IVT filtering described generically with no named vendor; Brand Safety Seal has no named third-party verifier. TAG Registered since 2018 (TAG-ID in its sellers.json); a c.2018 blog post claims TAG Certified Against Fraud, and no current seal was verified. No MRC accreditation, TAG certification or third-party viewability/IVT partner documented. [451, 603, 452, 604] | | Supply transparency | live | C | Advertisers see and bid on widget IDs and can download widget-level performance; the blocking guide does not confirm whether domain names are shown alongside widget IDs. Public sellers.json with named, non-confidential sellers (436 entries). Publishers must display 'clear and conspicuous' disclosure that widget content is third-party and paid. SSP guidance states 'no ads.txt necessary with our current demand' while publisher RTB requires an up-to-date ads.txt. [439, 440, 455, 605, 442, 606] | | Publisher controls | live (brand-safety tier gating); unconfirmed (category/advertiser blocking) | C | Publishers control which inventory requires the premium 'Brand Safety Seal' standard (yellow-flagged ads excluded from premium inventory such as Apple News); RTB demand is enabled/disabled via account rep; CNAME/subdomain setup; consent-management guidance; IVT filters. The publisher dashboard tour documents no advertiser-, category- or creative-blocking controls; such controls are unconfirmed. [604, 603, 606, 595] | | Commerce capabilities | live (Flipp deals content); announced (Smart Commerce) | D | Flipp integration (2025-03-18) serves geo-targeted grocery weekly specials on local-news publisher sites, with a vendor claim of 'up to 25%' yield increase; 'Smart Commerce' named as a future phase of the Mula OS (2026-03-09). No native product-unit, retail-media or affiliate commerce product documented. [607, 457] | | Interoperability | live | C | OpenRTB 2.5 with Native 1.1/1.2 on both DSP (buy) and SSP (sell) sides; DSP side accepts native, display/banner and video, ~2,000 QPS, 2–3 day integration, supply-chain data passed, Gzip; SSP side offers display/banner and native only (no video), POST only, honours floors, multiple bids per request, no user sync, no first-price auction support, no custom ext fields, reporting API/UI in EST. Prebid.js and Prebid Server 'revcontent' adapters (docs.prebid.org: media types display and native; TCF-EU 'No', USP 'No', COPPA 'No'; GVL ID 203; maintainer aziz@revcontent.com). Publisher RTB via PubMatic/BidSwitch. GAM standard and AMP unit setup; React Native SDK; Apple News via GAM 360. [441, 442, 443, 608, 609, 606, 595, 610] | | Geographic reach | live | C | US-centred (Sarasota HQ; ACH for US publishers; US press datelines; US zip/metro targeting). Brazil expansion with local team (2021-03) and Brazilian publisher signings; country targeting with in-country region refinement; international publisher reporting by geo; AWS endpoints in us-east-1, us-west-2, eu-west-1, ap-southeast-2; PayPal/wire for international publishers; CBD ads limited to US, CA, UK, AU, NZ, DE; political ads restricted in Brazil, EU, Middle East, North Africa and France pre-election. Share of impressions or spend by geography not disclosed. [611, 612, 441, 613, 614, 595] | Performance evidence: Tapstone (a native-ad buying platform) 'exceeded revenue goals by 40%' on Revcontent after launching 30+ boosts in Aug 2020; a single boost reached $20K revenue in Nov 2020 (vendor assertion, sponsor Revcontent press release quoting the client) [615]; Salon's CRO said 'Revcontent outperformed other content recommendation providers by 21 percent during Salon's test period'; Salon site loads 5x faster and ad viewability 'nearly doubled' year-over-year (vendor assertion, sponsor Revcontent press release (Sept 2020)) [594]; Mula OS: '25%+ RPS increase compared to control' and '18% increase in purchase intent with Vertical Video formats'; RevContent widgets '70% faster page loads than competitors'; '90% of demand is first-party' (vendor assertion, sponsor Joint RevContent–Mula press release (2026-03-09)) [457]; RevGPT-generated headlines 'consistently outperformed user-created alternatives' in testing (vendor assertion, sponsor Revcontent blog) [600]; Flipp demand activation yields 'up to 25%' increase in publisher yield (vendor assertion, sponsor Revcontent blog (co-marketing with Flipp)) [607]; Widget speed comparatives over time: '60% faster than competitors' (2020), '68% lighter than competitors' (2021), '70% faster page loads than competitors' (2026) (vendor assertion, sponsor Revcontent press releases) [594, 611, 457] ### Dianomi (active) *Dianomi plc (company number 04513809, England and Wales)* A small London-listed finance specialist that pays publishers 73% of revenue, under offer from Taboola at about 0.69 times revenue on upfront equity value. **SWOT (2026-09-27).** Small London-listed network selling per-click native ads beside finance and business news, with 78% of 2025 revenue from the US and a recommended Taboola takeover pending. - **Strengths.** Holds direct revenue-share contracts with 341 business, finance and lifestyle publishers at FY2025, including Reuters and CNN, with publisher churn of 2.9% by revenue. (record) [263, 460] Holds a finance niche with named clients such as Charles Schwab, Invesco and Bank of America, the specialism Taboola cited when it offered to buy the company. (record) [263, 461] Average revenue per click rose 7.4% to £0.58 in FY2025 while impressions fell 14.1%, a price gain rather than evidence of better advertiser outcomes. (record) [263] Returned to 2% revenue growth in H1 2026 (4.5% at constant currency), narrowed its adjusted EBITDA loss to £0.1m and held £6.0m cash with no debt. (record) [462, 463] - **Weaknesses.** Its documented algorithm optimises for engagement, clicks, views and revenue, and the conversions it tracks inform manual optimisation, with no conversion bidding objective documented. (record) [263, 464] No attribution window for reported conversions and no lift or incrementality partner is documented publicly as of September 2026, though DoubleVerify and IAS integrations are. (record) [465, 466] Revenue fell from £30.2m in FY2023 to £27.4m in FY2025, and adjusted EBITDA was a loss in FY2023, FY2024 and FY2025. (record) [467, 263] Published advertiser-outcome figures are Dianomi's own unnamed-client case studies, and its homepage claim of 8 billion+ monthly impressions conflicts with 39.1bn impressions reported in its FY2025 annual report. (record) [468, 469, 263] - **Opportunities.** If the Taboola scheme completes as expected before end-2026, the finance network would reach performance advertisers through Realize, Taboola's advertising platform, as Taboola's announcement sets out. (inference) [470, 461] If the Dappier answers engine, still in publisher talks at mid-2026, goes live, it would place ads inside the AI answers management blames for lost traffic. (inference) [253, 462, 263] The 67 new advertisers won in H1 2026 could widen demand, but growth depends on average spend, which slipped to £109k from £113k a year earlier. (inference) [462] - **Threats.** Impressions fell 14.1% in FY2025, which management blamed on AI summaries and searches that end without a click, and could fall again if publisher readership keeps shrinking. (inference) [263, 470] If Apple withdrew Dianomi's Apple News sales-agent approval, the channel behind 22% of FY2025 revenue would be at stake, a principal risk the annual report discloses. (inference) [263] Up to 24p per share depends on publishers adopting Taboola's standard terms and on revenue targets, giving the buyer a reason to move publisher contracts onto its terms. (inference) [470] - **What is sold.** Contextually targeted, ID-free ad placements (native content units, video, podcast audio units, IAB display sizes, and from April 2026 interactive units) served alongside editorial content on the websites, apps, Apple News feeds, email and AMP pages of premium business, finance and lifestyle publishers. Advertisers buy mainly per click (per view for video; some CPM), either through a Dianomi-managed account service or the My.Dianomi self-service platform. Adjacent products: Dianomi Audiences (buyable audience segments built from curated publisher lists, contextual keywords and first-party historical campaign data), Dianomi Insights (analytics, currently a value-added service), a programmatic route (viewable-CPM buying and Dianomi buying additional inventory programmatically), an affiliate vertical (Rakuten, Awin) and an announced Dappier AI 'financial answers engine' with conversational ad units. - **Who buys it.** Premium financial-services and business advertisers, predominantly in the US (78% of FY2025 revenue): the 2025 Annual Report names Charles Schwab, Invesco and Bank of America; recent wins named include Legal & General, UBS, LaSalle Investment Management, Dropbox, Pictet Asset Management, KeyBank, PwC, London Business School, Polen Capital and Northern Trust. Lifestyle advertisers contributed £2.8m in FY2025. The top 100 advertisers averaged £219k spend in FY2025; average advertiser spend in H1 2026 was £109k. An Investor Relations and Corporate Communications vertical for listed companies launched in H1 2026. - **How it earns.** Dianomi collects payment from advertisers on a per-click basis (per view for video ads; some advertisers on CPM), retains a share and passes the remainder to publishers under directly negotiated revenue-share agreements; cost of sales is the publisher revenue share, so gross margin (27.1% in FY2025, 28.9% in H1 2026) is Dianomi's net take. A small number of publishers receive minimum guarantees. Apple News, where Dianomi is an approved sales agent, produced 22% of FY2025 revenue. Additional revenue lines: programmatic (£1.7m FY2025), affiliate (£0.5m FY2025); Dianomi Audiences is not priced separately but was used on £3.0m of FY2025 advertising revenue. - **Supply and data dependencies.** Supply: direct revenue-share agreements with c.300-341 premium publishers (Reuters, CNN, Associated Press, WSJ, Condé Nast, Golf.com named), plus the Apple News channel as an approved Apple sales agent (22% of FY2025 revenue, flagged as a principal risk). Traffic dependency: FY2025 impressions fell 14.1% and devices 13.5% because of lower readership at non-subscription publishers as audiences consume AI summaries and zero-click search results. Data: targeting is contextual and ID-free ('does not rely on cookies or IDs'); optimisation uses Dianomi's own historical campaign click data (AI bidder embeddings, Dianomi Audiences). Currency: revenue is 78-79% USD while reporting is in GBP, producing a £0.7m constant-currency gap in FY2025 and a £0.4m FX loss on cash. - **Differentiation.** Vertical specialism: a curated, directly contracted network of business, finance and lifestyle publishers sold to financial-services advertisers, which Taboola's stated acquisition rationale singles out ('a highly specialised, premium ad network focused on finance'). Contextual, ID-free targeting positioned as privacy-first, with first-party historical campaign data (Dianomi Audiences, AI bidder embeddings) rather than third-party identifiers. Apple News approved sales agent status, giving access to publisher inventory in the Apple News app that publishers otherwise cannot monetise through Dianomi-type networks. Low publisher churn (2.9% on a revenue basis) and tier-one relationships (CNN whole-site, AP top-ten publisher) despite a shrinking open-web traffic base. Rising RPC (£0.54 to £0.58) indicating premium CPC pricing relative to general-purpose recommendation networks (no cross-vendor comparison available in this record). - **Risks.** Takeover execution: the Taboola scheme is recommended and backed by 75.3% irrevocables but not completed; contingent consideration depends on publishers accepting Taboola standard terms, signalling likely contract migration for Dianomi publishers post-close. Publisher traffic erosion: FY2025 impressions -14.1% and devices -13.5% attributed to AI summaries and zero-click search; the board explicitly cited this structural change as a reason to sell. Apple News dependence: 22% of FY2025 revenue via an approved-sales-agent status that Apple could withdraw. Advertiser demand cyclicality and concentration in financial services and the US (78-79% of revenue); the company itself notes winning new advertisers and publishers 'has been more challenging than in previous years'. Currency: GBP reporting against USD revenue (£0.7m constant-currency drag and £0.4m FX loss on cash in FY2025). Minimum guarantees to a small number of publishers could hurt profitability if performance falls short. Cash burn: cash fell from £8.8m to £5.8m in FY2025 (cash used in operations £2.1m), although H1 2026 stabilised at £6.0m. Key-person and small-team dependence (c.42 staff); the 2.7 announcement contemplates a 'limited reduction in headcount' and possible relocation of the US subsidiary into Taboola's New York office. - **Best fit.** Financial-services, asset-management, banking and B2B advertisers seeking contextual placement next to business and finance editorial in the US, UK and Australia on a CPC basis. Investor-relations and corporate-communications teams of listed companies targeting institutional, adviser and retail investor audiences (vertical launched 2026). Premium finance and business publishers wanting a curated, vetted-advertiser monetisation partner, including for Apple News inventory. Advertisers prioritising ID-free/contextual delivery and brand-safe premium environments over audience-graph reach. - **Poor fit.** Advertisers needing CPA/ROAS bidding, incrementality measurement or DSP-native programmatic buying at scale; conversions are tracked and reported but inform manual optimisation. Buyers requiring open-web reach outside business, finance and lifestyle verticals or outside the US/UK/Australia. Publishers needing granular self-serve advertiser or category blocking tools (none documented publicly). Anyone assuming Dianomi's standalone roadmap: post-acquisition, product direction is expected to migrate toward Taboola's Realize platform and standard publisher terms. - **Limits.** Small scale: FY2025 revenue £27.4m and c.42 staff; H1 2026 revenue £13.4m; loss-making at adjusted EBITDA level in FY2024, FY2025 and H1 2026. FY2025 advertiser count is not disclosed; comparable counts exist only for FY2024 (303 active) and half-year 'in the period' counts (225 in H1 2026), which are not the same measure as the annual figure. No CPA/ROAS bidding objective, attribution window, incrementality methodology or lift partner is documented publicly; conversion tracking (pixel, server-to-server, AppsFlyer and Adjust) and DoubleVerify and IAS integrations are documented in the help centre. Publisher-side blocking and creative approval controls are not documented publicly; advertisers can exclude up to 50 publishers per account through Supply Manager and adjust or block bids per publisher. Programmatic remains small (6% of FY2025 revenue) and volatile (H1 2026 supply revenue £116k vs £857k); it runs through named SSPs (Magnite, PubMatic, Microsoft Monetize, Adform), CPM-only and first-price. The AI bidder is described as developed in 2025 but its production status is not stated; Dappier AI answers engine has no commercial terms and no live publisher deployment reported as of H1 2026. 22% revenue dependence on the Apple News channel, which the company itself lists as a principal risk. Website marketing statistics (publisher count, monthly impressions, devices) are inconsistent with audited KPIs and with the 2.7 announcement figures. | Dimension | Availability | Evidence | Observed | |---|---|---|---| | Supply access | live | A | Direct revenue-share agreements with 341 publishers (FY2025 KPI; 294 'in the period' in H1 2026; marketing says '320+' and the 2.7 announcement 'over 250 premium publishers') of business, finance and lifestyle content including Reuters, CNN (whole site from 2025), Associated Press, WSJ, Condé Nast, Golf.com, Fox Business and Euronews; placements on desktop and mobile web, apps, Apple News feeds (as an approved Apple sales agent, 22% of FY2025 revenue), email and AMP. Reach stated as 'over 400 million devices per month' (annual report) / 424m average unique devices (KPI). [263, 462, 460, 470] | | Format support | live | C | Advertiser site lists native (single or multi-ad units), video (auto-play in view), podcast (user-initiated audio), display in IAB sizes (MPU, half page, leaderboard, billboard), canvas (up to 200 characters), verified CPC (pop-up verification page), app download, RSS feed, disclaimer and content hub units; publisher site lists in-article, below-article feed and right-rail placements. Dianomi Interactive (six interactive formats) launched April 2026. The annual report describes a transition 'from a predominantly native advertising model to a multi-format platform' as 'well advanced'. [468, 460, 616, 263] | | Advertiser workflow | live | A | Two core services: an account-managed service with a dedicated account manager, and self-service via the My.Dianomi platform for advertisers previously vetted by Dianomi; advertisers are vetted before onboarding. Pricing is per click (per view for video; some CPM); Dianomi may advise on CPC pricing to improve delivery. [263, 468] | | Creative tools | unconfirmed | D | No creative-building or generative tooling is documented; the interactive formats are described as ad units rather than a self-serve builder, and their cited performance figures come from 'Celtra Creative Insights', not from Dianomi campaign data. [616, 468] | | Optimisation | live | A | Ad optimisation is 'controlled by Dianomi's ad algorithm which places ads on website pages depending on the native content and the previous success of ads to maximize audience engagement, clicks, views and revenue'. During 2025 engineering 'developed an AI-driven bidder trained on years of Dianomi campaign click data' using 'a cascade of deep neural network models' to select the ad and compute a per-request bid within budget and pacing constraints; the H1 2026 report still lists it under initiatives Dianomi is 'investing selectively in', so live deployment is not confirmed. Objective stated as engagement/clicks/views/revenue; no documented CPA or ROAS bidding objective. [263, 462] | | Conversion-data access | live | C | The help centre documents an image conversion pixel, server-to-server events through Google Tag Manager or a direct server call, and AppsFlyer and Adjust app events, with conversion reporting by ad variant, campaign, device and publisher. No attribution window was found in the articles read. Conversions inform manual optimisation. [465, 617, 464] | | Measurement support | live | C | DoubleVerify blocking and monitoring and IAS monitoring are documented integrations, with third-party click and impression tracking and a reporting API; a real-time analytics dashboard and the Dianomi Insights tool are also offered. No brand-lift, holdout or incrementality tooling is documented. [466, 618, 619, 263] | | Supply transparency | live | C | Advertisers get publisher-level click and conversion reporting, account-level publisher exclusions of up to 50 through Supply Manager, and per-publisher bid adjustment or blocking. The network is negotiated directly with publishers. No pre-campaign site list or independent audit is documented. [619, 620, 621] | | Publisher controls | unconfirmed | C | The publisher page documents no publisher-side controls (category blocking, advertiser blocking, creative approval); it states Dianomi 'only works with trusted, brand-safe advertisers' and that advertisers are vetted centrally, and publishers get real-time revenue tracking via the 'Dianomi Dashboard'. Publisher-level blocking tools are therefore unconfirmed. [460, 263] | | Commerce capabilities | live | A | Affiliate advertising vertical launched in 2025 with 'strong working relationships' with Rakuten and Awin, generating £0.5m in FY2025 'from a standing start' after the affiliate sector's shift to CPC payment; Dianomi Interactive includes shoppable video, product feature hotspots, interactive gallery and customisation units. No native checkout, product-feed ingestion or retail-media integration documented. [263, 622, 616] | | Interoperability | live | C | Programmatic buying runs through named SSPs (Magnite, PubMatic, Microsoft Monetize, Adform) in open-market and private-marketplace deals, CPM-only and first-price; a reporting API exists. Programmatic revenue was £1.7m in FY2025 (6% of revenue). No agent-protocol implementation was found. [623, 619, 263] | | Geographic reach | live | A | Offices in London, New York (Dianomi Inc., Delaware-incorporated) and Sydney (Dianomi Pty Ltd). FY2025 revenue: USA £21.5m (78%), EMEA £5.1m (18%), APAC £0.9m (3%); H1 2026 USA 79.4%. The 2021 IPO use of proceeds targeted sales expansion in North America and EMEA. [263, 462, 624] | Performance evidence: For one US bank, 'Dianomi has been responsible for driving 74% of landing page visits' and 'CPA is nearly 10 times lower than benchmark' (vendor assertion, sponsor Dianomi (advertiser marketing page)) [468]; Interactive formats deliver '2.3% average engagement rate', '1.2x higher engagement than benchmark', '7.3% click-through rate', '4.6x higher than standard formats' (vendor assertion, sponsor Dianomi blog citing 'research from Celtra Creative Insights') [616]; Average revenue per click rose 7.4% to £0.58 in FY2025 while impressions fell 14.1% (direct record, sponsor Dianomi plc (audited annual report KPI)) [263]; Dianomi is used by '8 out of 10 largest asset management companies', '84% of US C-suite executives reached', '6 million customers driven monthly', '8 billion+ native ad impressions monthly' (vendor assertion, sponsor Dianomi homepage) [469] ### Readpeak (active) *Readpeak Oy (Finnish business ID 2615511-5); registered 2014-04-03 as MCN-Media Content News Oy, renamed Readpeak Oy on 2017-02-14; 'MCN-Media Content News' retained as a parallel trade name. Foreign entities found in registries: Readpeak AB (Sweden, org 559242-0490, registered 2020-02-12), Readpeak Norway NUF (Norway, org 929 123 166, registered 2022-05-23; branch of an unspecified foreign company), Readpeak AS (Norway, org 937 432 275), READPEAK LTD (England & Wales, no. 15479295, incorporated 2024-02-10), Readpeak SAS (France, Siren 104692926, registered 2026-04-29).* A Nordic self-serve native platform growing on minority investment, evidenced mostly from its own site. **SWOT (2026-09-27).** Helsinki-based, founder-led self-serve native platform selling mainly click-priced placements on European news sites, with the fullest documented pre-campaign publisher list among profiled vendors and a US launch in preparation. - **Strengths.** Support documents show a named list of available publishers before a campaign, with per-site bids and spend limits, matching the ANA's inclusion-list advice. (record) [297, 298, 89] Statutory accounts show the Finnish parent profitable every year from FY2021 to FY2025, with revenue rising from EUR 2.527m to EUR 7.644m (years to June). (record) [471] Advertisers can track conversions three ways: a consent-based 30-day cookie script, a cookieless script limited to the landing page, and server-to-server reports keyed to a click ID. (vendor documentation) [472, 473] Self-serve buying is open to businesses with a company email and VAT ID, verified usually within 24 hours, across four formats including Video Native priced per completed view. (vendor documentation) [474, 475, 476] - **Weaknesses.** Finnish-parent accounts show EUR 7.644m revenue for FY2025 (to June), and the company's EUR 15m expectation cannot be checked because group accounts are not public. (record) [471, 408] Pay-per-click and pay-per-completed-view are the only documented prices, click-through rate is the default goal, and audience targeting or return-on-spend bidding is not documented publicly as of September 2026. (vendor documentation) [477, 478, 479] Performance evidence is vendor-published: case studies with undefined benchmarks and one agency-run brand-lift study for a single Danish advertiser, with no significance testing reported. (vendor documentation) [480, 481, 482] Supply is unclear: the vendor claims 1,500+ news sites, its investor's page says over 1,000, and publisher terms and controls are not documented publicly as of September 2026. (record) [483, 484, 485] - **Opportunities.** If the US launch, described as in preparation on 11 September 2026, secures US publishers, pre-campaign site lists would match the ANA's inclusion-list advice to advertisers. (inference) [486, 297, 89] If markets added since November 2025 (France, Austria, Belgium, Italy, Poland) build local supply, they could lift growth that slowed to 6.8% at the Finnish parent in FY2025. (inference) [487, 486, 471] - **Threats.** If consolidation continues (Outbrain closed its Teads purchase in February 2025, Taboola made an offer for Dianomi in September 2026), Readpeak faces larger rivals for European publisher contracts. (inference) [45, 31] The planned US launch depends on building US publisher supply, none documented as of September 2026, while 46 people listed on LinkedIn already cover 14 claimed markets. (inference) [486, 488, 483] - **What is sold.** A self-serve (with account-management support) programmatic native advertising platform that places advertiser-supplied native units on partner news/publisher sites in Europe. Four formats: Content Native (drives to a blog/landing page), Product Native (product, event, front page or web store), Instant Article (article hosted without a separate landing page) and Video Native (in-feed video, max 15 s, bought on CPCV). Buying is CPC-bid (CPCV for video) at campaign and site level with geo/time/site/section/contextual targeting, site-level reporting, conversion tracking (JS script, consent-free script, server-to-server postback), an opt-in conversion-optimisation algorithm, automatic pacing, and AI-generated ad variations. - **Who buys it.** Brands and agencies in the Nordics and Western/Central Europe. Vendor claims range from 'over 3000 happy clients' and 'over 250 agencies' (homepage) to '3,500+ advertisers' (about page) and 'over 3,000 brands' (Sept 2026 release). Named case-study advertisers include BMW Nordics, Samsung, Audible, Magenta Telekom, Kia, LG Sweden, Toshiba, ABN AMRO, Verman (via Kinesso Finland), Austrian Tourism, Naturskyddsforeningen, Diabetesforeningen and IAB UK. Vendor states it serves 14 markets. - **How it earns.** Media margin on CPC/CPCV-priced campaigns bought through the platform (advertiser pays per click / completed view; publisher remuneration model not documented). Payment by credit card by default, invoicing on request; business ID/VAT ID required to run campaigns. Third-party DoubleVerify segments carry an additional displayed cost. Platform/SaaS fees and minimum budgets are not documented. - **Supply and data dependencies.** Supply: direct partnerships with publishers plus a 'curated premium network'; vendor claims 1,500+ premium news sites (Voland's portfolio page says 'over 1,000 publishers'). Privacy policy references auction IDs, bid request IDs and cookie syncing with third-party platforms, implying RTB-style delivery, but the split between direct publisher integrations and SSP/OpenRTB supply is not documented. Data: contextual/semantic signals and aggregated performance data (vendor states it does not rely on third-party cookies for targeting); DoubleVerify pre-bid brand-safety and contextual segments; Neuwo and MaxMind listed as integrations; consent handling via IAB Europe TCF (vendor states Vendor ID 290, not verified in this record). Conversion data via advertiser-installed tracking script (cookie-based up to 30 days, or consent-free cookieless landing-page-only) and S2S postback keyed on an rpcid click ID. - **Differentiation.** Site-level transparency and control: named publisher list, cherry-pick sites, per-site max bids and spend caps, saved presets (documented in support docs). Self-serve access for any business with a company email, verified within ~24 hours, credit-card payment by default; targets long-tail brands and agencies as well as majors. Premium news-publisher positioning in the Nordics (named placements on VG, Iltalehti, Dagens PS) rather than long-tail recommendation widgets. Consent-free cookieless tracking script and a proprietary cross-page time-spent metric as engagement currency. CPCV-priced in-feed Video Native (completed view = 100%). Founder-led, profitable (per Finnish accounts) and minority-backed by a growth investor rather than VC-scaled. - **Risks.** Scale claims are inconsistent across the vendor's own pages (3,000+ vs 3,500+ advertisers; 1,000+ vs 1,500+ publishers; nine vs 14 markets) and undated. Revenue communicated to the market (EUR 15m FY2025 expected) is roughly double the verifiable parent-only figure; growth in the Finnish entity slowed to 6.8% in FY2025. Rapid multi-market expansion (CH 2024, NL/UK 2024, FR 2025, AT/BE 2026, IT/PL 2026, US planned) on a small team; per-market supply depth outside the Nordics is unverified. CTO change within 12 months (Tuomo Tilli quoted Sept 2025; Sami Pippuri appointed Sept 2026). CPC buying with CTR-default optimisation creates click-quality incentives that the vendor mitigates with its own time-spent metric (proprietary, not independently audited). Consent-dependent cookie tracking (30 days) limits conversion attribution in high-refusal EU markets; the consent-free fallback captures landing-page activity only. Growth-fund minority investor (2022-vintage fund) implies a medium-term liquidity event; ownership could change before or after the cutoff (inference). Larger cross-border native platforms (Outbrain/Teads, Taboola) and publisher-owned native sales compete for the same premium news inventory (inference; not sourced in this record). - **Best fit.** Nordic and DACH/Benelux brands and agencies wanting transparent, site-selectable native placements on premium news sites with self-serve control (inference from documented controls and named placements). Content-led mid-funnel campaigns optimised to clicks and time spent rather than last-click ROAS. Advertisers able to install a tracking script or S2S postback and run 2+ weeks to use conversion optimisation. - **Poor fit.** US-centric buyers: no live US supply at cutoff (launch only 'in preparation'). Performance buyers needing CPA/ROAS bidding, audience/retargeting or lookalike targeting, or product-feed dynamic ads. Publishers seeking a documented monetisation product: no public publisher offering, controls or terms. Buyers requiring independently audited post-bid viewability/brand-safety reporting on every impression. - **Limits.** Only parent-only Finnish statutory figures (Readpeak Oy, EUR 7.644m FY to June 2025) are publicly verifiable; no consolidated group accounts were found, so the company-communicated EUR 15m (FY2025 expected) and EUR 12.0m (2024) cannot be reconciled. CPC/CPCV are the only documented buying models; no CPM, CPA or ROAS bidding and no value-based optimisation. Default optimisation is CTR-based; conversion optimisation is opt-in and needs advertiser-installed tracking plus a 7-14 day learning window. Targeting documented is geo (to zip level), time/day, site/section and contextual/semantic; no audience, retargeting, lookalike or first-party data onboarding is documented. Publisher-side product (integration method, controls, revenue share) is undocumented on the public site. Post-bid viewability/brand-safety measurement is not documented; DoubleVerify integration is pre-bid only per the support article. Small registered headcount (13 in the Finnish entity, 2 in Sweden; 46 listed on LinkedIn) relative to 13-14 claimed markets plus a planned US launch. Performance evidence is vendor-published case studies with undefined benchmarks; the one brand-lift study is agency-run for a single advertiser. | Dimension | Availability | Evidence | Observed | |---|---|---|---| | Supply access | live | D | Vendor states 'direct partnerships with publishers' and access to 'Europe's top publishers individually or through our curated premium network'; claims 1,500+ premium news sites (about page, Sept 2026 release) while its investor's portfolio page says 'over 1,000 publishers'. Named publisher placements in the BMW case: VG (NO), Iltalehti (FI), Dagens PS (SE), Fodboldnyheder (DK). Privacy policy lists auction IDs, bid request IDs and cookie syncing, consistent with RTB delivery, and the org chart includes a Head of Programmatic Partnerships; the share of supply reached via direct tags vs SSP/OpenRTB is not documented. [625, 483, 484, 626, 480, 486] | | Format support | live | C | Content Native, Product Native, Instant Article and Video Native. Creative spec: headline up to 70 chars (short 25), summary up to 120 (short 40), advertiser name max 25, CTA max 20, image 1200x628 recommended with no text or logos; video MP4 1280x720 16:9, 5-15 s (max 15 s), max 3.8 MB, audio off by default. Video Native bought on CPCV; Instant Article markets content 'without building a separate landing page'. Sponsored-labelling rules not documented in the guideline page read. [475, 477, 476, 625] | | Advertiser workflow | live | C | Self-serve signup with a company email; new accounts get 'limited functionality' until Readpeak verifies them ('usually takes up to 24 hours'); business ID/VAT ID and credit card required to create campaigns, invoicing on request via platform chat. Four user roles (Admin, Producer, Manager, Content Editor). Campaign setup: choose format, set max CPC bid, daily/weekly/lifetime budget, continuous or dated flight; select sites/sections; build creatives. Minimum budgets and platform fees not documented. [474, 477, 627] | | Creative tools | live | C | AI ad creation: from 1-4 initial variations plus landing-page content the platform generates '8 AI-generated ad variations' (headlines, teasers, CTAs); model/provider not named; no cost stated. Creatives are adapted into publisher templates ('transform into the correct fonts and scaled images'). A support article titled 'AI Ads Smart Variations' (2025-04-30) exists in the support sitemap but was not opened. [628, 625, 629, 298] | | Optimisation | live | C | Default optimisation 'focuses solely on click-through rate (CTR)'. Opt-in conversion optimisation uses a multi-step funnel (CTR, time on site, soft events such as scroll/form start, then final conversion), with goals ordered soft-to-final and a recommended 7-14 day learning period. Automatic pacing mode (optional, all users) adjusts bids intraday to land daily budget, favouring cheaper high-quality inventory when ahead. Marketing describes 'deep learning' over bid-stream, contextual and semantic signals. No ROAS or value-based bidding documented. [478, 630, 625] | | Conversion-data access | live | C | Two JS tracking scripts: a consent-requiring script using cookies for cross-page/cross-session tracking for 'up to 30 days', and a consent-free cookieless script limited to the landing page. Server-to-server: click ID passed as URL parameter rpcid; advertiser posts back to https://app.readpeak.com/url/ht with rpid, ev (event name) and cid. Conversions 'can be connected to the ad impression that led to the ad click'. GTM install guide, Google and Adform click-tracker integrations and a GA4 statistics article exist. Not documented: attribution window for S2S, revenue/value passing, offline or CRM uploads. [472, 473, 629] | | Measurement support | live | C | Reporting of impressions, clicks, CTR and conversions 'at publisher, section and segment level' with automated reporting and API access. Proprietary time-spent metric: pings 'every few seconds', tracked across all pages carrying the script, averaged across pages visited after an ad click; vendor documents how it differs from GA4 engaged sessions. DoubleVerify integration is pre-bid (50+ brand-safety segments, 400+ contextual segments, SIVT protection) selected per campaign at displayed extra cost; post-bid viewability/verification measurement not documented. Vendor announced TAG Brand Safety Certified (UK) in Sept 2025; TAG registry not checked. Homepage lists Google DCM, Adobe Analytics and Supermetrics integrations. [625, 631, 632, 633] | | Supply transparency | live | C | Advertisers see 'a comprehensive list of all available publishers' by name, can add/remove sites individually or in bulk, set per-site max bids and daily/weekly/lifetime spend limits, save site presets, and see site insights (relevance and volume scores). Platform page: 'See performance down to the site level, and cherry-pick the sites you want.' Site-level CPC bidding is stated on the homepage. [297, 298, 625] | | Publisher controls | unconfirmed | C | Advertiser-facing brand-safety policy: strict publisher vetting (legal compliance, original content, legitimate traffic), prohibited content categories, allow-lists for sites/categories/content, contextual avoidance of sensitive topics, and campaign review before go-live. No publisher-facing documentation found: publisher ad-approval workflow, category blocking, floor prices or revenue share are not documented (the site's page sitemap contains no publisher page). [634, 485] | | Commerce capabilities | unconfirmed | D | 'Product Native' promotes 'a product, event, front page or a web store'. No product-feed ingestion, catalogue/dynamic product ads, retail-media or shoppable unit is documented. [625, 477] | | Interoperability | live | C | Support sitemap lists 'Campaign Reports API Integration' (lastmod 2026-08-19); the guessed article URL returned 404 so API scope is unverified. Homepage integrations: Google DCM (Campaign Manager), Adobe Analytics, DoubleVerify, Neuwo, Supermetrics, MaxMind. Support articles for Google Click Tracker and Adform click tracker integrations exist. Privacy policy states IAB Europe TCF participation with Vendor ID 290 (the GVL fetch was truncated before ID 290, so unverified). No DSP/OpenRTB buying route, deal IDs or programmatic guaranteed documented for buyers. [629, 625, 626, 589] | | Geographic reach | live (Europe); announced (US) | A | Website in nine languages (EN, DA, DE, FI, FR, IT, NL, NO, SV). Markets claimed across releases: Finland, Sweden, Norway, Denmark (Nordics), UK, Germany, Switzerland, Netherlands, Belgium, France, Austria, Italy, Poland; vendor says '14 markets'. Registered entities: Finland (2014), Sweden (2020), Norway (2022 branch; plus Readpeak AS), UK (2024), France (2026). US: CEO titled 'CEO & US Market Lead' and release says 'preparing for US launch'; no US entity found. No Italian, Polish, German, Austrian, Swiss or Dutch entity surfaced in the searches run. [485, 483, 486, 635, 487, 636, 637, 638, 639] | Performance evidence: Native ads on Readpeak delivered 20% better ad recall, 30% better top-of-mind awareness and 5% higher unaided awareness than standard display, with equal favourability impact (Danish FMCG chain, OMD Denmark, ~DKK 500,000, 2 months 25 days, min. 333 respondents per group in display/native/control cells). (commissioned measurement, sponsor OMD Denmark with Readpeak (published on Readpeak's site)) [482]; BMW Nordics 12-month always-on: CTR 0.38-0.43% by country, CPC NOK 7.48 / DKK 3.59 / EUR 0.62 / SEK 5.65, ~65.7m impressions and ~273k clicks in total; vendor states CTR '47% higher' than benchmark and '50%+' outperformance vs competitor platforms. (vendor assertion, sponsor Readpeak (case study with BMW Northern Europe quote)) [480]; Verman (Finland, via Kinesso Finland) over 22 months: 170,000+ clicks, average CTR 0.60% (peaks >1%), average CPC EUR 0.70 vs EUR 3.63 for display brand campaigns, average CPM EUR 4 vs EUR 6.31 for display. (vendor assertion, sponsor Readpeak (case study with Kinesso Finland quote)) [481]; Video Native beta showed 'strong performance in key areas such as engagement rates, traffic quality, and view-through completion'; 51.8% of surveyed clients rated Video Native as having the highest growth potential. (vendor assertion, sponsor Readpeak) [476]; 'State of Native Advertising in the Nordics 2024': survey of 120 advertisers, agencies, publishers and ad-tech vendors via Surveypal in Aug-Sep 2024, distributed through Readpeak channels, IAB Finland, IAB Sweden and MarkkinointiKollektiivi. (vendor assertion, sponsor Readpeak) [640] ### Nativo (Life360) (acquired) *Nativo, Inc. (Delaware corporation; now a wholly owned subsidiary of Life360, Inc.)* A sponsored-content and native-display specialist bought by Life360, the family-location app company, for about $120 million. **SWOT (2026-09-27).** Nativo has been Life360's US advertising unit since January 2026, pairing a native publisher network and publisher tools with the Life360 app's audience, though outcome claims lack published methods. - **Strengths.** Life360 now reports advertising as its own line, $22.0M in Q2 2026, including its own in-app ads; Nativo itself is not reported separately. (record) [489, 490] Reach pairs the Life360 app, about 43M US ads-eligible users in June 2026, with 400+ premium publishers and 20k+ websites and apps, Life360 says. (vendor claim) [491] Publishers can connect through standard header-bidding auction routes, with Prebid.org listing Nativo adapters for Prebid.js and Prebid Server, plus documented Amazon TAM and Google Open Bidding options. (vendor documentation) [492, 493] Its website says it serves both sides: brands buy through managed service, self-serve or curated deals, and publishers use an ad renderer and creative hosting. (vendor claim) [494, 495, 496] - **Weaknesses.** Bidding to conversions, cost per acquisition or return on ad spend is not documented publicly as of September 2026, and conversion tracking is documented only through pixels. (vendor documentation) [186, 497, 498] Outcome claims lack published methods: Life360's 'up to 47%' higher call-to-action rates and grocery footfall lift state no design, and Brand Rank's test group is engaged viewers only. (record) [499, 500] Standalone Nativo revenue appears in no Life360 filing reviewed; the only figure, about $63 million for 2025, is an unaudited statement by Life360's CFO on an earnings call. (record) [501, 502] Publisher counts went from 'hundreds' to 'thousands' to '400+ premium' in ten months, and a site list is not documented publicly as of September 2026. (record) [364, 503, 491] - **Opportunities.** Travel, automotive and everyday retail, the categories Life360 says it is concentrating on, offer growth if its family and location data performs on off-app publisher inventory. (inference) [499, 364] If Life360 builds its ad infrastructure internationally, as it said in August 2026 it was preparing to do, the unit could sell beyond its all-US reach figures. (inference) [499, 491] If Life360 publishes the design behind its grocery footfall result, including 35,000 net-new incremental trips, buyers could treat it as evidence rather than a vendor claim. (inference) [491, 499] - **Threats.** With the trade name amortised over two years and nativo.com redirecting to Life360 Ads, the unit depends on retaining customer relationships Life360 booked at $38.2M. (inference) [501, 494] Ad use of a family-safety app's location data depends on regulator and user tolerance, with parents-only targeting and sensitive-location exclusions reported by AdExchanger as safeguards. (inference) [504, 364] - **What is sold.** Native and content-driven ad delivery across a directly integrated publisher network plus Life360's own app inventory. Advertiser-side: managed-service campaigns, a self-serve platform, programmatic direct/curated deals (SPARC predictive audiences, Curation), formats branded Native Article (in-feed unit driving to publisher-hosted branded content), Native Display, Native Video, Stories, plus Landing Page and Skins on the advertiser page; measurement add-ons Brand Rank (in-unit survey lift/rank). Publisher-side: an SSP (Prebid client/server, Amazon TAM, Google Open Bidding, OpenRTB, code-on-page), a SaaS ad renderer / creative hosting suite with audience extension, and an 'Exclusive Demand Marketplace'. - **Who buys it.** Brands and agencies (site names Novartis, MGM Resorts, HP, Mazda, SAP, GM, State Farm, AbbVie, Hilton, Edward Jones); Life360 says H2 2026 focus is travel, automotive and everyday retail. Publishers (site names Ziff Davis, AP, Raptive, Freestar, WebMD, Paramount, Newsweek, Yahoo, Dotdash Meredith, Time). Programmatic buyers via curated PMPs. - **How it earns.** Under Life360 the unit is reported as 'advertising revenue' recognised when impressions are delivered or services performed, on a gross basis when Life360 is principal and net when agent; channels are managed advertising arrangements, programmatic/open-marketplace, and other advertising-related services (professional services, analytics, platform access fees). Cost of advertising revenue includes traffic acquisition costs paid to publishers. Pre-acquisition Nativo pricing (CPM/CPC/rev-share terms) is not disclosed in any opened source. - **Supply and data dependencies.** Supply: direct publisher integrations ('100% publisher direct network', '400+ premium publishers and 20k+ websites and apps' per Life360 Q2 2026 deck) plus Life360 in-app inventory (~43M US ads-eligible MAU, June 2026). Data: Life360 first-party family/location signals (parents only, under-18 members excluded, sensitive locations excluded per AdExchanger), Nativo engagement signals, SPARC context/device/zip-census/ISP/time signals. Dependent on publisher header-bidding stacks (Prebid, Open Bidding, TAM) for programmatic supply. - **Differentiation.** Only native vendor in this set whose owner brings a first-party consumer app with ~43M US ads-eligible MAU and location/family signals; Life360 says reach expanded from ~16% to over 95% of US ad-eligible adults by adding Nativo's network (AdExchanger). Publisher-hosted branded-content unit (Native Article) plus a publisher-side ad renderer/SaaS, i.e. it sells tooling to publishers as well as media to brands. Sell-side interoperability documented at spec level (Prebid.js and Prebid Server adapters, Open Bidding, TAM, OpenRTB) rather than proprietary tag only. Cookieless audience product (SPARC) built from contextual/device/geo-census signals rather than identity graphs. - **Risks.** Brand sunset: 2-year trade-name life and the nativo.com redirect imply the Nativo brand is transitional; publisher and agency relationships carry 82% of identified intangible value, so attrition of the sales team is the main value risk. Revenue-quality mix: advertising revenue is a blend of gross (principal) and net (agent) recognition; 57% GAAP gross margin in Q2 2026 reflects traffic-acquisition costs to publishers. Concentration in managed service: Life360 attributes growth 'primarily' to managed advertising; self-serve and programmatic contribution is not quantified. Guidance dependence on H2 seasonality: FY2026 $98-115M requires Q4 roughly double Q1 ($19.7M); H1 delivered $41.6M. Investor reaction: ASX coverage reported the share price fell after the announcement (headlines only; articles not opened). Privacy exposure: location-derived audiences on a family-safety app; Life360 states parents-only targeting and sensitive-location exclusions (reported, not documented in filings opened). - **Best fit.** Mid-funnel branded-content and consideration campaigns needing publisher-hosted articles with engagement metrics. Advertisers wanting family/parent audiences with footfall outcomes in US retail, travel and automotive (Life360's stated H2 2026 focus). Publishers seeking a native ad renderer plus audience extension with header-bidding-compatible demand. - **Poor fit.** Performance buyers optimising to CPA/ROAS at scale — no documented outcome-bidding objective and pixel-only conversion tracking. Non-US campaigns — platform reach and guidance are US-only. Buyers requiring a full list of DSP pathways or documented third-party verification integrations. - **Limits.** Nativo standalone financials are not disclosed in Life360 filings; pro forma results are omitted as 'not material'; the only revenue figure, about $63M for 2025, is an unaudited statement by Life360's CFO on the March 2026 earnings call. Headline price ~$120M (65/35 cash-stock) vs ASC 805 consideration $104.0M — the filings do not reconcile the gap. Publisher-scale claims move between 'hundreds', 'thousands', '400+ premium publishers' and '20k+ websites and apps' within ten months; units are not consistent. No optimisation objectives beyond audience expansion are documented; conversion tracking is pixel-only (no server-side/CAPI found). DSP integrations are named only in part: Life360's August 2026 deck cites 'direct integrations with key DSPs incl. Amazon'; others are undisclosed. US-only ad platform per Life360's own 'internationally' framing; no non-US market evidence. 'Nativo Insights' not found on any opened page; 'Nativo IQ' appears in one site summary but no product page was located (guessed URL 404). | Dimension | Availability | Evidence | Observed | |---|---|---|---| | Supply access | live | A | Direct publisher integrations plus Life360 in-app inventory. Life360 Q2 2026 deck: '400+ premium publishers and 20k+ websites and apps', '~43M Life360 U.S. In-App Reach' (ads-eligible MAU, June 2026), '~234M U.S Off-Site Reach' (Comscore). Nov 2025 release said 'hundreds of publishers across CTV, mobile, and premium web and app'; Mar 2026 release said 'thousands of publishers'. Advertiser page claims '100% publisher direct network' and 'MFA-free inventory'. [491, 364, 503, 641] | | Format support | live | C | Publisher docs define four formats: Native Article (in-feed unit driving to publisher-hosted branded content, mid-funnel), Native Display (in-feed, lower-funnel), Native Video (in-feed non-incentivized), Stories (mobile in-feed immersive). Advertiser page lists Landing Page, Stories, Skins, Display. SSP and self-serve pages say native, video, display. Nov 2025 release mentions CTV among environments; no CTV format is documented on the site. [642, 641, 493, 497, 364] | | Advertiser workflow | live | C | Three routes documented: managed service ('expert team handle every detail'), self-serve ('launch, manage, and refine campaigns', 'Set your own budgets, formats, and targeting', no minimum spend stated), programmatic direct via curated PMPs/Curation. Advertiser resource centre has sections Getting Started, Building Your Campaign, Advanced Setup, Campaign QA, Tracking & Targeting, Reporting & Analysis. Self-serve changelog (release 4.6.1, undated) adds Nativo Predictive Audiences to Campaign Quick Setup. Life360 says revenue is 'primarily' managed advertising. [641, 497, 495, 643, 644, 489] | | Creative tools | live | C | 'Build or upload creative and go live fast' (self-serve); publisher SaaS: 'advanced ad renderer', 'creative automation and dynamic optimization', Creative Hosting Suite; Stories slide-description editor (release 4.6.1). Nov 2025 investor deck lists 'Advanced Ad Renderer' for custom formats. No generative-AI creative tool documented. [497, 496, 644, 645] | | Optimisation | live | D | SPARC / Nativo Predictive Audiences: cookieless lookalike expansion from engagement signals (IAB placement category, placement location, device/browser, zip-level census, ISP, day/time; '5 billion data points daily'), claims '3x better performance than cookies' and '2x more likely to convert'. Publisher SaaS cites 'dynamic optimization'. No documented optimisation objectives (e.g., CPA bidding, ROAS) found in opened pages; format docs map to funnel objectives only. [498, 644, 642] | | Conversion-data access | live | C | Nativo Conversion Pixels: created per advertiser; configurable click-through and view-through windows (30 days recommended); first- or last-interaction attribution; dynamic conversion value via 'optional_float'; pixels auto-paired to new budgets; prior converters can be excluded. No server-side/conversions-API or cookieless conversion option documented. [186] | | Measurement support | live | C | Brand Rank: in-unit surveys with control/test groups producing 'Via Rank' and 'Via Lift'; claims '3.4x average lift' (Nativo data, 2024). Reporting API ('programmatic way to access data on the Nativo Platform'). Adelaide AU-score study published by Nativo. Life360 cites store-visit lift and CTA-rate comparisons without stated methodology. No third-party verification partners (IAS, DV, MOAT) named on opened pages. [500, 646, 647, 499] | | Supply transparency | live | D | Claims 'publisher-first supply chain' with 'no MFA, max 2-hop Schain, and only clean, direct inventory'; publisher docs cover ads.txt/app-ads.txt; Prebid adapter supports supply chain. No site-list, sellers.json statement or MFA definition found on opened pages. [648, 649, 492] | | Publisher controls | live | C | Safety Blocking UI: block advertiser domains (single or bulk .txt/.csv upload) and IAB-style categories org-wide; section-level blocking and demand-channel blocking referenced; blocks reversible. No creative-approval workflow documented. Publisher revenue-share terms not disclosed (Revenue Adjustments and Tipalti FAQ sections exist). [650, 649] | | Commerce capabilities | unconfirmed | E | No retail-media, product-feed or shoppable native unit documented on any opened page. Conversion pixel supports dynamic transaction values. Life360 reports footfall (store-visit) results for a grocer. [186, 499] | | Interoperability | live | A | SSP is 'Integration-Agnostic': Prebid Client, Prebid Server, Amazon TAM, Google Open Bidding, oRTB, code-on-page; also standard JavaScript and Google Ad Manager integrations in publisher docs. Prebid.org lists bidder 'nativo' (GVL ID 263) for banner, video and native in Prebid.js and Prebid Server (Go and Java), TCF-EU yes, USP yes, GPP 'some', schain yes, floors yes, first-party data no, COPPA no, DSA no; adapter requires prior setup via prebiddev@nativo.com; Prebid Server adapter is a pass-through of OpenRTB requests. Buy-side DSP names not disclosed ('top DSPs'). Mobile SDK for native/video/display in apps. [492, 493, 649, 651] | | Geographic reach | live | A | All reach figures are US ('U.S. In-App Reach', 'U.S Off-Site Reach', '95% of U.S. adults'); Life360 said in Aug 2026 it is preparing 'to build out the same infrastructure internationally', implying the ad platform is US-only today. Prebid adapter supports TCF-EU consent. No non-US Nativo office or market claim found. [499, 491, 492] | Performance evidence: Nativo inventory averaged 28.21 AU, 40% above Adelaide's native benchmark and 21% above its display benchmark; native in-feed article ~30 AU (+48%) (commissioned measurement, sponsor Published by Nativo; commissioning/payment not stated; Adelaide CEO quoted) [647]; Campaigns using Life360 audience data see call-to-action rates up to 47% higher than third-party targeting; retail and pets 'more than two times' (vendor assertion, sponsor Life360 (shareholder letter)) [499]; A top-10 US grocery chain: 132,000 store visits, 35,000 net-new incremental trips, +43.5% blended in-store visit uplift from a single campaign (vendor assertion, sponsor Life360) [499, 491]; SPARC audiences '3x better performance than cookies' and '2x more likely to convert' (vendor assertion, sponsor Nativo) [498]; Brand Rank: '3.4x average lift as a result of content engagement' (Nativo data, 2024) (vendor assertion, sponsor Nativo) [500]; Publishers: '275% eCPM lift over standard display ads', '2x increase in deal size'; 'typical integration time 2 weeks' (vendor assertion, sponsor Nativo / Life360 investor deck) [648, 645]; Explore Georgia: 2.2x time spent, +54% time spent vs other media partners, 152% CTA rate above benchmark, 123k additional hours engaging (vendor assertion, sponsor Nativo case study) [494]; Nativo programmatic revenue grew more than 120% YoY in Q2 2025 (vendor assertion, sponsor Nativo press release) [652] ### TripleLift (active) *TripleLift Inc. (as named in its sellers.json; DUNS 063519038, TAG-ID 6c33edb13117fd86)* A Vista-owned programmatic native exchange that has cut staff three times since 2020 and now sells CTV and display as much as native. **SWOT (2026-09-27).** A Vista-owned sell-side platform that offers in-feed native alongside CTV and retail-media formats, with a public seller list but no published revenue and only vendor-backed performance evidence. - **Strengths.** Its public seller list (sellers.json) showed 7,941 entries on 27 September 2026, none confidential, and 1,268 distinct publisher domains once Google Open Bidding duplicates are removed. (vendor documentation) [505] Builds each native ad from the title, image and text in the buyer's winning bid, per its OpenRTB spec, so DSP buyers need no publisher-specific creative. (vendor documentation) [506, 507] Supports the industry's native bid standard, OpenRTB Native 1.2, in documentation updated 14 September 2026, plus Prebid header bidding and Google Open Bidding. (vendor documentation) [508, 509, 505] Amazon named it a third-party supply source for product-data native ads in December 2023, and it says it joined Amazon DSP's certified supply programme in October 2025. (record) [510, 511] - **Weaknesses.** Revenue, take rate (its share of spend) and profit are not documented publicly as of September 2026, leaving no public basis to compare its scale with listed rivals. (record) [512, 513, 514] Missed the ANA's deadline for log-level data, one of five SSPs, and Adalytics said it may have transacted on a 22-site MFA sample, like its peers. (record) [89, 515] Every performance figure found is vendor-stated or vendor-commissioned, and no public study with published design and results tests whether its native ads cause extra sales or conversions. (vendor documentation) [516, 517, 518] Conversion tracking tools and native placement standards are not documented publicly as of September 2026, and most of its knowledge base sits behind a login. (vendor documentation) [519, 518] - **Opportunities.** If ANA benchmark marketers keep buying mostly through private marketplaces (81.6% of spend in Q3 2025), TL Direct's deal-building could draw budget, provided deals screen out made-for-advertising sites. (inference) [290, 520, 515] If AI-agent buying scales, TL Direct's agent interface (MCP, generally available since June 2026) fits the transport AdCP specifies, though no AdCP or AAMP implementation is documented. (inference) [520, 377, 521] Its stated 2025 entry into 15 international markets and the September 2026 Multilocal path to European DSP clients could widen demand, depending on volumes neither release gives. (inference) [513, 522] - **Threats.** If Amazon steers demand to the native ads its Publisher Services launched in May 2025, TripleLift's commerce formats bought through Amazon DSP could lose volume. (inference) [74, 511, 523] After three CEO changes since 2022 and layoffs in 2023 and 2025, delivery of TL Spark and TL Direct depends on an executive team completed in August 2026. (inference) [524, 366, 525] - **What is sold.** Sell-side programmatic access (SSP) to open-web, app, CTV and offsite retail-media inventory, packaged with creative transformation. Web formats listed: Native, Image, Cinemagraph, Scroll, Window, Carousel, Reveal, Pharma (FDA-style ISI), Outstream, Instream, Branded Video, Display and others. CTV formats listed: Spots, Pause Ads, Product Spotlight, QR Spots, Social to CTV, Split Screen, Dynamic Overlay and L-Bar. Around these sit curated deals, free proprietary behavioural and contextual audience segments, and measurement add-ons. Native is one format family among many, not the whole product. - **Who buys it.** Demand side: agencies, brands and curators, buying through DSPs (in 2021 the company claimed 100% of the top 20 DSPs; its knowledge base lists DV360, Viant, Basis, Amazon DSP and Yahoo activation guides), through the self-serve TL Direct (GA 2026-06-10), or through managed service. Supply side: publishers and CTV/streaming owners that monetise through TripleLift, including directly and via Google Open Bidding and resellers. - **How it earns.** Unconfirmed in any primary document. As an SSP it presumably takes a share of media transacted, but no take rate, fee schedule or revenue figure has been published. TL Direct pricing is not disclosed. The company markets audience segments as carrying no extra data fees. - **Supply and data dependencies.** Depends on DSP demand (notably Amazon DSP for commerce formats and the Certified Supply Exchange), on Google Open Bidding for a large share of seller IDs (2,295 of 7,941 sellers.json entries carry the -EB suffix and google.com domain), on reseller intermediaries (for example Freestar and Publift), on retailer data and product feeds (Amazon ASIN assets, Walmart SKU feed), on CTV partners (DIRECTV Advertising for programmatic pause ads), and on the first-party data technology from the 2022 acquisition of 1plusX. - **Differentiation.** Sell-side creative transformation: it renders native (and CTV or commerce) executions from DSP-supplied assets, so DSP buyers can reach in-feed native without publisher-specific creative (vendor-documented since 2014). An early programmatic-native pedigree: it claimed the first RTB in-feed native exchange (2014) and describes itself as the 'largest exchange for DSPs to bid on native'. Both claims are unverified. Commerce-native on third-party supply through Amazon DSP (nREC) and a Walmart product feed. Amazon named TripleLift a third-party native supply source for its REC units in December 2023; CSE membership is confirmed only by TripleLift. A combination of CTV non-standard formats (pause ads with DIRECTV, overlays, Product Spotlight) and a self-serve deal tool with MCP access (TL Direct). - **Risks.** Ownership and exit risk: Vista has held a majority since 2021 (five-plus years). There have been three CEO transitions (Berry to Clark in 2022, an interim Office of the CEO in 2024, Helmreich in 2025) and three layoff rounds. A change of control is plausible but not reported (inference). Supply-quality risk: MFA exposure was documented by Adalytics in 2024, and 1,755 non-EB intermediary seller IDs appear in sellers.json. Dependence on Amazon DSP for commerce growth, and on Google Open Bidding for a large share of seller IDs. Status of the 1plusX data asset is unclear: 1plusx.com did not serve content on 2026-09-27, and the platform page no longer names 1plusX. Agentic claims (TL Spark, MCP access) are young and undocumented beyond press releases. - **Best fit.** DSP-led buyers who want open-web in-feed native, or high-impact formats, without producing publisher-specific creative. Amazon DSP advertisers extending retail-media creative (nREC) to open-web and CTV supply. Brand and upper-funnel CTV buyers testing non-standard units (pause ads, overlays). - **Poor fit.** Advertisers who need audited, reproducible conversion or incrementality evidence. None is public. Buyers who require log-level data from every SSP as a condition of spend (on the ANA 2023 record). Performance advertisers seeking recommendation-widget, CPC-priced native of the Taboola, Outbrain or MGID type. TripleLift's model is SSP/CPM programmatic, although its pricing is not published (inference). - **Limits.** No audited or disclosed revenue, take rate or profitability, so scale cannot be compared with Taboola, Outbrain/Teads or Magnite. Native is no longer the company's lead positioning. It now presents itself as a 'Creative SSP' across CTV, retail media and the open web, and the 2025 year-end release contains no native-specific statement. Transparency record is mixed: the company did not provide log-level data to the ANA 2023 study, and Adalytics observed MFA delivery, including in PMPs. Publisher-side controls and native placement standards are not publicly documented. The knowledge base is login-gated. Every performance claim found is vendor-asserted or vendor-commissioned. | Dimension | Availability | Evidence | Observed | |---|---|---|---| | Supply access | live | A | Open-web, app, CTV and offsite retail-media supply. Vendor claims: 'over 5,000 direct publishers across 40 verticals', '99% of the top 100 publishers', '80% of the Comscore 200', '70% of our inventory via direct integrations (1-hop)'; all undated. The sellers.json snapshot (2026-09-27) has 7,941 entries: 3,816 PUBLISHER, 4,048 INTERMEDIARY, 77 BOTH. 2,295 IDs are Google Open Bidding (-EB) duplicates. Non-EB entries cover 1,268 distinct publisher/both domains. CTV includes programmatic pause ads with DIRECTV Advertising. [653, 505, 513] | | Format support | live | D | Web: Native, Image, Cinemagraph, Scroll, Window, Carousel, Reveal, Pharma, Adaptive Commerce Solutions, Outstream, Instream, Branded Video, Canvas Video, Display, Collection, Flipbook, Multi-Action, Vertical Video. CTV: CTV Spots, Product Spotlight, QR Spots, Social to CTV, Split Screen, Dynamic Overlay, L-Bar, Pause Ads. Native in-feed placements are described as appearing 'directly into the publisher's feed'. [654, 655] | | Advertiser workflow | live | C | Three routes are listed: DSP integrations, TL Direct self-serve, and managed service. TL Direct became generally available on 2026-06-10 for 'brands, buyers, publishers, and curators' to 'build and launch deals in minutes'. The release does not say whether buying bypasses a DSP, and no pricing is disclosed. The knowledge base lists DSP activation guides for DV360, Viant, Basis, Amazon and Yahoo. [520, 654, 519] | | Creative tools | live | C | Automated native rendering. Since 2014 it has used 'computer vision and dynamic templating' to render many executions from one asset set. The OpenRTB spec says 'TripleLift creates a native ad through the assets specified in the winning bid'. Custom creative formats and 'Creative Capture' are also marketed. The Forrester TEI covers 'Custom Creative Formats' with creative assembly 'at no additional cost'. [507, 506, 516] | | Optimisation | announced | D | TL Spark (launched 2026-04-21) is described as coordinating supply, creative, audiences, measurement and optimisation. It is said to be 'augmenting human planners and traders' and to reallocate spend 'toward what delivers the best outcomes'. The TL Spark page gives no live or beta status, no customer count and no methods. The platform page claims '+47% ROAS lift vs benchmarks' and '+50% more efficient CPC' without stating a basis. [521, 656, 518] | | Conversion-data access | unconfirmed | E | No public documentation was found of an advertiser pixel, a conversion API, or CPA/ROAS-based buying on TripleLift itself. Outcome figures (ROAS, new-to-brand) appear in retail-media case releases, where retailer data flows through Amazon DSP. The knowledge base is mostly behind a login. [518, 657, 519] | | Measurement support | live | C | Attention: Adelaide AU-guaranteed deal packages, CTV first (2023). Brand lift: one brewing campaign 'third-party verified by Cint'. Fraud: rate '0.4% (validated by Human)'. The company also uses proprietary eye-tracking. The Forrester TEI was commissioned by TripleLift. No named sales-lift or incrementality partner was documented on the platform page. [658, 517, 653, 516] | | Supply transparency | live | A | sellers.json is public, with zero confidential or passthrough entries (consistent with the '100% sellers.json transparency' claim), but 52% of entries are INTERMEDIARY (31% after removing Google Open Bidding duplicates). In the December 2023 ANA study TripleLift declined, or could not provide, log-level data in the required timeframe. Adalytics (Jan-2024 data) observed ads transacted via TripleLift for 8,451 brands on sampled MFA sites, including via PMP deals, despite a stated MFA block. [505, 89, 515] | | Publisher controls | unconfirmed | C | The Prebid adapter documents support for deals, the Floors Module, schain and GDPR/USP (GPP 'some'). Public native placement standards were unreadable: the Salesforce support page returned an error, and the new knowledge base (from 2026-09-08) is login-gated. No public documentation of creative blocklists or ad-quality controls was read. [509, 519, 508] | | Commerce capabilities | live | C | Amazon Native Responsive eCommerce (REC) on third-party supply, using ASIN assets such as price, star ratings and Prime logo (beta 2024-05-15 with Flywheel, Kepler, Tinuiti, SparkX). Joined the Amazon DSP Certified Supply Exchange on 2025-10-14 (nREC, OLV, CTV). Offers 'Walmart SKU ads via the product feed and TripleLift tags'. CTV Product Spotlight. Amazon's REC announcement of 2023-12-08 names TripleLift as a third-party native supply source; Amazon's CSE launch page (2024-06-13) names only Magnite, PubMatic and Index Exchange; no Amazon-side confirmation of TripleLift's membership was found. [659, 511, 657, 660, 510] | | Interoperability | live | C | Supports OpenRTB Dynamic Native Ads API v1.2 (docs updated 2026-09-14, including the standard Native 1.2 aurlsupport field), a legacy OpenRTB 2.3 native spec, and Prebid.js/Prebid Server ('display, video' media types). Google Open Bidding is used (-EB seller IDs). Amazon DSP CSE. TL Direct offers 'UI, API, and MCP-based access'. The IAB Tech Lab Agentic Advertising Management Protocols are cited only as 'proposed workflow standards', with no stated implementation. No AdCP support was found. [508, 506, 509, 505, 520, 521] | | Geographic reach | live | C | US-headquartered. Six offices reported in 2025. DACH expansion in August 2022. Year-end 2025 release: expanded into '15 international markets', a new Paris office, Canada and UK leadership, and a Numatec LatAm partnership. Audiences was US at launch, with AU, CA, FR, DE and UK planned for 2024 (rollout not verified). Multilocal (2026) is framed as a path to European DSP clients. [513, 661, 366, 662, 522] | Performance evidence: 272% ROI, $2.2M NPV, $3.70 ROAS, $1.3M incremental profit and $357K avoided labour over three years for 'Custom Creative Formats' (commissioned measurement, sponsor TripleLift commissioned Forrester Consulting) [516]; CTV campaigns deliver 'up to 26x industry benchmarks' (+6.3% brand favourability for one US steakhouse campaign) (vendor assertion, sponsor TripleLift) [517]; TripleLift CTV inventory performed 8% above Adelaide's CTV attention benchmark (commissioned measurement, sponsor Partnership announcement (TripleLift and Adelaide are commercial partners)) [658]; TripleLift Audiences 'regularly matches or exceeds the performance of cookies' (vendor assertion, sponsor TripleLift) [662]; Homepage: '3.6x brand lift in CTV', '7.9x ad recall vs benchmarks'; platform page: '+47% ROAS lift vs benchmarks', '+61% high attention scores' (vendor assertion, sponsor TripleLift) [654, 518]; Microsoft Store achieved '475 percent ROAS (unadjusted)' on native RTB (2014) (vendor assertion, sponsor TripleLift) [507]; Prime Day 2026: spend on TripleLift inventory via Amazon DSP +45.2% YoY globally (vendor assertion, sponsor TripleLift) [523] ### Sharethrough (Equativ) (acquired) *Sharethrough Inc. (sellers.json contact: 'Sharethrough Inc, 5455 Gaspe Ave #730, Montreal, Quebec, H2T 3B3, Canada'); per its counsel, all issued and outstanding shares were sold to Equativ on 2024-06-12* An early native SSP, bought by Equativ in 2024 and operating under the Equativ brand since 2025, where native is one channel among several. **SWOT (2026-09-27).** Native advertising pioneer founded in 2008 and bought by Equativ in June 2024, whose exchange still runs with native as one format among many. - **Strengths.** Open-web reach shows in the Sharethrough seller file, which listed 5,206 entries on 27 September 2026, including 2,769 publisher entries and 1,932 distinct publisher domains. (vendor documentation) [526] Seller lists (sellers.json) for both the Sharethrough and Equativ exchanges are public with zero confidential entries, and the 2022 supply policy requires ads.txt and sellers.json compliance. (vendor documentation) [526, 527, 528] Documentation covers the standard programmatic routes: OpenRTB Native 1.2, Prebid.js and Prebid Server adapters, and native asset specs for six named DSPs. (vendor documentation) [529, 530, 531] An exchange-wide block on made-for-advertising sites has run since May 2025, Equativ says, using Jounce Media lists across both the Equativ and Sharethrough exchanges. (vendor claim) [532] - **Weaknesses.** Native is one listed format among many: the April 2026 creative-enhancement launch does not name it, and the supplier API does not support native video. (vendor documentation) [53, 533, 529] Native creative follows each DSP's asset specs, and no route for conversion data or outcome optimisation at the exchange is documented publicly as of September 2026. (vendor documentation) [531, 529] All native performance figures found are vendor-published: Equativ claims 2.8x higher click-through rate than competitors' native on the same sites, with no period, sample or method given. (vendor claim) [534] The exchange's revenue share (its take rate) and any fee for curated deals are not documented publicly as of September 2026. (vendor documentation) [530, 535] - **Opportunities.** If OpenAI confirms the ChatGPT Ads partnership Equativ announced on 22 September 2026, its native-copy experience could win planning and creative work, while OpenAI keeps delivery decisions. (inference) [536, 537] If curation keeps driving growth, as Equativ said it helped do in Q1 2024, native could gain spend inside multi-format deal packages on its Maestro platform. (inference) [52, 538] - **Threats.** If buyers cut duplicate routes to the same inventory, one of the two exchange identities may lose spend, since all 2,704 Equativ seller IDs sit in Sharethrough's file. (inference) [526, 527] Google's alleged conduct, which Equativ's October 2025 lawsuit says significantly damaged its ad-server and exchange businesses, remains a risk unless a court grants relief. (inference) [539] - **What is sold.** Inside Equativ, the Sharethrough exchange is a supply-side ad exchange seat. Buyers reach it through DSP deal IDs or the open auction; publishers reach it through Prebid, the Prebid Server adapter or direct integration. Its inventory covers display, video, CTV and native, and native is rendered through 'Display Native' (TrueTemplate), which formats DSP-supplied native assets to match each site. Equativ packages the exchange with the Maestro curation platform (deal creation, meta deals, AI planning agent), creative enhancements (captions, sports ticker, countdown, shoppable overlays), GreenPMPs, CTV ad serving/SSAI and retail media. Since 2026-09-22 it also says it offers planning and creative support for ChatGPT Ads. - **Who buys it.** Agencies, trading desks and brands buying programmatically through DSPs (the support centre lists native specs for DV360, The Trade Desk, Yahoo DSP, Xandr, Amazon DSP and DeepIntent), and curators and self-serve buyers in Maestro (Equativ claims 500+ self-serve clients, 500+ agencies and all Big Six holding companies). On the sell side are publishers and app owners monetising through the exchange. - **How it earns.** Sell-side revenue share on cleared spend: Prebid documents that Sharethrough bids are returned 'net', with 'revenue sharing already taken into account'. The take rate is not published. Equativ also reports 'net recurring revenue' (definition not published): $100M in 2022 standalone and 'above $200M' combined at the June 2024 merger. Creative Enhancements are activated in Maestro 'with no fee'. No curation-fee schedule was found. - **Supply and data dependencies.** Depends on publisher and intermediary supply (5,206 entries in the Sharethrough seller file, 2,152 of them INTERMEDIARY and 1,135 of those named as Google Open Bidding 'OB' routes) and on DSP demand (native assets are built to DSP specs). Quality relies on third-party vendors: HUMAN for IVT, Confiant for malware, Zvelo for brand safety, AdVerif.ai for misinformation, White Bullet for piracy and Jounce Media for MFA lists. Addressability relies on alternative IDs and data partners (30+ data providers claimed). The ChatGPT Ads work depends entirely on OpenAI's ads system, which OpenAI says 'controls all delivery decisions'. - **Differentiation.** Lineage: the vendor says it brought in-feed native to content publishers and that TrueTemplate renders DSP native assets in each site's style (vendor claim, not independently verified). Combined sell-side and curation stack: SSP, publisher ad server, CTV SSAI, Maestro curation with AI planning agent, retail media and creative enhancements. Supply-quality posture: exchange-wide MFA block using Jounce lists, fully non-confidential seller files, HUMAN pre-bid IVT filtering, GreenPMPs. Early claimed position as a ChatGPT Ads planning/creative partner that cites native-copy expertise (not confirmed by OpenAI in sources opened). - **Risks.** Brand dilution: Sharethrough's native identity is being absorbed. The native marketing page redirects to a generic advertiser page, and 'native' appears only as a list item in Equativ positioning. Supply-path duplication: all Equativ seller IDs are cross-listed in the Sharethrough seller file, so buyers doing SPO may see two routes to the same inventory. Ownership: private-equity (Bridgepoint) ownership since 2023 implies an eventual exit event; no sale process was found as of the cutoff. Exposure to Google: Equativ's own lawsuit (2025-10-13) alleges damage to its ad-server and exchange businesses. Remedies or damages are uncertain. Headcount boilerplate moved from 'over 750 across 20 countries' (2025) to 'over 720 across 19 countries' (September 2026), which may indicate consolidation (inference). The ChatGPT Ads role depends on OpenAI, which states that its ads system controls all delivery decisions. - **Best fit.** Programmatic buyers who want native in-feed placements on open-web publisher sites through their existing DSP (DV360, The Trade Desk, Yahoo DSP, Amazon DSP, DeepIntent) with deal IDs or curated PMPs. Omnichannel buyers who want native, display, video and CTV curated in one Maestro deal, especially in the US, Canada, France and the rest of Europe. Buyers whose supply-quality requirements include MFA exclusion and transparent seller files. - **Poor fit.** Performance advertisers who need a native-specific CPC/CPA buying platform with conversion optimisation (a recommendation-widget platform fits better). Buyers of native video through OpenRTB native (not supported on the Equativ supplier API). Anyone needing independently measured native outcome performance: none was found. - **Limits.** Native is now one format inside an omnichannel SSP and curation stack. The 2026 enhancement launch lists in-stream/CTV, out-stream and display, not native. Native creative is DSP-built ('We follow DSP specs for Native inventory'), so creative quality and optimisation sit with the buyer's DSP. The Equativ supplier OpenRTB API documents 'Native video is not supported at this time'. No documented conversion-data path or exchange-side outcome optimisation for native. No public take rate or curation fee schedule. Equativ's 'net recurring revenue' is not defined. Much native documentation is legacy: the support section is titled 'Outdated', and the STX snapshot is undated. | Dimension | Availability | Evidence | Observed | |---|---|---|---| | Supply access | live | C | Open-web publisher and app supply through the Sharethrough exchange seat. Its sellers.json has 5,206 entries: 2,769 PUBLISHER, 2,152 INTERMEDIARY, 285 BOTH and 0 confidential, with 1,932 distinct publisher/both domains excluding google.com. The Equativ seat has 2,704 entries. Equativ marketing claims 'over 500K publishers, apps, and leading media channels', '37B daily impression requests' (Display Native page) and analysis of '400K+ sites and apps'. Supply policy requires that a majority of traffic come from the US, UK, Canada and EMEA. [526, 527, 663, 534, 528] | | Format support | live | C | Native: 'Display Native' built on TrueTemplate, marketed as matching fonts, sizes and colours per site. The Equativ supplier OpenRTB API supports Native Ads 1.2 and states 'Native video is not supported at this time'. The Equativ ad server has Native (RTB), Native web/app and Easy Native publisher templates. Non-native formats are display, OLV outstream/instream, CTV (VAST 2.0-4.0; no VPAID or MRAID), shoppable and high-impact. The Prebid adapters for both seats list display, video and native. [534, 531, 530, 664, 529, 665] | | Advertiser workflow | live | C | Programmatic only: DSP deal IDs (help articles on accepting 'a Sharethrough or Equativ Deal ID in DV360' and opting into the Sharethrough exchange), plus the Maestro curation UI for deal creation, 'meta deals', a data marketplace, custom business rules, live reporting and a reporting API. Equativ claims 500+ self-serve clients and 30K+ deals. No native-specific managed service or self-serve native campaign builder is documented. [531, 538, 663] | | Creative tools | live | C | For native, 'We follow DSP specs for Native inventory': assets come from the buyer's DSP and TrueTemplate renders them. Creative Enhancements (automatic headlines, dynamic captions, sports ticker, countdown, shoppable overlays/QR) cover in-stream/CTV, out-stream and display, and native is not named. The legacy Headline Analyzer (headlines.sharethrough.com) still resolves (HTTP 200); its functionality was not inspected. The Native Ad Generator sits behind a login. [531, 533, 54] | | Optimisation | live | D | Supply-side curation and planning: the Maestro Media Planning Agent turns briefs into PMP configurations and auto-bundles meta deals. Agents for optimisation, troubleshooting and reporting were announced as an expansion. Bid optimisation toward clicks, CPA or ROAS sits with the buyer's DSP; no exchange-side outcome optimiser for native is documented. The legacy guide (outdated section) recommended $4-6 CPM starting bids and cited CTRs 'around 1%'. [666, 667] | | Conversion-data access | unconfirmed | E | No documentation found of conversion pixels, CAPI-style conversion feeds or advertiser conversion data flowing to the exchange. Conversion data is assumed (not verified) to stay in buyers' DSPs. | | Measurement support | live | C | The OpenRTB native spec requires eventtrackers (impression via img and js) and native.privacy. HUMAN (TAG-accredited) filters IVT pre-bid and post-bid, and the Maestro page lists a reporting API. Enhancement claims cite a TVision eye-tracking study and in-house research. No MRC accreditation of Equativ or Sharethrough measurement was found. [529, 668, 663, 533] | | Supply transparency | live | C | Both seller files are public, with 0 confidential entries in either. schain is supported (Prebid; the OpenRTB spec supports source.schain). The OpenRTB spec includes a regs.ext.dsa transparency object. Onboarding requires ads.txt and sellers.json compliance, and the supply policy bans arbitrage and resale. MFA was blocked exchange-wide from 2025-05-16 using Jounce Media lists. Complication: all 2,704 Equativ seller IDs are cross-listed in the Sharethrough file, and the Equativ file lists 'Sharethrough SSP Global Supply' as an INTERMEDIARY, so a buyer can meet the same publisher through two exchange identities. [526, 527, 530, 529, 528, 532] | | Publisher controls | live | C | Publisher-set floors (Prebid Floors Module support; format-level native bidfloor in OpenRTB), advertiser-domain and category blocking (Prebid ORTB blocking 'partial' for Sharethrough and 'yes' for Equativ), and publisher-defined native templates in the Equativ ad server built from macros ([native_title], [native_imageUrl], etc.). A PubMaestro publisher UI exists (pubmaestro.sharethrough.com, HTTP 200) but was not inspected. [530, 664, 529, 665, 54] | | Commerce capabilities | live | D | Equativ Shoppable Ads (launched 2025-10-28) cover CTV via QR, video, display and mobile in-app. Retail media comes through the Kamino Retail acquisition (completed by January 2025). No native commerce/product-listing unit on open-web native placements is documented. [669, 538] | | Interoperability | live | C | OpenRTB 2.6 with Native Ads 1.2 on the supplier API and OpenRTB 2.6 fields for bidders. Sharethrough has Prebid.js and Prebid Server adapters (GVL ID 80); an optional eqtvNetworkId routes Sharethrough-adapter requests to 'the Equativ exchange'. Equativ has a Prebid.js adapter (GVL 45, no PBS adapter, deals 'no'). Equativ 'direct integration with more than 70 DSPs'. AdCP and the Agentic RTB Framework are named as open protocols for customer agents (announced). An MCP-server page appears in site navigation but was not opened. [530, 664, 529, 670, 538, 666] | | Geographic reach | live | D | Equativ reports a team in 19 countries (September 2026 boilerplate). The ChatGPT Ads partnership covers 12 listed markets (BE, BR, CA, FR, DE, IT, JP, MX, PL, ES, UK, US). The Sharethrough supply policy requires a majority of site traffic from the US, UK, Canada and EMEA. Equativ is a founding member of the European Media Marketplace. The US was 40% of Equativ revenue in 2022, before the deal; no post-merger regional split was found. [536, 528, 671, 672] | Performance evidence: Display Native delivers '1.4x better performance than industry native CTR benchmarks' and '2.8x higher CTR compared to native ads delivered on the same sites as the competition' (vendor assertion, sponsor Equativ) [534]; Legacy Sharethrough homepage figures: '91% more likely to pay attention', '14% more likely to comprehend', '40% increase in VCR', '2X lift in CTR' for various enhancements (vendor assertion, sponsor Sharethrough/Equativ) [54]; Sports Ticker: viewers '14% more likely to pay attention to a CTV commercial' and '19% more likely to pay attention to the full commercial' (commissioned measurement, sponsor Equativ (study attributed to an eye-tracking study with TVision; commissioning terms not stated)) [533]; Captions raise comprehension by 56%; Countdown makes viewers 43% more likely to remember a promotion start date; Shoppable Ads show a 9% lift in brand-name attention and 30% and 53% comprehension gains (vendor assertion, sponsor Equativ) [533, 669]; Legacy STX: 'Every impression is 100% viewable (verified by MOAT)'; CTRs 'around 1%' (vendor assertion, sponsor Sharethrough) [673, 667]; ChatGPT Ads through Equativ: advertiser testimonial (BNP Paribas, via Havas Media France) that messages were adapted to conversation contexts (vendor assertion, sponsor Equativ) [536] ### The wider universe The matrix is drawn from a wider list of 68 companies. For each one, a primary page was opened. The list covers the core market, nearby substitutes and new surfaces. It is a reasoned sample, not a census. Adult-only, gambling-only and ad-fraud networks were left out on purpose. Companies that surfaced only as names were listed but not scored. A company was scored if two tests held. First, it sells or delivers open-web native as a main or clearly separable business, or native is bought through it at scale. Second, at least four of the twelve dimensions had evidence, counting a dimension documented as not applicable. **Vendor universe** | Company / product | Legal entity / owner | HQ | Scope group | Native formats | Buying routes | Main geographies | Status at cutoff | Src | |---|---|---|---|---|---|---|---|---| | Taboola | Taboola.com Ltd. (Nasdaq: TBLA) | US business address (New York); incorporated outside US per EDGAR code L3 | core native | native recommendation units; native, display, carousel, vertical, motion, performance video, app promotion | self-serve (Realize), managed; programmatic not verified on opened pages | global; US-led | active (8-K Q2 2026 results 2026-08-05) | [674, 386, 30, 675] | | Realize (Taboola ad platform) | Taboola.com Ltd. | US (New York) | core native | native, display, performance video, vertical, carousel, app promotion across 'mail, mobile, and premium editorial environments' | self-serve; managed (dedicated account management for agencies); CPC/CPA objectives | global (8 language versions) | active (realize.com live 2026-09-27; cited in 8-K 2026-08-05) | [386, 30] | | Teads Holding Co. (Outbrain lineage) | Teads Holding Co. (Nasdaq: TEAD), Delaware; formerly Outbrain Inc. until 2025-06-09 | US (New York) | core native | native recommendation (Outbrain lineage), in-read native video, display, CTV; 'Ad Formats' and Teads Studio | managed (Enterprise), self-serve (Growing Brands), programmatic via ad manager/Zemanta DSP | global; 50 markets claimed; 30+ countries staffed | active (8-K Q2 2026 results 2026-08-06) | [676, 271, 387] | | Zemanta (Teads DSP) | Teads Holding Co. | US (parent) | core native | native DSP buying across 'extended Teads inventory' | self-serve DSP / programmatic | global | active (product absorbed) (zemanta.com states 'Zemanta / Outbrain DSP is now part of the Teads offering' (accessed 2026-09-27)) | [563] | | MGID | MGID Inc. | not stated on opened page | core native | native recommendation widgets, Exit Story, rich media, contextual targeting, internal recirculation | self-serve, managed, programmatic | global | active (© 2026 MGID Inc.) | [427, 437] | | Revcontent | RevContent, LLC | not stated on opened page | core native | native widgets; cookieless native; email monetization | self-serve ('no minimum spending requirements'); managed/programmatic not detailed | not stated (US publisher logos) | active (© 2026 RevContent, LLC) | [367] | | Dianomi | Dianomi (UK Company Reg. 4513809) | UK | core native | native placements next to financial/business editorial; premium native and contextual display; Apple News partnership | marketplace/managed; pay-per-engagement | US, UK, Australia (offices NY/London/Sydney) | active (© Dianomi 2026) | [469] | | Readpeak | Readpeak Inc. (per footer) | unresolved (footer 'Inc.'; Nordic/European operations) | core native | Content Native, Product Native, Instant Article, Video Native | self-serve programmatic native platform | Europe (Nordics, Benelux, DACH, FR, IT) | active (© 2026 Readpeak Inc.) | [625] | | Nativo (Life360 Ads) | Nativo, Inc. (Delaware), owned by Life360, Inc. | US (El Segundo, CA) | core native | custom native formats via advanced ad renderer; sponsored content | managed, programmatic (SSP), self-serve | US | acquired (active under Life360) (acquisition completed 2026-01-02, reported in a filing of 2026-01-05 (~$120M, 65% cash/35% stock); nativo.com 301 to ads.life360.com) | [494, 677] | | TripleLift | TripleLift (ownership not stated on page) | US (New York) | core native | native Image, Cinemagraph, Scroll, Window, Carousel, Reveal, Pharma; also outstream/instream video, CTV, retail media, display | TL Direct (self-serve), DSP integrations, managed service | North America; US/Canada/UK named for some formats | active (© 2026 footer) | [654] | | Sharethrough | Equativ Group (Sharethrough 'An Equativ Group Company') | France (Equativ); Sharethrough HQ not stated | core native | native, enhanced display/video/CTV ad enhancements, curation | programmatic exchange (SSP), direct | global | acquired by Equativ (2024-06-12); operates under the Equativ brand since 2025-06-09 | [54] | | Kargo | Kargo | not stated on opened page | core native (positioning now 'high-impact') | high-impact creative formats across display, OLV, CTV, social, retail media; native not foregrounded on homepage | managed, programmatic; 'invitation-only marketplace' | not stated | active (homepage live 2026-09-27 ('Over two decades strong')) | [678] | | StackAdapt (native channel) | StackAdapt Inc. | Canada | adjacent-substitute (DSP with native channel) | in-feed native, content recommendation, in-ad native, native video, outstream, mobile in-app native | self-serve DSP, managed | global (EMEA, NORAM, APAC) | active (© 2014–2026 StackAdapt Inc.) | [679, 680, 681] | | Yieldmo | Yieldmo | not stated on opened page | core native (historic); now 'creative and media results company' | interactive display, enriched video, custom builds, CTV; native not named on homepage | programmatic; end-to-end managed service; contextual curation | not stated | active (© 2026 Yieldmo) | [682] | | Media.net | Media.net (ownership not stated) | not stated on opened page | core native (contextual/native SSP) | 'unique formats that match how users consume content in walled gardens'; contextual | SSP/programmatic with DSP partners (The Trade Desk, Criteo, DV360) | global | active (© 2025 Media.net) | [683] | | Adyoulike (OpenWeb) | OpenWeb (acquired April 2022) | France/UK/US offices (Neuilly-sur-Seine, London, New York) | core native | Automatic Stories, Native Stories, Native Social, Native Video, Native Display, Retail Stories | programmatic (50 DSPs connected claimed); direct publishers | US, UK, France | active; Opera link unconfirmed (© 2026 adyoulike.com; no Opera mention) | [369, 66] | | plista | unknown | Germany (historic) | core native (historic) | content recommendation/native widgets (historic) | unknown | DACH (historic) | reported ceasing operations (March 2024); domain has no address records at the cutoff | [372, 684, 371] | | Engageya | unknown | Israel (historic) | core native (historic) | content recommendation widgets (historic) | unknown | unknown | retired (service discontinued 2024-09-30 per its homepage notice, captured by a third party) | [372, 370] | | ZergNet | Static Media | US | excluded from core (organic editorial recommendation) | editorial recommendation widgets, 'ad-free', '100% pure editorial content' | publisher network participation; no paid ad buying route stated | US | active (not a paid native seller) (© 2026 ZergNet.com; About page) | [685] | | Content.ad | unknown | US (historic) | core native (historic) | native widgets (historic) | unknown | unknown | unconfirmed; domain has no address records at the cutoff; closure undated | [372] | | NewsBreak | Particle Media, Inc. | US | core native (app feed) | in-feed native, in-article, related article, in-feed/immersive video, carousel, newsletter | self-serve (Target CPA; CPM/CPC), managed support | US (local) | active (© 2026 Particle Media, Inc.) | [686] | | SmartNews | SmartNews, Inc. | Japan/US | core native (app feed) | in-feed native ('Ads as Content') | SmartNews Ads self-serve (launched July 2025), managed | Japan, US | active (corporate site news items dated 2025) | [687] | | Primis | Primis; page states part of Interpublic Group (IPG) and Omnicom portfolio | Israel (Tel Aviv) | core native (native video discovery) | video discovery player/units for publishers; pre-roll opportunities | programmatic marketplace/auction | global | active (© 2026 footer); not scored: no paid native unit documented (round-2 close-out) | [688] | | Connatix / JWP Connatix / JWX | JWP, Inc. (JWX) | not stated on opened page | core native (in-article/outstream video) | publisher video platform; premium ad formats; QuantumPath advertising orchestration | programmatic, managed | not stated | merged/rebranded (active as JWX) (redirect chain connatix.com -> jwpconnatix.com -> jwx.com observed 2026-09-27; © 2026) | [689] | | Seedtag | Seedtag | not stated on opened page | core native (contextual in-image/in-article lineage; now 'Contextual OS') | display, OLV, CTV via contextual AI 'Liz'; in-image/in-article not named on homepage | curation platform; managed/programmatic not detailed | EN/ES/PT/FR/DE/IT markets | active (homepage live 2026-09-27) | [690] | | GumGum | GumGum, Inc. | US | core native (in-image/in-screen lineage; now high-impact/SSP) | high-impact (Hang Time, desktop skin, in-video squeezeback), CTV, OLV, display | managed, programmatic ('Top 15 global omnichannel SSP') | 19 markets; US, Japan, France, Spain, Denmark sites | active (© 2025 GumGum, Inc. footer) | [691] | | Undertone | Perion Network Ltd. (Nasdaq) | Israel | core native/high-impact (brand absorbed) | high-impact formats (historic); Perion One channels: CTV, DOOH, audio, retail, web, search | managed/platform (Perion One) | 40+ markets | brand retired/absorbed (undertone.com 301 to perion.com; no Undertone mention; © 2026 Perion Network Ltd.) | [692] | | Sovrn | Sovrn, Inc. | US | adjacent-substitute (exchange + affiliate) | omnichannel ad exchange; native not named in extract; commerce/affiliate tools; Signal attention yield | programmatic | not stated | active (homepage live 2026-09-27) | [693] | | Ezoic | Ezoic Inc. | US | core native (publisher monetization incl. native placements) | native 'woven into the article', sticky video, anchor, rewarded, interstitial | direct campaign launch via bid system; sales team | open web (not stated) | active (© 2026 Ezoic Inc.) | [694] | | PropellerAds | Propeller Ads Ltd (Cyprus, HE353513) | Cyprus | adjacent-substitute (performance/push network) | popunder, push, in-page push, interactive, Telegram Mini Apps, social; native not headline format | self-serve (NIKO AI), managed, DSP/oRTB | 195+ geos; Tier-1 and emerging | active (homepage live 2026-09-27; founded 2011) | [695] | | Adsterra | AD MARKET LIMITED | Cyprus | adjacent-substitute (performance network with native banners) | Native Banners, popunder, Social Bar, in-page push, interstitial, banners, SmartLink | self-serve, managed, RTB | 248 geos | active (© 2013–2026 Adsterra.com) | [696] | | AdNow | RENODO MEDIA LTD | Bulgaria (Plovdiv) | core native (long-tail widgets) | native widgets: content discovery, product recommendations | self-serve (adnow.one), SSP for publishers | 114 countries claimed; many localized sites | active (© 2026) | [697] | | BidVertiser | BidVertiser / Bpath, LTD | not stated on opened page | adjacent-substitute (direct/performance network) | Native Ads, direct navigation, pop under, push, in-page push, contextual pops | self-serve | not stated | active ('Serving you Since 2008') | [698] | | Google AdSense native | Google LLC | US | core native (platform) | In-feed, In-article, Multiplex (content-recommendation-style grid) | publisher self-serve; demand via Google Ads auction | global | active (help page shows all three formats with no deprecation notes (accessed 2026-09-27)) | [699] | | Google Ad Manager native | Google LLC | US | core native (platform) | native styles (element-based, publisher-styled) with required 'Ad/Sponsored' badge and AdChoices | direct-sold and programmatic via Ad Manager; OpenRTB native not verified on opened page | global | active (help page accessed 2026-09-27) | [700] | | Amazon Ads display (formerly Sponsored Display) | Amazon.com, Inc. (Amazon Ads) | US | adjacent-substitute (retail media off-site display) | display on Amazon properties (Twitch, Fire TV, Echo Show) and 'premium placements across the open internet'; Amazon-generated or own creative | self-serve via Ads Console Campaign Manager | not stated | active (renamed) (page states 'Sponsored Display is now known as display ads' (accessed 2026-09-27)) | [701] | | Amazon Publisher Services (APS) | A9.com LLC (Amazon) | US | adjacent-substitute (publisher supply infrastructure) | TAM, UAM header bidding; Connections Marketplace; Signal IQ; Amazon Publisher Cloud; formats not specified | server-to-server marketplace for publishers | not stated | active (© 2025 A9.com LLC) | [702] | | Amazon Sponsored Products | Amazon.com, Inc. (Amazon Ads) | US | adjacent-substitute (retail search native) | sponsored product units in shopping results and product pages; also other Amazon O&O and third-party destinations | self-serve, CPC, no monthly fees | multi-country (page lists Americas, EU, UK, MENA, AU, IN, JP, SG, ZA) | active (page accessed 2026-09-27) | [211] | | Microsoft Advertising native (Audience ads) | Microsoft Corporation | US | core native (platform) | native audience ads matching 'news feed, article page, inbox, or content recommendation area'; display audience ads | Microsoft Advertising self-serve | not stated; 'up to 1 billion customers' | active (page accessed 2026-09-27) | [703] | | Microsoft pubCenter | Microsoft Corporation | US | core native (publisher side) | native/display for publishers (not verified) | publisher self-serve (not verified) | unknown | unconfirmed (pubcenter.microsoft.com HTTP 403 on 2026-09-27) | [372] | | Meta Audience Network native | Meta Platforms, Inc. | US | core native (in-app platform) | Native Ad API units: Sponsored label, AdOptionsView, advertiser name, title, CTA, image, icon | Meta Ads auction demand; publisher SDK integration (iOS/Android) | global | active (developer docs accessed 2026-09-27) | [704] | | AppLovin MAX native | AppLovin Corporation | US | core native (in-app platform) | native ads (manual integration); Ad Placer auto-insertion into RecyclerView streams | MAX mediation/auction; templates deprecated (no-fill from end Q2 2025) | global | active (docs accessed 2026-09-27) | [705] | | Unity Ads / LevelPlay native | Unity Software Inc. | US | core native (in-app platform) | native ads in LevelPlay SDK (Android, iOS, Flutter, React Native) | LevelPlay mediation; Unity/ironSource exchanges | global | active (LevelPlay native ads documented at cutoff; legacy ironSource URL redirects) | [706, 372, 707] | | Digital Turbine | Digital Turbine, Inc. (Nasdaq: APPS) | US (Austin, TX) | adjacent-substitute (on-device app distribution / in-app) | on-device integrations, Launchpad app distribution (2026-06-02); native format not specified | platform/managed via device and app partners | global (1B+ devices claimed) | active (FQ1 2027 results 2026-08-04) | [708] | | Meta Feed/Reels (Facebook, Instagram) | Meta Platforms, Inc. | US | adjacent-substitute (social in-feed) | in-feed and Reels ads | Ads Manager self-serve, managed, Marketing API (routes not on opened page) | global | active (Q2 2026 results 2026-07-29) | [709] | | TikTok | TikTok (ByteDance; legal entity not on page) | not stated | adjacent-substitute (social in-feed) | in-feed full-screen sound-on ads on For You feed | TikTok Ads Manager self-serve; minimum budgets $30 NA/EMEA, $20 APAC | global; 'over a billion users' | active (© 2026 site) | [710] | | LinkedIn Sponsored Content | LinkedIn Corporation (Microsoft) | US | adjacent-substitute (professional in-feed) | single image, video, carousel, document, event, thought leader, CTV, article/newsletter ads | Campaign Manager self-serve; LinkedIn Audience Network | global | active (page accessed 2026-09-27) | [711] | | X (formerly Twitter) | X Corp. | US | adjacent-substitute (social in-timeline) | promoted posts (not verified) | X Ads self-serve/managed (not verified) | global | unconfirmed (business.x.com HTTP 402; developer.x.com 402; help.x.com 403 (2026-09-27)) | [712] | | Reddit | Reddit, Inc. (Delaware) | US (San Francisco) | adjacent-substitute (community in-feed) | in-feed/conversation placements (formats not verified on opened page) | Reddit Ads self-serve/managed (not verified) | global | active (entity); formats unconfirmed (EDGAR 8-K results filings 2026-07-30, 2026-04-30; marketing domains blocked) | [713, 372] | | Pinterest | Pinterest (entity not on page) | US | adjacent-substitute (discovery in-feed) | image, video, carousel, collections, showcase, quiz, idea, premiere spotlight | Ads Manager self-serve; managed consultation | US site; global | active (page accessed 2026-09-27) | [714] | | Snap | Snap Inc. | US | adjacent-substitute (social/stories) | single image/video, story ads, collection, AR lenses, Spotlight, dynamic product ads | Snapchat Ads Manager self-serve; managed/partners | 25+ countries claimed | active (page accessed 2026-09-27 (data footnotes 2024)) | [715] | | Google Demand Gen / Discover | Google LLC | US | adjacent-substitute (feed-style across Google surfaces) | image (1.91:1, 4:5, 1:1), video, carousel (2–10 cards), product feeds on YouTube incl. Shorts, Discover, Gmail, Maps, GDN | Google Ads self-serve; 'social-like buying flow' | global | active (help page; GDN campaigns migrating from June 2026) | [716] | | Criteo (commerce media) | Criteo S.A. | France | adjacent-substitute (commerce media / retail media) | onsite sponsored products, offsite commerce display; Commerce Growth, Commerce Max (DSP), Commerce Yield (retail media), Commerce Grid (SSP) | self-serve and managed DSP/SSP | global | active (© Criteo 2026) | [717] | | Amazon DSP | Amazon.com, Inc. (Amazon Ads) | US | adjacent-substitute (DSP) | audio, display, online video, Streaming TV, in-store; Prime Video, Twitch, Fire TV, Amazon Publisher Direct, third-party exchanges | self-service; managed ($50k typical minimum); RTB, PMP, private auction, programmatic guaranteed | US, CA, MX, BR, 8 EU, 2 ME, 4 APAC (per page) | active (page accessed 2026-09-27) | [718] | | The Trade Desk | The Trade Desk, Inc. | US | adjacent-substitute (DSP with native channel) | Native listed among channels (CTV, display, native, audio, video, DOOH, mobile) | self-service DSP (Kokai) | global | active (Kokai Zuma release 2026-08-27 per page; © 2026) | [719] | | Google Display & Video 360 | Google LLC | US | adjacent-substitute (DSP) | display, video, YouTube (Google Preferred, Reserve, TrueView), CTV; native not named on opened page | programmatic DSP with automated bidding | global | active (page accessed 2026-09-27) | [720] | | OpenAI — ads in ChatGPT | OpenAI | US | emerging AI surface | labeled sponsored unit at bottom of answers (OpenAI); via StackAdapt: 1:1 256x256 image + headline + body 'below relevant conversations as clearly labeled, separate responses' | OpenAI direct (self-serve Ads Manager since May 2026) and agency and technology partners, including StackAdapt | 40+ countries (OpenAI, September 2026) | live (announced 2026-01-16; US test began 2026-02-09; multi-country by 2026-09) | [240, 681, 679, 721, 722] | | Perplexity ads | Perplexity AI | US | emerging AI surface | sponsored follow-up questions; paid media beside answers; 'sponsored' label | direct with launch partners (Indeed, Whole Foods Market, UM, PMG) | US to start | retired (phased out during 2025; FT reported 2026-02-17 no plans to return) | [723, 372, 724, 247] | | Google AI Mode / AI Overviews ads | Google LLC | US | emerging AI surface | Conversational Discovery ads; Highlighted Answers (sponsored suggestions in AI Mode lists); AI-powered Shopping ads; Business Agent for Leads | Google Ads via AI Max for Search, AI Max for Shopping, Performance Max; Direct Offers pilot (Jan 2026) | not specified | testing/announced (announced 2026-05-20) | [725, 258] | | Microsoft Copilot ads | Microsoft Corporation | US | emerging AI surface | ads inside Copilot answers (formats not listed on the page opened) | Microsoft Advertising campaigns (unverified) | unknown | live (vendor-reported metrics, Feb-May 2025) | [372, 726, 703, 254] | | Amazon Rufus / Alexa for Shopping ads | Amazon.com, Inc. | US | emerging AI surface | Sponsored Products and Sponsored Brands prompts that can open or appear in Alexa for Shopping (formerly Rufus) conversations | Amazon Ads, CPC, auto-enrolled | US | live (US general availability 2026-03-25) | [727, 211, 255, 256, 255] | | Dable | Dable Inc. | South Korea (Seoul) | core native (unseeded; Asia) | DNA content-ad network native units; wheres.ai app DSP; SalesBoost retail media | network/managed; DSP | Korea, Asia | active (© 2026 Dable Inc.) | [728] | | popIn Discovery | popIn Corporation (ownership not shown) | Japan (Tokyo) | core native (unseeded; APAC) | native and display via DSP with funnel prediction | DSP; exchanges incl. Google ADX, Microsoft, Yahoo, Amazon, MGID, Xandr | Japan, APAC (JA/EN/ZH/KO) | active (site live 2026-09-27) | [729] | | LiveIntent | Zeta Global (acquisition closed 21 October 2024) | US (not stated on page) | core native (unseeded; email/newsletter native) | Native Curated Packages, Native Ad Blueprints in email/newsletters | programmatic and direct-sold | not stated | acquired (closed 2024-10-21) | [730, 731] | | EX.CO | EX.CO | not stated on opened page | core native (unseeded; in-content video) | in-content video player, contextual video recommendations, premium video content marketplace; CTV | programmatic (ML yield engine), managed | global | active (site live 2026-09-27 (data to mid-2025)); not scored: no live paid native unit documented; Playbuzz branded content ended by 2019 (round-2 close-out) | [732] | | Bidtellect | Simpli.fi | US | core native (unseeded; native DSP absorbed) | contextual and native advertising capabilities within Simpli.fi | Simpli.fi programmatic platform | US | acquired ('Bidtellect is now part of Simpli.fi' (© 2023)) | [733] | | Strossle / Azerion Native | Azerion Group ('Powered by MGID') | Sweden (Stockholm) / Netherlands (Azerion) | core native (unseeded; Nordic/European) | native/content recommendation for publishers and advertisers | via Azerion; MGID technology | Nordics, Europe | rebranded/absorbed (strossle.com notice observed 2026-09-27) | [437] | | Yahoo native inventory (via Taboola Realize) | Yahoo Inc. (supply); sold through Taboola.com Ltd. | US | core native (supply partner) | native units on Yahoo mail/editorial properties listed as Realize placements | via Realize self-serve/managed | US-led | active (as Realize placement) (realize.com lists Yahoo among placements (2026-09-27)) | [386] | ## 10. Enterprise adoption *Enterprise buyers can now see more of the programmatic supply chain than at any time since 2016. They can see almost none of the native one. What they can check, test and take with them decides if native gets a line on the plan.* > **In short.** > The two scaled native platforms report different kinds of buyer. Taboola had about 2,200 Scaled Advertisers in the fourth quarter of 2025, working with it directly or through agencies; a significant majority of its revenue came from those working with it directly. Teads Holding Co. stresses master service agreements with holding companies. It also cites about 50 joint business partnerships that average $4 million a year. Neither filing has an outcome guarantee for buyers, or a published spending floor. Both do describe campaigns priced on results, where the buyer pays per acquisition. On their native self-serve platforms, both show placements only after the ads run. Buyers get site-level reports once a campaign has data. Controls start with lists of sites to exclude. The public documentation shows no list of the sites on offer before a campaign. Teads' programmatic buying tool, Outbrain DSP, does document publisher include lists at setup [264]. Some marketers opted into the ANA's log-level benchmarks. Among them, the median share of spend on made-for-advertising sites fell from 10% in 2023 to under 1% in 2025. That is a result for those who took part. It says nothing about native supply. The playbook claims native is being bought as outcome infrastructure. In this paper's reading, that is still an aspiration. The contracts and controls on record describe a click business. This chapter turns the evidence into real choices. They are the ones a buyer, an agency and a publisher face. It treats the author's Performance & Native playbook as a set of claims to test. Those claims are about how enterprise buying works. It also re-checks the playbook's market facts against primary sources at the cutoff. ### Who buys, and through what At the cutoff, the record shows four enterprise buying routes. Managed service, run by the vendor's own sales and key-account teams. The vendor's self-serve buying tool. At Teads it sits under master service agreements with holding companies. At Taboola it is Realize, with an optional Claude Skill front end. DSP and programmatic links through OpenRTB Native. And, since June 2026, publisher-side native supply open to SSP demand through Teads' EngageOS with Magnite's demand server [46, 1, 153, 265, 266, 158, 267, 268]. A DSP is the software buyers use to bid for ad space. An SSP is the tool sellers use to offer it. Taboola's and Teads' product pages state no spending floor. The ranges quoted for them come from third-party blogs. MGID's help centre sets a $100 minimum deposit for self-serve accounts. It recommends $650, and its blog gives a $50 minimum campaign budget [269, 270]. The two scaled platforms describe different customers. Taboola's annual report counts about 2,200 Scaled Advertisers, working with it directly or through agencies, and says a significant majority of its revenue came from those working with it directly rather than through an agency. Its ten largest clients made up under 10% of revenue [1]. Teads describes master service agreements with agency groups for its buying tool. It also describes a team just for large global brands. And it cites about 50 joint business partnerships. They average $4 million of spend a year. In 2025, 93% of its revenue came direct from buyers on its own platforms [46]. Its second-quarter 2026 results listed renewals with Stellantis, LVMH, Warner Bros. and Dyson. The same release suspended guidance, its public forecast. The reason was swings in direct-response and small-business demand [271]. Teads reported the renewals next to weakness in direct response and small business. Its reports do not show enterprise revenue or profit on their own. No single "holding-company corridor" story fits both. > **The playbook's operating model, labelled.** > The playbook carries some sales maths. It starts with 120 qualified opportunities at a 20% close rate. It states that this yields 23 deals at $144,000–230,000 annual contract value. But 120 at 20% is 24, which would give $3.46–5.52 million of contract value. The page's $3.3 million matches 23 deals at the lowest value. The playbook also gives $3.25–4.7 million of booked revenue per full-time seller. And it cites the author's own track record at Hearts & Science and Verve Group. These figures are the author's illustrative operating model and his own claims. The page labels them as such. No industry benchmark found in this research supports or contradicts them. They are not used as evidence anywhere in this paper [272]. ### What the playbook got right, and what has moved Twenty market and competitive facts in the playbook were re-checked against primary sources. Most structural claims hold. Outbrain closed its purchase of Teads on 3 February 2025. The firm became Teads Holding Co. in June [45, 273, 46]. Omnicom closed its purchase of Interpublic on 26 November 2025 [274]. GroupM became WPP Media on 28 May 2025 [275, 276]. Life360 closed its purchase of Nativo on 2 January 2026 [277, 59]. Reddit's Max campaigns entered beta in January 2026. They opened to all advertisers in September 2026 [278]. Taboola's Realize+ entered expanded beta in the second quarter. The Trade Desk shipped Kokai Zuma on 27 August 2026 [279, 153, 280]. Several details are stale or unsupported. At the cutoff, Omnicom Media's roster is OMD, PHD, UM, Initiative and Hearts United. Hearts United combines Hearts & Science and Mediahub. It launched on 28 August 2026 with about $9.1 billion of 2025 billings. The playbook lists the two as separate brands. Omnicom's own site still shows Mediahub on its own [281, 282, 272]. The playbook names Annalect and Acxiom as Omnicom's data backbone. Omnicom Media's site names Omni, Acxiom Real ID and Flywheel. It does not mention Annalect [282]. Trade press says Omnicom retired the Annalect and Kinesso brands in mid-2026 [283]. The playbook calls Criteo the subject of acquisition speculation. Criteo's own releases show it moved its legal home from France to Luxembourg. The move was completed on 29 July 2026. The releases also show an approved move to a US legal home, expected in January 2027. They make no reference to a sale [284, 272]. In July 2026, Bloomberg reported a takeover approach from Vista Equity Partners and Quinti Capital. Criteo declined to comment. So the playbook's "acquisition speculation" matches the public record [285]. CitrusAd now trades as Epsilon Retail Media [286, 287, 272]. And the line "MFA's direct share has fallen sharply (ANA)" rests on two measures. They cannot be compared. Both are discussed below. ### What a buyer can inspect The ANA's programmatic transparency work is the main source on what a large buyer can see. It was built on log-level data, not on vendor reports. Log-level data is the record of each impression bought. Its 2023 study covered 21 marketers. It found made-for-advertising (MFA) sites took 21% of impressions and 15% of spend. Their CPMs, or prices per thousand impressions, were 25% below other sites. And they passed standard screens for viewability, invalid traffic and brand safety [89]. Some marketers then opted into the ANA's quarterly benchmarks. The ANA's December 2024 release said their average MFA spend fell to 6.2% in 2024. The ANA's own first-quarter 2025 report restates that value as 1.1%. It then sat at 0.4% to 0.6% of spend through 2025. It rose to 1.1% in the first quarter of 2026. That was roughly flat across its performance groups. That quarter the ANA named "AI slop" as an emerging sub-type [288, 289]. The ANA says its waterfall, average and median values cannot be compared [94, 95, 96, 97, 98]. Private marketplaces are invite-only auctions. In the third quarter of 2025, 81.6% of participants' combined spend went through them. For the median buyer it was 88.8%. Their median domain count fell to 26,372 [97, 290]. Three things stop that series from meaning "solved". First, the participants, from 21 marketers in the 2023 study to 86 in the first quarter of 2026, chose to join the log-level benchmarks. So it is a result for that group. The top quartile still had up to 27.4% of web spend on MFA sites. That was in the third quarter of 2025 [97]. Second, the market-wide series from Jounce Media stops in mid-2023, at about 30% of web auctions. No market-wide figure for 2025 or 2026 was accessible [89, 291]. Third, none of it measures native supply. Recommendation platforms are walled-off buying routes. They do not pass through the DSP logs the ANA studied. For native, transparency has to come from the platforms. Their own docs show what they offer. Both give reports by placement after the ads run. Outbrain and Teads offer publisher and section reports "once your campaign has gathered some data". By default, a campaign can exclude up to 30 publishers and 100 sections. Realize has a sites tab with per-site results and a setting to block sites. It has a tool called SpendGuard that caps spend by itself. It also has a pre-bid brand-safety setting whose supplier is not documented [292, 293, 294, 295, 296]. Neither native self-serve platform documents a list of sites before the campaign. Outbrain DSP, though, accepts publisher include lists, which name the sites to buy [264]. A smaller regional platform shows it can be done. Readpeak's docs show a full named list of its publishers on offer before a campaign. They also show per-site bids and spend limits [297, 298]. The ANA's own advice is the reverse of what these controls support. It favours inclusion lists over exclusion lists. Exclusion lists, it says, are "largely ineffective" against domains that appear daily [89]. Taboola's answer to MFA exposure is a curated tier called Taboola Select. The tier is described as premium editorial inventory "free from MFA properties, as verified and ensured by Jounce Media". Taboola publishes no MFA claim for the rest of its network. Its own policy page defines MFA sites as ones that "use paid ads on major platforms to lure real humans". The ANA and Adalytics show its widgets playing that role [299, 300]. Both vendors publish landing-page rules. Teads sets numeric caps. It allows no more than five ad placements and one video player per screen. It bars auto-refresh above three times a minute [301]. Realize requires an "Advertorial" label, sponsor disclosure and contact details. It bans false-urgency timers [302, 303]. Neither publishes figures on rejections, takedowns or audits. So having a rule does not show it is enforced. | What a buyer wants | Taboola / Realize | Outbrain / Teads | Sources | |---|---|---|---| | Site-level performance report | yes, after delivery | yes, by publisher and section, after delivery | [295, 296, 292] | | Pre-campaign inventory list | not found in public documentation (Sept 2026) | native platform: not found (Sept 2026); Outbrain DSP: publisher include lists at setup | [296, 292, 264] | | Exclusion controls | block sites; SpendGuard automatic caps | 30 publishers and 100 sections by default | [296, 295, 293, 294] | | Inclusion list | not found in public documentation | not found in public documentation | [296, 292, 89] | | MFA-free tier | Taboola Select, Jounce-verified, curated tier only | not found | [299, 300, 301, 304] | | Landing-page rules | labels, disclosure, no false urgency; no density cap found | numeric ad-density and refresh caps | [302, 303, 301] | | Enforcement statistics | none published | none published | [302, 301] | | Attribution windows | click 1–30 days (default 30); view 1–24 hours (default 24) | legacy pixel 24-hour matching; 72-hour window for imported conversions | [181, 220, 221] | | Realised lag distribution | not found in public documentation | not found in public documentation | [181, 220] | | Exportable learnings or logs | public docs: aggregate reports and audience upload; contract terms unknown | public docs: aggregate reports; API by request; contract terms unknown | [231, 233, 234, 235] | ### What a buyer can test Chapter 7 showed that no randomised incrementality study of open-web native is public. Incrementality means the extra results the ads caused. So the test has to be the buyer's. The decision tool below starts from the question a buyer is really asking. It matches that question to the test that answers it. It sets out what each test must assume, where it fails and what it still cannot say. | Buyer question | Test that fits | What the inference needs | Where it fails | Cannot tell you | |---|---|---|---|---| | Did this native campaign cause sales we would not otherwise have had? | A randomised holdout at user or geography level, with the platform excluding the control group from delivery, or a geo experiment with synthetic controls where user-level holdouts are not offered. | The control group is drawn from the same population the platform would have targeted; the platform does not re-target the control through another channel; the outcome is measured the same way in both groups; spend and duration are large enough to see a small effect against noisy sales. | Small budgets or short flights: across 25 large experiments the median confidence interval on return was over 100 percentage points wide. Any exposed-versus-unexposed or clicker-versus-non-clicker comparison, which is what native vendor brand-lift studies use, overstates lift. | Long-run brand effects beyond the post-treatment window; effects in channels the test did not measure; whether the same result holds at a different spend level. [162, 164, 165, 171, 734] | | Is the platform's reported cost per acquisition true? | Reconcile platform-reported conversions to your own analytics or CRM by click identifier, with the attribution window, view-through rules and realised lag distribution agreed before the campaign, as the IAB/MRC retail-media guidelines require. | The documented attribution windows are applied the same way on both sides (Taboola documents click windows of 1 to 30 days, default 30, and view windows of 1 to 24 hours, default 24); the realised lag distribution, which public documentation does not show, is supplied; your conversion definition matches the platform's optimisation event; deduplication against other channels is possible. | When the optimisation event is a proxy (page views when tracking is absent), when windows change mid-flight, or when view-through credit is included without a disclosed rule. | Say whether the conversions were incremental; reconciliation solves counting, not causation. [184, 145, 178, 228, 181] | | Which creative or landing page performs better? | An A/B split inside the platform on the same placements and period, judged on the downstream outcome rather than the click. | Random assignment across the split, which matters more than exactly equal budgets; enough conversions to separate the arms; if the platform shifts spend adaptively between arms, an analysis that accounts for it. | When the platform's optimiser reallocates spend between arms and the analysis ignores it, or when the outcome is click-through, which in native units includes accidental clicks. | Tell you whether either arm beat no advertising at all. [135, 159, 136] | | Is native more efficient than my other channels? | A marketing mix model calibrated with channel-level experiments on the same outcome, region and window, replicated over time before results are treated as anchors. | Enough spend variation over time; the experiment's counterfactual (reduced spend) matches the model's (zero spend) or the difference is adjusted; the calibration experiments are recent. | Small channels with little variation are poorly identified; there is no single formula to turn an experiment into a model prior; older or shorter experiments add uncertainty. | Attribute results to individual placements or creatives; separate a measurement change from a performance change when platform-attributed conversions are among the model's inputs; a model fitted to independently measured sales does not depend on the platform's window. [187, 188, 189, 191] | | Where did my ads actually run, and was the traffic real? | A pre-campaign site list and a post-campaign placement-level log from the platform, matched against your own inclusion list and a made-for-advertising classification you have chosen and understand. | The platform exposes site-level data at a granularity that reveals subdomains; the classifier you use is disclosed and consistent; invalid-traffic filtering is applied before billing. | Public documentation at both scaled platforms describes site reports after delivery and exclusion-first controls, though a pre-campaign list may be negotiable; classifications differ between vendors; subdomain rewriting has defeated verification tags before. | Prove the reader read anything; viewability and validity are delivery-quality baselines, not outcome measures. [89, 292, 296, 92, 97] | | Are we reaching new customers or paying for people who would have bought anyway? | A holdout with new-to-brand defined before the campaign, using a disclosed lookback period, or a retargeting-only versus prospecting-only split with separate holdouts. | A stable customer identifier; a new-to-brand window the retailer or platform discloses; separation of prospecting and retargeting budgets. | Frequent, already-engaged users are the easiest to reach and the least responsive; dynamic product retargeting underperformed generic creative on average in the one field experiment that tested it. | Distinguish brand-search harvesting from genuine demand creation without a separate search experiment. [175, 173, 174, 184] | | Can I leave this platform without losing what it learned? | An export test before you scale: request placement-level logs, audience definitions and conversion history, and check what the reporting API returns. | Your contract grants data access; the platform's API returns more than aggregates. | Public documentation at both scaled native platforms describes aggregate reports; exports of audiences, click-level logs and model learnings were not found and may depend on the contract; conversion history feeds the platform's model. | Recover the optimisation state; a new platform starts its learning phase from zero, which the published rules put at five to ten conversions a day. [231, 233, 234, 235, 145, 139] | | Did the campaign move brand metrics? | A survey-based brand-lift study with a randomised control group that is served a public-service or unrelated ad, not an exposed-versus-unexposed comparison. | Random assignment before exposure; the control ad is delivered through the same targeting logic; response rates similar across arms. A public-service control ad can itself change behaviour, so it is a comparison ad, not a no-ad counterfactual. | Exposed-versus-unexposed designs suffer activity bias, overstating effects by up to 350% in one controlled test; clicker-versus-non-clicker designs select on the outcome. | Convert an attitude change into a sales figure; the meta-analytic effect of disclosure on behavioural intention is null. [165, 166, 167, 108] | The contract matters as much as the test. Under the ANA's contract guidance, an agency must disclose when it acts as principal rather than agent. That is when it buys media itself and resells it. Affiliate definitions must reach the top of the holding company. Mark-ups on principal inventory must be capped. The advertiser must get access to transaction data. And auditor agreements must be attached as contract exhibits [305, 306]. For a native buy, the matching asks are narrower and more specific. The first is the attribution window and the realised lag distribution behind reported conversions. A conversion is the action the platform counts as a result. The window is how long after an ad a conversion can still be credited to it. The lag distribution shows how long the reported conversions actually took. Both should be disclosed before the campaign, as the MRC's outcomes standard requires of measurement providers and the retail-media guidelines require of retail media [183, 184]. If the platform prices per acquisition, ask how an acquisition, or an incremental action, is measured. The second ask is the site list, before the ads run. The third is the definition of the conversion the platform optimised to. The docs show this can be a page view when tracking is absent [145]. The fourth is an export of the placement log. The windows and conversion definitions are documented, at least by Taboola [181]. Three items were not found in public documentation. One is the lag distribution. Another is a pre-delivery site list at the two large platforms. The last is a placement-log export. Buyers may be able to get them by contract. ### Guarantees, packaging and the outcome promise The playbook argues native must be sold as outcome infrastructure to reach enterprise budgets. The contracts looked at here run the other way. In both scaled platforms' filings, the only guarantee language is about publishers. The platform promises them a minimum per thousand impressions [46, 1, 266, 158, 267]. The clearest guarantee to buyers found in the wider market is Vevo's "Attention Guaranteed" with Adelaide. Announced in January 2026, it promises a minimum attention score, not a sale. It discloses no make-good, meaning what the seller gives back if the promise is missed [307, 308]. In May 2025, Teads' own blog described a pilot. It was for "Attention-to-Outcome Guarantees" built on Adelaide scores. No terms, markets or current status were published [309]. An attention guarantee is not an outcome guarantee. The playbook's phrase "attention-to-outcome guarantees" overstates what has been promised by contract in public. The signal from buyers is mixed. A Taboola-sponsored survey covered 200 senior marketers at large US and UK firms. It found 80% would spend more on the open web if it had the automation of walled gardens. Those are the big closed platforms [153, 310]. Stated intent is cheap. In the same quarter, Teads suspended guidance. The cause was volatile direct-response and small-business demand. The evidence supports two points. Enterprise budgets for native follow proof that the buyer's own measurement team will accept. And the two large native platforms do not document the tools such a team would need. Those tools are a holdout, a disclosed lag distribution and a pre-campaign inventory list. A holdout is a group kept from seeing the ads. ### The publisher's decision For the publisher, adoption is a decision about where the widget sits on the page. The recommendation widget is the block of paid and editorial links, often below a story. The cases below show financial incentives that can conflict with quality and with control over placement. The record shows publishers turning up clickbait in the units when they could [79]. It shows them swapping $1–2 million guarantees for revenue shares. They did it to gain freedom over placement [79, 311]. In one documented case, a publisher ran a paid-traffic subdomain for seven years. One observed session on it carried more than 200 ads [92, 93]. Against that, the widget is one of the few income lines that need no sales team. Teads' EngageOS proposes one auction for ads and for links to the site's own stories. The aim is to optimise a whole session's yield. No adoption or lift figures have been disclosed yet [265, 271]. Jounce Media has a "cheap reach" category for ad space under 15% viewability. It includes widgets below the fold, where the reader must scroll to see them. Jounce names Outbrain and Taboola among the operators. Buyers called it a necessary evil [312]. A publisher choosing a native partner in 2026 faces a choice. One path is guaranteed pay, with exclusivity. The other is flexible pay, without it. Chapter 4 showed which way the money has been moving. ### The rules, in three jurisdictions The legal floor under all of this asks for much the same things. Where it differs is enforcement. In the US, Section 5 of the FTC Act sets the test. An ad format is deceptive if it materially misleads reasonable consumers about the ad's commercial nature or source. Deception is judged on the ad's overall net impression. The burden falls on "everyone who participates directly or indirectly in creating or presenting native ads". That includes publishers, agencies and affiliate networks [125, 313, 2]. Lord & Taylor, settled in 2016, is the best-known FTC native case. It involved a paid magazine piece and about 50 undisclosed influencer posts. But it is not the only one. In February 2019 the FTC approved final orders against Creaxion and Inside Publications. Inside Publications had run paid ads disguised as articles in its magazine [314]. Enforcement since 2023 has gone to endorsements and reviews. The Consumer Reviews and Testimonials Rule was pleaded in a 2026 order. The FTC's Legal Library lists roughly ten review-related matters dated 2023 to 2026, by last-updated date; one, the Rytr order, was set aside in December 2025 [315, 316, 317, 318]. This paper did not run a defined case search. So it makes no claim about enforcement trends. Section 5 of the FTC Act is the statute. The 2015 policy statement and guide, and the Endorsement Guides, are the agency's reading of it. The reviews rule is a trade regulation rule. State privacy law governs behavioural targeting of native units. California's CCPA as amended covers "sharing" personal information for cross-context behavioural advertising. That means ads based on what a person does on other firms' sites and apps. A business that sells or shares personal information must honour valid opt-out preference signals. It generally must also post the "Do Not Sell or Share" link. It may leave the link off only if it processes the signals without friction and makes the related disclosures, as the regulations define. The regulations say plainly that the link is not an alternative to honouring the signal [319]. Since the first regulations took effect in 2020, a compliant browser signal has counted as an opt-out of sale. Since 29 March 2023 that also covers sharing, for the browser and any linked profile. An amended package took effect on 1 January 2026 [320, 321, 319, 322, 323]. From 1 January 2027, California's Opt Me Out Act applies. It makes it unlawful to maintain a browser without an opt-out signal that consumers can set [324]. As of 24 November 2025, nineteen states had signed comprehensive privacy laws, by the IAPP tracker's convention. Of these, seventeen had an explicit right to opt out of targeted ads. Counts differ by date, and by whether narrower laws such as Florida's are included [325, 326]. By the cutoff, at least two more states had signed. Oklahoma signed in March 2026, and Alabama in April 2026. Both laws take effect in 2027 [327, 328]. The EU bans undisclosed advertorials outright. An advertorial is an ad made to look like a story. Undisclosed advertorials are item 11 on Annex I of the Unfair Commercial Practices Directive. Everything on that list is unfair in all circumstances. Item 11a adds undisclosed paid search results or ranking [329]. The Digital Services Act has set rules for online platforms since 17 February 2024. Micro and small firms are exempt. Platforms it covers must label ads clearly with who placed them and why the user sees them. They may not show ads based on profiling that uses special categories of personal data. Nor may they do so when they are aware, with reasonable certainty, that the user is a minor. Very large platforms must also keep a public archive of their ads [330, 331, 332]. The AI Act's Article 50 transparency duties applied from 2 August 2026. They cover telling people when they deal with AI. They cover marking generated content so machines can read it. They also cover deepfakes. And they cover AI-generated text published to inform the public on matters of public interest. Such text is exempt if it has had human review or editorial control. Someone must also hold editorial responsibility for it. So they do not require the same label on all AI-assisted ad copy [333]. The AI Omnibus delayed the separate high-risk duties to December 2027 and August 2028. It also gave generative systems already on the market more time. They have until 2 December 2026 to meet the Article 50(2) duty. That is the duty to mark content so machines can read it. The other Article 50 duties applied from 2 August 2026 [334, 335]. No rule found expressly governs paid placements inside AI assistant answers. There, the closest authorities are the paid-search precedents and the platform labelling duties [336, 337, 338, 339, 340]. The UK's CAP Code requires marketing to be obviously identifiable. Advertorials must be headed as such. Schedule 20 of the Digital Markets, Competition and Consumers Act has applied since 6 April 2025. It makes undisclosed paid editorial promotion and fake reviews unfair practices. It lets the competition authority enforce this directly [341, 342]. The ASA is the UK ad regulator. In February 2026, it upheld a complaint against a medicated weight-loss advertorial on Mumsnet. The ad had only a small "Created by" credit. In April 2026 the ASA upheld another, against a paid search ad. The ad presented a trader's site as an independent review body [343, 344, 345]. Industry guidance sits below all of this. IAB's Native Advertising Playbook 2.0 of May 2019 requires a disclosure on every native type. It defers to the FTC's 2015 documents. It has no legal force and sets no exact wording. So the "Promoted" labels that the FTC calls ambiguous stay in use [3]. The tool below lists each source with its effective date and scope. Nothing in this section is legal advice. **Disclosure and governance instruments** | Jurisdiction | Instrument | What it requires for native / disclosure | Effective | Scope | Law or practice | Src | |---|---|---|---|---|---|---| | US | FTC Enforcement Policy Statement on Deceptively Formatted Advertisements | Ad format deceptive under Section 5 if it materially misleads reasonable consumers about commercial nature or source (net impression); applies to advertorials, native/sponsored content, paid search results, infomercials | 2015-12-22 | Legal standard (Commission statement interpreting Section 5 FTC Act); all commercial speech; advertisers and other participants | agency guidance interpreting Section 5 (FTC Act) | [125] | | US | FTC Native Advertising: A Guide for Businesses | Disclosure in front of/above headline, prominent, plain terms ('Ad', 'Advertisement', 'Paid Advertisement', 'Sponsored Advertising Content'); 'Promoted'/'Promoted Stories' called ambiguous; applies to advertisers, publishers, agencies, affiliate networks | 2015-12 | Staff guidance (no safe harbour) | agency guidance interpreting Section 5 (FTC Act) | [2] | | US | FTC Endorsement Guides, 16 CFR Part 255 (2023 revision) | Clear and conspicuous disclosure of material connections; responsibility on brand and endorser, not platform; '#ad'/'Paid ad' at start of post | 2023-07-26 | Interpretive guides (basis for Section 5 actions); endorsers, advertisers, intermediaries | legal obligation | [735] | | US | Trade Regulation Rule on the Use of Consumer Reviews and Testimonials, 16 CFR Part 465 | Bans fake/AI-generated reviews and testimonials, paid-sentiment reviews, undisclosed insider reviews, company-controlled 'independent' review sites, review suppression, fake social-media indicators; civil penalties | 2024-10-21 | Binding rule; reviews and testimonials only (not native ad formatting) | legal obligation | [736] | | US-CA | CCPA/CPRA, Civ. Code 1798.135 and 1798.140 | 'Do Not Sell or Share' link or honouring opt-out preference signal for sale/sharing; 'sharing' = disclosure for cross-context behavioural advertising | Sharing provisions since 2023-01-01; current text amended effective 2025-01-01 | Statute; businesses meeting CCPA thresholds | legal obligation | [321] | | US-CA | CCPA Regulations s.7025 (opt-out preference signals), 2025 update package | Compliant signal must be processed as valid opt-out of sale/sharing for the browser/device and associated (incl. pseudonymous) profiles | Package effective 2026-01-01 (s.7025 in force since 2023) | Binding regulation | legal obligation | [319] | | US-CA | AB 566 California Opt Me Out Act (Ch. 465, Stats. 2025) | Browsers must include consumer-configurable functionality to send an opt-out preference signal | 2027-01-01 | Statute; businesses that develop or maintain a browser | legal obligation | [324] | | US (19 states at 24 Nov 2025; at least 21 by the cutoff) | Comprehensive state privacy laws (IAPP chart) | Opt-out of sale; 17 of the 19 in the IAPP chart include opt-out of profiling/targeted advertising; Oklahoma (signed Mar 2026) and Alabama (Apr 2026) take effect in 2027 | 2020-01-01 (CA) to 2026-01-01 (IN, KY, RI) | Statutes; thresholds vary by state | legal obligation | [325, 327, 328] | | EU | Directive 2005/29/EC (UCPD) Annex I items 11 and 11a; Art. 7(2) | Undisclosed paid editorial promotion and undisclosed paid search results/ranking are unfair in all circumstances; failure to identify commercial intent is a misleading omission | Item 11 via national transposition (2007); item 11a from 2022-05-28 | Directive transposed in member states; traders | legal obligation | [329] | | EU | Digital Services Act, Regulation (EU) 2022/2065, Art. 26 and 28 (as summarised by the Commission) | Ads clearly labelled with advertiser identity and main targeting parameters; no ads based on profiling with sensitive data; no profiling-based ads to minors; VLOP ad repository | VLOPs/VLOSEs late Aug 2023; all platforms 2024-02-17 | Regulation; online platforms (not pure publisher sites) | legal obligation | [330] | | EU | AI Act, Regulation (EU) 2024/1689, Art. 50 (transparency), as amended by AI Omnibus | Disclose AI interaction; machine-readable marking of AI-generated content; label deepfakes and AI text informing the public | 2026-08-02; Art. 50(2) marking for generative systems already on the market by 2026-12-02 | Regulation; providers and deployers of covered AI systems; not advertising-specific | legal obligation | [337, 334, 335] | | EU | AI Omnibus Regulation (OJ L_202601744) | Delays high-risk rules to 2027-12-02 (Annex III) and 2028-08-02 (Annex I); nudification ban from 2026-12-02; gives generative systems placed on the market before 2026-08-02 until 2026-12-02 to meet the Art. 50(2) marking duty | 2026-07-27 entry into force | Regulation amending AI Act | legal obligation | [340, 334, 335] | | EU | Digital Fairness Act (proposal) | Expected to address dark patterns, addictive design, influencer marketing, personalisation | Proposal expected Q4 2026 (not adopted at cutoff) | Not yet law | proposal, not yet law | [737] | | UK | CAP Code rules 2.1-2.4 | Marketing communications obviously identifiable; advertorials labelled e.g. 'advertisement feature'; no falsely posing as consumer | Current edition | Self-regulatory (ASA); industry practice with statutory backstop | self-regulatory code (ASA), backed by DMCC Schedule 20 | [341] | | UK | DMCC Act 2024 Schedule 20 paras 12-13 | Undisclosed paid editorial promotion and fake/concealed-incentive reviews unfair in all circumstances | 2025-04-06 | Statute; CMA direct enforcement | legal obligation | [342] | | US/global | IAB Native Advertising Playbook 2.0 | All native ad types must carry a disclosure conveying the content is paid, large and visible enough to notice; defers to FTC 2015 documents | 2019-05 | Industry practice, not law | industry practice (no legal force) | [3] | > **Author's illustrative sizing, not a benchmark.** > Is a holdout worth running? That depends on the baseline conversion rate, the smallest effect worth detecting and how many people can be randomised. Spend alone does not settle it. Lewis and Rao's 25 large field experiments show how hard it is. They spent about $112,000 each on average and still produced median confidence intervals on return more than 100 points wide [162]. > > The table shows the sample a simple test needs. It assumes randomised eligible users, one yes-or-no conversion over a fixed window, equal test and control groups, a two-sided 5% significance level and 80% power. These are assumptions for illustration, not native benchmarks. > > | Control conversion rate | Rate to detect | Relative lift | Users per group | Total users | > |---|---|---|---|---| > | 0.5% | 0.55% | 10% | 327,922 | 655,844 | > | 1.0% | 1.10% | 10% | 163,095 | 326,190 | > | 2.0% | 2.20% | 10% | 80,682 | 161,364 | > > The table uses the standard normal approximation for two proportions: users per group = [1.96√(2m(1−m)) + 0.8416√(p₀(1−p₀) + p₁(1−p₁))]² ÷ (p₁ − p₀)², where p₀ and p₁ are the two rates and m is their mean. > > The users are everyone randomised, including people who never click. So the sample cannot be multiplied by a cost per click. Turning it into a budget needs the test group's reach, impressions per person reached, click-through rate and cost per click. The test's length depends on how many eligible users arrive each week and how long conversions take to arrive. A geo test needs its own calculation over regions and weeks. A longer test or pooled flights can help, but clusters, spillover between groups and changing conditions add their own error. So run the power calculation before recommending a holdout. If the design cannot resolve an effect that matters to the decision, say so, and cap the spend that stays unmeasured. > **For the publisher, in a checklist.** > Ask for the revenue share or the guaranteed rate per thousand page views, and its term. Ask how exclusivity is defined and what it covers. Ask which advertisers and categories you can block yourself, and whether blocking goes through an account manager. Check whether your pages could fall inside a buyer's made-for-advertising filter. For scale: Taboola records about 63.5% of revenue as traffic acquisition cost and Teads about 59.3%. Guarantees ran at about 15% of Taboola's supply cost in 2025. > **For the buyer, in three sentences.** > Treat the native platform's reported conversions as a claim, not a measurement, until the window, the lag and the definition are disclosed. Run a power calculation first: if a test can detect an effect that matters to the decision, run your own holdout or geo test and expect it to be noisy; if it cannot, say so and cap native spend as unmeasured. Ask for the site list before the campaign and the placement log after it, and price the vendor's answer into the deal. ## 11. What changes next *The two large native players on the open web are moving in different directions. One is spreading beyond the recommendation unit, or widget. That is the box of suggested links on a publisher's page. The other is being squeezed by the math of a format sold against page views. Agents, protocols and answer engines change who buys and where the ad sits. They do not change whether the click was worth anything.* > **In short.** > In the same quarter of 2026 Taboola grew ex-TAC gross profit 11.8%. Under its old definition the gain was about 4.7%. Ex-TAC means net of traffic acquisition cost (TAC), the money paid to publishers. Taboola raised its profit guidance but cut its revenue range. Teads Holding Co. shrank revenue 17% in that quarter. It suspended guidance, citing swings in direct response and "open-web headwinds". Teads' annual report speaks in the past tense. It says AI answers in search have cut click-throughs to publishers and affected its revenue. Taboola's filings say nothing of the kind. Two agent protocol stacks exist: AdCP and AAMP. Neither set of pages refers to the other. None of the large native platforms searched shows up as having built on either. Most automated buying systems whose docs were opened keep a human approval step for launch. Realize+ reallocates budget automatically within rules the buyer sets. The scenarios that matter turn on three signs a reader can watch. They are publisher sessions, guarantee costs and whether anyone publishes a holdout. Guarantees are the minimum payments promised to publishers. A holdout keeps the ads from a control group. It tests what the ads really add. This chapter sorts three things. It splits what was announced from what was done. It splits a retired brand from a retired product. And it asks what agentic buying can do for native beyond the optimisation chapter 6 described. Agentic buying means software agents do the buying. ### Two trajectories, one quarter The second quarter of 2026 gives a useful contrast in reported results. It is not a natural experiment, though. Nothing assigned the two firms their different conditions. Teads Holding Co. reported revenue down 17% to $284.6 million. Ex-TAC fell 14%. Adjusted EBITDA fell 74% to $7.0 million, and it posted a $42.5 million net loss. Then on 6 August it suspended all guidance. That included the roughly $100 million adjusted EBITDA target it had reaffirmed a quarter earlier. It cited "the volatility of the Direct Response and SME business" and "open-web headwinds" [271, 346, 29]. Its restructuring in late 2025 cut staff by about 10%. It replaced its chief commercial officer, chief marketing officer and North America managing director [29, 346]. The company casts its enterprise and connected TV business as growing. It names direct response and small business as the problem. That points at the performance and widget lineage, the part that was Outbrain. Its segment reports do not show that lineage's results on their own. Taboola reported the opposite. Ex-TAC gross profit rose 11.8%. Adjusted EBITDA rose 22.8% and beat guidance. It raised full-year guidance for both [30, 347]. In the same release its revenue came in below its own range for the second quarter. Its full-year revenue range was cut by $76–106 million at the endpoints. Management did not highlight the cut [348, 347, 30]. There is a precedent. In August 2024 Taboola cut its revenue guidance by about $170 million but held ex-TAC. It explained that some Yahoo revenue "will now only be reflected in ex-TAC" [349, 350, 41, 351]. Management did not tie the 2026 cut to presentation. On the earnings call it gave two causes. One was leaving publisher ties, largely in Greater China. The other was Google phasing out its Explore More feature [352]. Take the definition used before the second quarter. On that basis the raised ex-TAC range would be about $760 million to $771 million. That is flat to slightly down against May. Three things could each push revenue and ex-TAC apart. One is more revenue shown net. Another, described in chapter 4, is the new non-cash add-back. The last is a shift in demand or mix. Management disclosed no bridge between them, so the paper does not choose. ![Revenue guidance fell while profit guidance rose, but on the old definition profit guidance was flat.](figures/fig-guidance-divergence.svg) *Figure fig-guidance-divergence. **Revenue guidance fell while profit guidance rose, but on the old definition profit guidance was flat.** Taboola’s full-year 2026 guidance ranges at each 2026 update. Between May and August the revenue range was cut by $76–106 million at the endpoints while the ex-TAC gross profit and adjusted EBITDA ranges were raised. The August ex-TAC range includes a newly added-back non-cash write-off; on the earlier definition it would be about $760–771 million, flat to slightly down against May. On the call, management attributed the revenue cut to leaving publisher relationships, largely in Greater China, and to Google’s deprecation of Explore More. Each panel is drawn on its own axis, which starts above zero; the chart shows ranges and midpoints, so position rather than bar length carries the value. Units: USD millions, ranges and midpoints. Scope: Taboola FY2026 guidance as filed on 25 February, 6 May and 5 August 2026. Sources [348], [347], [30], [352]. Caveat: Midpoints are the paper’s recomputation from the guided ranges. Non-zero axes make small movements visible; the dollar labels give the size. Data: data/figure-data.json#guidance* The two firms also differ in how they disclose the shock from AI search. Teads' 2025 annual report says generative AI answers in search "reduced [users'] need to click through to original publisher websites". It says the resulting fall in partner traffic "has affected, and could in the future have a significant effect on, our revenue" [46]. Taboola's filings never use the phrase "AI Overviews". Its 2026 quarterly reports do not contain "generative AI". Its 2026 releases say nothing about search traffic [353, 30, 347, 348]. One reading is that the exposure differs, since Yahoo and Microsoft are portals with direct traffic. The other is that their disclosure habits differ. The record cannot tell them apart. ### The traffic shock, measured where it can be The pressure on the widget sits at the top of the publisher's funnel. People Inc.'s core sessions fell 5% over the full year 2025. In the fourth quarter of 2025, they fell 13%. In the first quarter of 2026 they fell 17%. In the second quarter they fell 22%. The company tied the fall "primarily to the impact of the growing prominence of Google AI Overviews" [354, 355, 356, 215]. Its digital revenue still grew 6% in the second quarter of 2026, driven by performance marketing and a Meta content licence. Advertising held flat, as higher programmatic rates offset lower volumes [215]. Ozone's benchmark data, as reported by Digiday, covers the second quarter. It puts publisher ad requests down 32–37% year on year in the United States. In the United Kingdom they fell 39–41%. In the UK, eCPMs rose 30% in June and made up part of the loss. An eCPM is the price paid per thousand ads shown. US programmatic spend fell 44% over the first half [357]. Those are the alliance's own benchmarks of premium inventory, not a neutral measure. They suggest that a large premium publisher can reprice. The paper infers that a long-tail widget host, without that pricing power, cannot do so as easily; no data here measures it. Dianomi blamed its 2025 fall in impressions, 14% by its own total, on "AI-powered content discovery tools". It then grew 2% in the first half of 2026, from a lowered base [358, 359]. These series do not measure the same thing. People Inc. reports sessions, Ozone reports ad requests and Dianomi reports impressions. A session can hold more than one page. A page can send more than one request. A request may go unfilled. The paper infers that fewer sessions and requests mean fewer widgets shown. That is because the widget is sold against page views. No series measures widget inventory directly. ### Retirements: brands, products and the difference The record of failures has to tell a retired name from a retired product. Most of the "exits" in this market are the former. | Entity | What happened | Announced | Completed | What continued | Sources | |---|---|---|---|---|---| | Sharethrough | acquired by Equativ; operates under the Equativ brand | 2024-06-12 | 2025-06-09 | native listed among Equativ's channels; a Sharethrough-branded site still live | [360, 54] | | Microsoft Start | brand reverted to MSN | 2024-11 | 2024-11 | Taboola's Microsoft supply relationship, at 5% or more of revenue through 2025 | [361, 1] | | Ligatus | acquired by Outbrain, effective 2019-04-01 | 2019 | 2019 | entity still listed as a Teads Holding Co. subsidiary in 2025 | [362, 363] | | Nativo | acquired by Life360 | 2025-11-10 | 2026-01-02 | product line combined with Life360's location and family data | [364, 365, 59] | | Outbrain | acquired Teads, then took its name | 2024-08-01 | 2025-02-03 | Outbrain lineage products under the Teads brand | [46, 29] | | TripleLift | third layoff round in five years, 2025-07-11 | 2025-07 | 2025-07 | independent, Vista-owned | [366] | | Revcontent, ZergNet | sites live at the cutoff; no ownership notices | | | operating | [367, 368] | | Adyoulike | acquired by OpenWeb | 2022-04 | 2022 | site live; current product status not independently established | [66, 369] | | Engageya | discontinued service | | 2024-09-30 | its own homepage notice, captured by a third party | [370] | | plista | reported to be ceasing operations | 2024-03 | | domain has no address records at the cutoff | [371, 372] | | Content.ad | domain has no address records at the cutoff | | | status undated | [373] | This project could open just one documented case of a publisher dropping a widget. It is nine years old. In 2017, Outside magazine replaced Outbrain with in-house widgets. Outbrain was about 10% of its digital ad revenue. The in-house widgets earned higher click-through. The piece that reported this also noted that Outbrain "accounts for 30% of revenue for some publishers" [374]. The paper infers that dependence on revenue at that scale is one reason more publishers have not followed. No population-level series of widget removals was found. **AI and agentic capability map** | Platform / product | Capability | Status | Date | Independently observable? | What the record supports | Src | |---|---|---|---|---|---|---| | Taboola | Realize: Self-serve performance platform beyond search/social; native, display, vertical, video, carousel, app formats; CPA/ROAS objectives | live | 2025-02-26 | yes (public site) / no (performance) | Launch language from Q4 2024 8-K; 600M DAU claim; 11,000+ publishers | [351, 386] | | Taboola | Realize+: 'Agentic system that turns advertiser goals into outcomes' | beta | 2026-04-23 | no | Release says Realize+ makes and executes campaign decisions within goals the advertiser sets, and opened a Claude Skill for campaign setup and optimisation. No outcome data published. | [153, 386] | | Taboola | Abby: AI Ad Assistant: Conversational campaign setup, targeting, budget and creative edits; question-led | live | 2024-10-17 (announced) | partly (product page) | Vendor claim: 75% faster to live in early tests, no method | [392, 393] | | Taboola | GenAI Ad Maker: Generative creative | live | 2026-09 (listed in site footer) | no | Product page returned 404 | [393] | | Taboola | Realize ID: Cross-device audience profiling/data product | live | 2026-09 (site banner) | no | | [386, 393] | | Teads | Predictive AI engine / Teads Ad Manager: Outcome optimisation across screens; TAM self-serve | live | 2026-08-06 | partly | CTV HomeScreen placements integrated into TAM | [387, 271] | | Teads | Teads EngageOS: AI-powered publisher operating system unifying editorial and ad inventory | live | 2026-08-06 (Q2 2026 release) | no | Publisher-side, not advertiser automation | [271] | | Teads | Generative creative / agentic buying: — | unconfirmed | 2026-09-27 | no | No advertiser-side generative or agentic buying product found in the Q2 2026 release; the conversational AI SDK (publisher side) is listed separately. | [387, 271] | | MGID | MIA: Diagnostic AI assistant on live account data; recommendations only | live | 2026-07-16 | partly (blog) | 'the decision to act remains yours' | [394] | | MGID | CTR Guard / CPA Tune / smart bidding: Automated optimisation | live | 2026-09-27 (site) | partly | | [267] | | MGID | Generative creative tools: text-to-image, image-to-image, title and description generation in 29 languages, performance prediction | live | 2024-02-15 | no | Vendor announcement; no performance evidence. | [154] | | Google | Ads in AI Overviews: Search/Shopping ads inside AI Overviews; desktop expansion | live | 2025-05-21 | yes (public surface) | Markets not specified in extract | [257] | | Google | Ads in AI Mode / AI Search ad formats: Ads in AI Mode; Gemini-built formats tailored to query; Direct Offers pilot | live | 2025-05-21 to 2026-05-20 | yes (public surface) | No adoption figures | [257, 258] | | Google | Ask Advisor / Ads Advisor: Cross-product Gemini agent; agentic safety/policy features | announced | 2026-05-20 | no | Approval model not stated | [258] | | Google | AI Max for Search; Performance Max; Demand Gen: Goal-based automated campaigns | live | 2026-08-27 | yes (product) | Demand Gen '30% increase' claim, internal data H2 2025 | [257, 398] | | Google | Asset Studio Multimodal Video Creation: Generative video creative | live | 2026-08-27 | partly | | [398] | | Google | Business Agent for YouTube ads: Conversational agent inside ads | beta | 2026-09-16 | no | | [399] | | Meta | Advantage+ sales campaigns; Advantage+ creative (gen AI): Automated audience/placement/budget; AI creative variations | live | 2026-09-03 | yes (product) | Case-study evidence only | [400] | | Meta | Business Agent (WhatsApp): Conversational commerce agent | live | 2026-09-03 | no | | [400] | | Meta | Ads inside Meta AI assistant: — | unconfirmed | 2026-09-27 | no | No primary opened | | | Amazon Ads | Ads Agent: Natural-language campaign build/optimisation; SQL over AMC; human approval before launch | live | 2025-11-11 | partly (product page) | AMC users; US DSP Multimedia Solutions users | [395] | | Amazon Ads | Creative Agent: Generative display/video/audio creative at no cost | live | 2025-11-11 | partly | | [396] | | Amazon Ads | Campaign Manager: Unified AI campaign console | beta | 2025-11 (unBoxed 2025) | no | | [397] | | Amazon | Sponsored Products and Sponsored Brands prompts: sponsored prompts that can open a Rufus dialogue | live | 2026-03-25 | no | Establishes sponsored prompts that can open Rufus; does not establish that Rufus answers themselves carry ads. | [255] | | OpenAI | ChatGPT Ads: Ads to free-tier users; product feeds, geo/platform targeting, custom audiences, outcome-optimised bidding; self-serve Ads Manager | live | 2026-01-16 (announced); 2026-02-09 (US test began); 2026-05 (Ads Manager); 2026-09-01 (40+ countries, reported) | reported yes (Wired observation cited by Wikipedia; not opened) | $1B annualised run rate self-reported by OpenAI | [242, 243, 241] | | OpenAI + Stripe | Instant Checkout / ACP: Agentic checkout in ChatGPT; merchants pay fee; 'not sponsored' (Sept 2025) | beta | 2025-09-29 | yes (open spec) | Adjacent commerce rail, not advertising | [259, 260, 261] | | Perplexity | Sponsored answers / follow-up ads: Ads beneath answers (2024 test) | retired | phased out 2025; reported 2026-02-18 | yes (reported) | Subscription-first pivot | [248, 249] | | Microsoft | Copilot ads: ads inside Microsoft Copilot answers | live | 2025-08-06 | no | Vendor first-party figure of 73% higher click-through than search; formats and availability not listed on the page opened. | [254] | | The Trade Desk | Kokai: AI-driven DSP platform | live | 2026-02-25 | no (adoption % not in release) | | [401] | | Apostra (ex-Scope3) | Agentic media platform on AdCP: Buyer/seller agents; channels include native | live | 2026-09-23 | no (no customer names or volumes) | Rebrand from Scope3 | [391, 390] | | Prebid.org | Prebid Sales Agent: Open-source MCP/A2A sales agent implementing AdCP 3.1.1; GAM/Broadstreet/Kevel/Triton | beta | 2026-09-27 (repo state) | yes (open repo, 38 stars) | No native format documented | [382, 383] | | AdCP (Ad Context Protocol) (AgenticAdvertising.org (pending 501(c)(6), Delaware)) | Protocol version stable v3.1.24; rc v3.2.0-rc.6 (site shows 3.2.0-rc.3; docs 3.1.24); native format named: no | beta | 2026-09-23 (3.1.24); 2026-09-24 (rc.6) | 141+ members claimed (Yahoo, PubMatic, Scope3/Apostra, Samba TV, LG Ads, Kargo, Wunderkind, Sightly); implementations: Prebid Sales Agent (alpha, AdCP 3.1.1), Apostra platform; first agent-to-agent buy claimed 2025-10-16 (LG Ads) | | [375, 376, 377, 384, 382, 391] | | AAMP (Agentic Advertising Management Protocols) (IAB Tech Lab — Agentic Task Force) | Protocol version AAMP 3.0 (includes OpenProposal); prior 2.3, 2.0; native format named: no | live | index: Released 2026-02-26; page last updated 2026-09-22 | Components: Agentic Direct, Buyer Agent, Seller Agent, Open Proposal, Registry Agent Example, ARTF, Agentic Audiences; page mentions Google Ad Manager integration and a Databricks webinar; no adopter list | | [378, 379, 380] | | ARTF (Agentic Real Time Framework) (IAB Tech Lab — Agentic Working Group) | Protocol version v1.0; native format named: no | beta | 2026 (public comment at access) | none named | | [378, 380] | | Agentic Audiences (formerly UCP) (IAB Tech Lab — Agentic Task Force) | Protocol version not versioned; native format named: no | beta | 2026 | none named | | [378, 380] | | ACP (Agentic Commerce Protocol) (OpenAI and Stripe (founding maintainers)) | Protocol version spec release 2026-04-17 (cart, feed, orders, auth, MCP); native format named: n/a (commerce) | beta | 2026-04-17 | ChatGPT Instant Checkout (US; Etsy live at launch, Shopify merchants to follow; 'approved partners') | | [259, 260, 261] | | UCP (Universal Commerce Protocol, Google) (Google) | Protocol version not stated; native format named: n/a (commerce) | unconfirmed | 2026-05-20 (expansion); 2026-09-16 (cart transfer, checkout testing) | Tapestry, lululemon, Minted named as partners in Sept 2026 post | | [258, 399] | | OpenRTB Dynamic Native Ads API (IAB Tech Lab) | Protocol version v1.2; native format named: yes (the native spec) | live | 2017-07 | industry-wide (not enumerated here) | | [378] | | MCP / A2A (transports) (Anthropic / Google (referenced)) | Protocol version not stated; native format named: n/a | unconfirmed | — | AdCP planning-time tasks; Prebid Sales Agent; AAMP page | | [379, 377, 384, 382] | | Taboola | DeeperDive advertising engine: ads inside a generative AI answer engine on publisher sites, opened to other AI companies | live | 2026-06-16 | partly (vendor usage figures) | Vendor reports tens of millions of answers a month and 7M+ users. | [51] | | Teads | Conversational AI SDK: conversational ads for publishers; MCP integration listed as coming soon | beta | 2025-11-12 | yes (developer documentation) | Publisher-side tool; no advertiser-side agentic buying. | [252] | | Dianomi + Dappier | Conversational answer engine with brand-agent ad units on publisher sites | announced | 2026-03 | no | No deployment evidence found. | [253] | | OpenAI | Sponsored Agents: click-to-chat ads that open a brand agent in ChatGPT | beta | 2026-09-16 | announced by OpenAI; no outside observation of results | OpenAI's announcement and help centre describe the product and keep the sponsored chat separate from ChatGPT's answers and the user's conversation. No advertiser integration documentation. Trade press named Wayfair as a pilot advertiser. | [244, 245, 246] | ### Agents and protocols: shipped, demonstrated, announced The question on AI and agentic buying asked what truly autonomous buying can do beyond current optimisation. The record at the cutoff has three layers. The map above labels each item as live, beta, announced, retired or unconfirmed. It gives the date and says whether anyone outside the vendor has seen it. The protocol layer is real. It is also split in two, and the two stacks do not refer to each other. AdCP is governed by AgenticAdvertising.org, which calls itself a pending trade association. On 23 September 2026 AdCP showed a stable release line at v3.1.24. It showed a release candidate for 3.2 the next day. Its home page showed a third version string. Eight tags were cut in sixteen days [375, 376, 377]. IAB Tech Lab's effort is AAMP, short for Agentic Advertising Management Protocols. It was listed as released on 26 February 2026 at the umbrella level. But its real-time part was still in public comment. Its audiences part had no release listed [378, 379, 380]. On 22 September 2026 Tech Lab announced AAMP 3.0. Its new OpenProposal spec is in public comment until 22 October [381]. Neither set of pages mentions the other. Only one open source build of AdCP for sellers was opened, Prebid Sales Agent. It calls itself alpha [382, 383, 377, 384]. The group claims the first media buy between agents ran on 16 October 2025 with LG Ads. It discloses no buyer, amount or third-party observer [385, 375]. The native format's own programmatic contract is OpenRTB Dynamic Native Ads. It was finalised as version 1.2 in March 2017; the Tech Lab index lists July 2017. The agent specs, by contrast, change each week [378, 379, 376]. The large native platforms do not appear in any of it. Taboola, Teads and MGID are missing from AdCP's member logos. Nor are they on the Prebid Sales Agent's list of ad servers. Take Taboola's and Teads' results for the second quarter of 2026. Neither mentions AdCP, AAMP, MCP or buying between agents [375, 382, 30, 271, 386, 387]. AdCP can describe native. Its 3.1.24 docs refer to native formats, and its catalogue lists native entries. But a proposal for a standard native format is put off to the 3.2 track. It waits on an audit against OpenRTB Native 1.2 [388, 389]. On 23 September 2026, the commercial platform built by AdCP's co-creator was rebranded from Scope3 to Apostra. It lists native among its channels. It describes live campaigns in six countries but names no customer and no dollar figure [377, 384, 390, 391]. The vendor universe lists Kargo among sellers with native roots that are now pitched as high-impact. Kargo is among AdCP's self-reported members [384, 385]. But joining is not building on it. No live buying through it was found. Agents may reach native supply through a middleman before any large native vendor adopts a protocol. The vendor layer is mostly assistants and automation within limits. The tasks worth testing are separate. They are planning, tool use, budget changes, placement changes, making creative and watching campaigns. A marketing label does not say which of these are handed over. Taboola's Abby, announced in October 2024, guides campaign setup by asking questions. It also edits creative. Its one performance claim is that early campaigns went live 75% faster than manual setup. It gives no sample or method [392, 393]. Realize+ was announced on 23 April 2026 after a beta. It is marketed as "an agentic system that turns advertiser goals into outcomes". The pitch is that it keeps making and carrying out campaign decisions. The same release opened the platform to a Claude Skill for setting up and tuning campaigns [153]. Taboola's filings disclose no Realize, Abby or AI metric of any kind [351, 386, 393, 30]. MGID's MIA, launched in July 2026, reads live account data and explains problems. And "the decision to act remains yours" [394, 267]. MGID has offered AI image and text tools since 2024 [154]. Teads' release for the second quarter names predictive AI and EngageOS for publishers. And it names a bundle for connected TV [271, 387]. Its developer site carries an SDK for conversational ads in public beta. It lists MCP support as coming soon [252]. The platform layer is where autonomy is closest. It is still gated. Amazon's Ads Agent builds campaigns from a media plan and tunes them in plain language. But "campaigns only launch after you review and approve" [395, 396, 397]. Google showed advertiser agents at its May 2026 event. Its beta for a YouTube business agent came in September 2026. Next to these sits a Demand Gen claim. It cites a 30% average rise in conversions, with no stated baseline [258, 398, 399]. Meta's Advantage+ case studies report large lifts for selected advertisers. Two of them come from lift studies. None covers the whole platform [400]. The Trade Desk's results mention Kokai only in words, with no figures [401]. Most systems whose docs were opened keep some human approval step. Amazon has its review step. AdCP has its governance check and audit log. MGID has its "yours to decide", and Taboola its question-led setup [395, 377, 394, 392, 375]. Realize+ and Apostra describe the system acting on its own within rules the buyer sets. That is a weaker gate than a sign-off on each campaign. An approval step does not rule out the system acting alone inside set limits. What neither has published is what happened when it ran. On this record, what agentic buying changes for native is the buyer. Take an agent that can read a seller's supply, quality, pricing rules and how the seller defines outcomes. It will choose sellers that expose them in a form machines can read. AdCP defines tasks to report what ran and to send back results. Amazon opens its clean room to plain-language queries. Among the native vendors profiled in chapter 9, only Taboola documents an interface for agents in the material this review opened. It is an official Realize MCP server that can read and change campaigns. Taboola published it in July 2026 [402, 403]. None of those vendors documents a way for agents to set goals or report outcomes beyond its own API [377, 395, 386, 387, 267]. Wrapping an existing reporting API for agents is a small job for engineers. As positioning, it is a big one. No protocol changes the evidence problem from chapter 7. An agent that optimises to the conversions a platform reports is doing attribution faster. The incrementality question passes through untouched. That question is whether the ads caused conversions that would not have happened anyway. ### Scenarios, and what would tell you which one is running Three scenarios fit the record. Each names what it rests on and the case against it. Each names indicators to monitor. These are signals, not decisive tests. Where a direct test exists, the box names it and says whether the data are disclosed. > **Scenario one: the format is left behind.** > The large platforms keep growing profit. They do it by selling display, video and connected TV. They sell through the buyer ties and publisher contracts they built for widgets. The widget becomes one placement among many. Rests on: advertisers value the platform's conversion model and supply access. They value these more than the native format. Case against: the buyer base is mid-market and scattered. And once the format edge is gone, Realize competes with all DSPs. What to watch: Taboola's revenue guidance against its ex-TAC guidance. The share of Realize spend on placements that are not native. The company does not yet disclose it. DeeperDive's reported answer volume and advertiser uptake. Teads' share of revenue from connected TV, 13% in the second quarter of 2026. Time frame: the fiscal 2026 annual reports, early 2027. Signal against: Taboola's fiscal 2026 revenue growing faster than its ex-TAC gross profit with no new non-native line disclosed. Accounting presentation and mix can move that ratio, so it does not settle the question. Direct test: the native and non-native split of Realize revenue or spend, which Taboola does not disclose. > **Scenario two: the widget business shrinks with publisher traffic.** > AI answers keep cutting publisher sessions. The widget's impressions fall with them. Guarantees become too costly, and the second tier keeps consolidating. Rests on: publishers cannot replace search traffic fast enough. And portal supply (Yahoo, Microsoft) does not shield the largest platform forever. Case against: People Inc. grew digital revenue through a 22% fall in sessions. Scarcity has raised eCPMs in some markets. What to watch: People Inc. core sessions and Dianomi impressions. Taboola's cost of guarantees as a share of TAC. It fell from 18% to 13% between 2024 and the second quarter of 2026. Revenue and TAC lines tied to Yahoo in Taboola's quarterly filings. The pending Dianomi completion and any further deals in the second tier. Time frame: the four quarters to mid-2027. Signal against: People Inc.'s core sessions coming within 5% of the prior year for two quarters in a row. One publisher's recovery would not rule out a wider contraction. Direct test: widget impression volumes across a stable set of publishers, which no platform or measurement firm publishes. > **Scenario three: proof arrives.** > A native platform publishes an incrementality study. It works with an independent partner. The study is randomised or run by region. It discloses attribution windows and how long conversions lag the ad. It shows site lists before the ads run. Enterprise buyers respond with brand budgets, not direct-response tests. Rests on: the platforms have the volume to run such tests. And the results are good enough to publish. Case against: no platform has published one with design and results. A practitioner account shows at least one geo test ran; its lift estimate was not published. Vendor designs for brand lift still compare those who saw the ads with those who did not. And a small measured lift may be worse for sales than an unmeasured claim. What to watch: any holdout study by Taboola, Teads or MGID that names its measurement partner. Disclosure of how long conversions actually lag. Time frame: through the 2026 annual reports and help pages dated to mid-2027. Signal against: neither appears by then. Direct test: a published study followed by a measured shift of brand budgets toward the platform, which would need spend data by budget type that neither platform discloses. | Indicator | Latest value | Period | What it would signal | Sources | |---|---|---|---|---| | People Inc. core sessions, year on year | −22% | Q2 2026 | AI answers eroding search-referred sessions at a premium publisher; the series went −5%, −13%, −17%, −22% | [215] | | Ozone publisher ad-request volume | −32% to −37% US; −39% to −41% UK | Q2 2026 | fewer ad requests on premium display; fewer widget slots is an inference | [357] | | Ozone eCPM | +7% US; +30% UK | June 2026 | scarcity pricing offsetting volume loss in the UK, not in the US | [357] | | Teads Holding Co. revenue, year on year | −17% | Q2 2026 | decline across the combined company; the filings do not isolate the Outbrain lineage | [271] | | Teads Holding Co. adjusted free cash flow | −$37.9M | H1 2026 | cash burn against total debt of $614.5M carrying value at 30 June 2026 (10% notes due 2030) | [271, 82] | | Teads Holding Co. share price against Nasdaq's $1.00 minimum | second deficiency notice, 11 August 2026 | deadline 8 February 2027 | listing pressure on the second scaled platform | [84] | | Taboola ex-TAC gross profit, year on year | +11.8% reported (about +4.7% under the previous adjustment convention) | Q2 2026 | a signal for scenario one: ex-TAC outgrowing revenue fits a shift beyond the widget; the gap between the two figures is an adjustment-convention sensitivity | [30, 47] | | Taboola FY2026 revenue guidance midpoint | $1,943M (Aug) vs $2,034M (May) | 2026 | top line lowered while profit expectations rose | [30, 348, 347] | | Yahoo agreement amendments | seven, latest effective 2026-03-01 | 2023–2026 | repeated re-basing of performance terms in the largest native supply contract | [404, 405, 406] | | Taboola cost of guarantees, share of TAC | about 13% | Q2 2026 | down from about 18% in 2024; falling guarantees mean cheaper supply or less exclusive supply | [47, 1] | | Dianomi revenue, year on year | +2% H1 2026; +14% July–August | 2026 | small-cap financial native recovering from a 2025 decline | [359] | The paper's read is that scenarios one and two are running at once, at different firms. Scenario three has not started in public. That fits what chapters 3 through 7 found. The moat is supply. The platforms are building beyond the widget, though neither discloses how much growth those lines add. And public causal evidence is missing. The paper cannot say why it is missing. Plausible reasons include these. Tests may have been run and kept private. Measuring effects this small is costly and noisy. The incentive to publish may be weak. And one format is hard to isolate inside a multi-format platform. The paper does not choose between them. ### Failure modes for this paper's own outlook Three things would change the conclusions. The first is a made-for-advertising series for the whole market for 2025 or 2026. It would show whether widget arbitrage has collapsed or grown. No source this project could open provides one. The second is an incrementality result for native alone. Its design and results would be public. It would move the evidence chapter from absence to a finding. The third is Taboola's next annual report. It will show whether the 2026 revenue cut was presentation or demand. It will also show whether its filings begin to say what Teads' already say about AI search. All three are dated events a reader can check. The handoff note lists them. ## 12. Method, limits and corrections *How the paper was built and what the verification pass caught. What could not be reached, and each change made after the challenge passes. It is published so readers can check the paper rather than trust it.* > **In short.** > Separate AI research agents worked from a shared brief. They produced 12 research workstreams, one sweep of the vendor universe and 9 vendor profiles. Each piece then went to a separate verifier agent told to refute it. The research opened and logged 977 sources and put 642 claims in the ledger. A verifier re-checked 311 of those claims against their sources. The verification, argument, scoring and completeness reviews led to 275 substantive corrections. A single scorer scored 24 vendors on 12 dimensions. Of all cells, 58% could be scored. The rest are marked n/e or n/a rather than filled. This is AI-assisted research with a documented process. It is not an external audit, not peer review and not independently validated. ### What the paper set out to answer The paper asks one question. What is native advertising in 2026, who earns from it, and what evidence backs what buyers are told about it? The brief split that into twelve sub-questions, one per workstream. They were fixed before research started: 1. What counts as native? How big is it, by each measurer's own definition? 2. Who sells and distributes it? How concentrated is supply? 3. What does a dollar of native revenue pay for? What do publishers keep? 4. Do readers recognise it as advertising? Does the creative or the disclosure change that? 5. How do the recommendation and delivery systems decide what to show? 6. Is there causal evidence that native spend adds sales or conversions? 7. How clean and transparent is the supply? 8. How do commerce links and first-party data change the unit? 9. How do large buyers buy and control it? 10. What have AI and agentic buying changed? What is still announced rather than deployed? 11. Which rules on disclosure, privacy and AI apply? How are they enforced? 12. Where has the model failed? What would show which way it is heading? Chapters 1 to 11 answer them, roughly in that order. The vendor universe and the nine vendor profiles answer a thirteenth, practical question. Who could a US buyer actually use, and what can each one document? ### How it was built The research ran as parallel streams from one brief. Before any research began, the brief fixed the definitions, the three scope groups, the source hierarchy and the evidence classes. Twelve workstreams covered definitions and market measurement, market structure, financial and publisher economics, and consumer and creative evidence. They also covered delivery technology, causal effectiveness, supply quality, commerce and data, and enterprise buying. Then came AI and agentic execution, disclosure and governance, and failure modes. A thirteenth stream built the vendor universe but did not score it. Nine agents built vendor profiles for Taboola, Teads Holding Co., MGID, Revcontent, Dianomi, Readpeak, Nativo, TripleLift and Sharethrough. Each stream wrote a structured evidence file. It lists sources with their dates of publication, event and access. For each source it also gives geography, period, funding and access limits. It logs findings, with the supporting passage, scope, evidence class, confidence and counterevidence. And it holds data tables for each number that might be charted. Each file then went to a separate verifier agent. The verifier had not seen the research, and it was told to refute it. It picked the claims that mattered and opened the cited source. It checked denominator, unit, geography, date, baseline and population. It looked for a contrary or more primary source. Then it gave a verdict: supported as stated, supported with qualification, contradicted or unresolved. Verifiers wrote their own files and did not edit the research. The verifiers checked up to seventeen claims per file. So a verifier re-checked 317 of 642 ledger claims. Claims outside that sample are marked unresolved in the ledger. A note says they rest on the source the research pass opened. The numerical and citation review traced each number that matters in the prose back to its source. That review is described below. Synthesis, scoring and drafting came next. This work was AI-assisted and in the author's voice, for his review before publication. The paper's argument was not fixed in advance. The brief's candidate theses were tested against the evidence. The headline findings in the abstract are the ones that survived. Several of the author's own starting positions did not, and the handoff note lists them. ![Every stage was built to remove things, and the counts are computed, not typed.](figures/fig-pipeline.svg) *Figure fig-pipeline. **Every stage was built to remove things, and the counts are computed, not typed.** The five stages of this project with the actual outputs of each. The verifier sample covers the material claims per file, not every claim; the rest are labelled unresolved rather than assumed. Corrections include withdrawn claims and score changes from the review passes. Units: counts. Scope: this research project, as of the cutoff. Caveat: Counts describe process volume, not evidence quality; a claim that survived verification is supported by its source, not proven true. Data: data/stats.json* ### How sources were found, and what an absence means There was no bibliographic database search. There was also no formal screening protocol of the kind a systematic review uses. Each research agent worked from the brief's leads and the source hierarchy. Filings, regulators and standards bodies came first. Then came peer-reviewed studies, official technical docs, measurement publishers and the trade press. Vendor material came last. The searches in each workstream mixed vendor and product names, format terms and the workstream's topic. The format terms were native, in-feed, content recommendation, sponsored content and advertorial. Topics included traffic acquisition cost, ad recognition, incrementality, made-for-advertising and agentic buying. Searches ran across the open web, SEC EDGAR, regulator sites, and company developer and help centres. They also used arXiv, DOI resolvers and scholarly metadata services. Nothing published after 27 September 2026 was used. A search result was a lead, never evidence. Each cited source was opened, and the record notes how much of it was read. Sources with limited access were handled the same way throughout. If a source could be read only in part, the register labels it. The label says abstract, partial extract, paywalled excerpt, or access error with text reached by other means. Nothing is quoted from the part that was not read. A study read only as an abstract supplies no effect size beyond what the abstract states. That method limits what an absence can mean. When this paper says a document, study or feature was not found, it means a narrow thing. It was not found in the sources this review opened, in public material, by the cutoff date. It does not mean the thing does not exist. And it says nothing about private contracts. Where a vendor might hold the item privately, the text says so. ### What the process is not Two agents reading the same source are not two independent measurements. A model that checks another model's work is not an external audit. When this paper says a claim was verified, it means a second agent opened the source and read the passage. When it says a figure was recomputed, it means a script recomputed it from the source rows. The formula is in the derived-checks file. No third party has reviewed the paper. No scored vendor has seen it. Nobody funded or commissioned it. Scoring ran as an AI-assisted first pass under the author's rules, then a revision once the verifier records and the final reviews were read, and a re-audit of the format scores after the second outside review. The corrections ledger lists every score that changed. The rules file was written before the first pass, but it carries no independent timestamp. There is no second scorer. So there is no inter-rater reliability statistic, which measures how far two scorers agree. This paper's central complaint is that native vendors grade their own homework. For such a paper, that is the biggest gap left in the method. It is named here rather than left for a reader to find. ### Sampling and scoring, stated plainly The vendor universe is a reasoned sample, not a census. It started from named seeds. It added each native specialist found while researching the twelve workstreams. A primary page was opened for each row. Adult-only, gambling-only and pure ad-fraud networks were left out on purpose. Some vendors surfaced only as names, with no source that could be opened. They were listed but not scored. The matrix scores 24 vendors. Each sells or delivers open-web native as its main business or a clearly separable one. Or it is an intermediary through which native is bought at scale. Each could also be evidenced on at least four of the twelve dimensions, counting a dimension documented as not applicable. Budget comparators such as Meta's own feeds and TikTok are compared on revenue in chapter 2; Meta Audience Network is scored as a publisher-side intermediary. They are not scored, because their business model is not comparable on these dimensions. The twelve dimensions, their definitions and their score anchors were fixed in `data/scoring-rules.json` before scoring began. Scores run 0 to 5. They are bounded by the evidence, not absolute. A 5 means the capability is documented in operational detail. The source is a filing, an official document or an independent observation. A 3 means it is real but thinly evidenced, narrow or in beta. A 1 means adjacent or incidental. A 0 is used only where a source documents absence. The label n/e means the record gave no basis to score. The label n/a means the dimension does not apply to the vendor's model. Each scored cell carries an availability label. It carries an evidence grade, from A (filing or official document) to E (none). It also carries a date, a written reason and at least one source. The build fails if any of those is missing. The same standard applies to each vendor. A large help centre earns grade C, not a higher score. The first draft published no composite score. The reason was that a composite mixes capability with how much evidence there is, and rewards vendors that document more. The final paper carries one, the Native Quotient (NQ). It was added after scoring, at the author's instruction. It is built the way his Contextual Quotient is built, so the two share a construction and a way of reading them. Their scores and bands are relative to different capabilities and samples and are not interchangeable. The design answers part of the objection, not all of it. It has three groups: Depth, Reach and Standing. Each group is the mean of the dimensions scored inside it. So an n/e cell is skipped, not counted as zero. The groups combine at 45/25/30. The result is then shrunk toward the field mean in proportion to how few dimensions are scored. Shrinkage damps a thin record, but chapter 9 shows it does not make missing evidence neutral. The dimension scores were fixed before NQ existed, and none was changed to fit it. The format scores were later re-audited against their anchor after the second outside review; those changes followed the anchor, not the index. Bands are reported instead of ranks, with thresholds set from the field mean. The reweighting test in chapter 9 shows why. Under four other weightings, positions move by up to 5 places. Coverage is the number of dimensions scored out of twelve. It is shown for each row so that thin evidence stays visible. Of 288 cells, 168 carry a score, 83 are n/e and 37 are n/a. ### What could not be reached Access limits shaped this paper. The register records them per source. Of 977 sources, 90 are recorded as not read in full; for a further 137, added by verifiers, the extent read was not recorded. The ones that matter are these. EMARKETER's pages returned access errors to direct requests. Its executive summaries were read by other means. So the 2024 value and any split after 2019 are missing. Chapter 2 says so. OpenAI's and Perplexity's own announcement pages could not be opened. So the ChatGPT Ads and Perplexity facts rest on text reached by other means or on second-hand reports. They are labelled that way. EUR-Lex rejected requests during the research pass. The Digital Services Act articles were later read in the Official Journal text. The AI Act's Article 50 was read from a word-for-word unofficial copy. The AI Omnibus was read from Commission material. Several academic papers could be read only as abstracts. Several SEC filings were too long for the fetch tool. It returned the business and risk sections and only parts of the financial notes. So the reconciliations drawn from them are marked derived. One tooling limit hit the later streams. The research agents shared a web-search quota, and it ran out part-way through the run. So the governance, failure-mode, universe and profile streams reached sources by direct address, not by search. So did all the verifiers. So discovery in those streams is narrower than in the earlier ones. The completeness review below treats that as a known bias rather than pretending it away. Two fetches of PDF documents returned summaries with tables that were not in the documents. Each figure from a PDF in this paper was re-extracted word for word by script. The incident is recorded so a reader knows the class of error the process had to guard against. ### The challenge passes, and what they changed After the verifier pass, an interim draft got a separate, detailed review with one hundred findings. Each finding was checked against its cited source where one could be reached. Most were accepted. A few were accepted in part, with reasons. The changes are in the corrections ledger, tagged with the review's finding number. That review is a critique of a draft, not an audit or a peer review. Five reviews were then run on the final text. Their findings, and what was done about each, are in the corrections ledger below, tagged F, and in the handoff note. A fact review checked each number that matters against the source record. It did the same for each date, entity, ownership claim, product status, technical claim, citation and superlative. An argument review attacked the central thesis. It asked whether the scope was chosen to favour it, and whether native and adjacent markets were lumped together. It asked whether a weak comparator flattered the result, whether correlation became cause, and whether counterexamples changed the reading. A scoring review challenged the sample, the dimensions, evidence consistency and the treatment of missing data. A completeness review looked for missing stakeholders, formats, geographies, failed businesses and contrary studies. A numerical and citation review reran each derived calculation and resolved each reference ID. It checked that each cited page supports the sentence it is attached to. It also matched each figure to the canonical dataset. A second outside review of the rebuilt package returned 38 findings, tagged R2 in the ledger. Where a finding made a factual claim, it was checked against the primary source before anything changed. One was rejected: Taboola files a 10-K, not a 20-F. The rest led to corrections in the text, the scoring, the NQ method, the print edition and the package. Two research leads in the brief itself were wrong, and are recorded. Sharethrough's merger with Equativ was announced in June 2024, not 2023, and Adyoulike was bought by OpenWeb, not Opera. **Corrections ledger** | ID | Original claim or score | Challenge | Evidence checked | Verdict | Change made | Reason | Affected | |---|---|---|---|---|---|---|---| | X01 | Research brief: Sharethrough merged with Equativ in 2023. | Research lead checked against deal announcements. | Equativ announced and closed the Sharethrough deal on 12 June 2024; 2023 is when Bridgepoint took its stake in Equativ. [66] | contradicted | Chapters 3 and 11 and the Sharethrough profile use June 2024. | The brief was wrong. | Ch. 3, 11; Sharethrough profile | | X02 | Research brief: Adyoulike was bought by Opera. | Research lead checked against the acquisition record. | OpenWeb acquired Adyoulike in April 2022, for a reported $100 million. [66] | contradicted | Chapters 3 and 11 name OpenWeb; the scenario table carries the 2022 acquisition. | The brief was wrong. | Ch. 3, 11 | | X03 | First draft: no recognition research has been published since 2022. | Verifier D searched for post-2022 studies. | A 2022 US experiment (600 adults) and a 2024 recognition study (567 students) were found. [104, 105] | contradicted | Chapter 5 cites both and says neither is a representative replication. | The absence claim was false. | Ch. 5, 12 | | X04 | Huebner et al.: participants classified ads by product visibility, logos and buttons rather than by the label. | Verifier D re-read the paper. | Disclosure labels were the second most cited cue (95 mentions), after product visibility (178); the authors call disclosures an important indicator. [116] | contradicted | Chapter 5 reports the ranking of cues with the label second. | The draft dropped the label count. | Ch. 5 | | X05 | Taboola GenAI release: AI ads beat human ads on click-through within advertiser, campaign and day; conversions not reported. | Verifier E and the interim-draft review re-read the release and located the paper. | The 0.76% vs 0.65% is the raw comparison; under the tightest controls the ads performed comparably. The paper found no significant conversion effect and its sample totals differ by version. [142, 160] | contradicted | Chapter 6 separates raw and controlled results, cites the paper and states the version differences. | The draft overstated the result. | Ch. 6 | | X06 | Under CPA objectives on CPC billing, the vendor absorbs the CPC-to-CPA translation risk. | Verifier E, then finding 02 of the interim-draft review. | The advertiser still pays per click whether or not a conversion follows; MGID calls target CPA a directional goal. [145, 139] | contradicted | Chapter 6 and the delivery figure say automation transfers bid authority, not financial risk. | Billing terms decide who carries the risk. | Ch. 6; delivery figure | | X07 | eMarketer native values for 2020 to 2025 are not public: a six-year gap. | Verifier A searched for public eMarketer charts. | eMarketer's January 2023 chart gives 2019 to 2023 ($47.29B to $97.46B) and restates 2019 from $43.9B to $47.3B. [22] | contradicted | Chapter 2 and the spend figure add the 2023 vintage; the gap is 2024 only. | A public source existed. | Ch. 2, 12; spend figure | | X08 | Taboola FY2026 guidance is revenue $1,993M to $2,054M (February range). | Verifier A checked the latest update before the cutoff. | The August 2026 update superseded it: revenue $1,930M to $1,956M, ex-TAC $772M to $783M. | contradicted | All guidance references use the August ranges; the guidance figure shows every update. | Superseded figure. | Ch. 4, 11; guidance figure | | X09 | The August 2026 revenue cut was not explained. | Verifier C read the earnings-call transcript. | Management attributed the cut to leaving publisher relationships, largely in Greater China, and to Google deprecating Explore More; the new ex-TAC range includes an added-back write-off. [352] | supported with qualification | Chapter 11 and the guidance figure report the explanation and the like-for-like range of about $760M to $771M. | The explanation was on the call, not in the release. | Ch. 11; guidance figure | | X10 | Teads Holding Co. half-year adjusted EBITDA was $7.8 million. | Verifier C re-read the Q2 release. | The release reports $7.7 million. [83] | contradicted | Chapter 4 uses $7.7 million. | Transcription error. | Ch. 4 | | X11 | Taboola's traffic acquisition cost is its publisher payout ratio. | Verifier C and finding 32 of the interim-draft review. | TAC also includes exchange purchases, up-front and incentive payments and non-cash amortisation; Yahoo-related presentation changed in 2024. | supported with qualification | Chapter 4 and the economics figure describe TAC as a recorded cost, not a payout ratio. | Definition was too narrow. | Ch. 4; economics figure | | X12 | Amazon Publisher Services Native ads are filled by Amazon advertiser demand. | Verifier B re-read the announcement. | The May 2025 release announces the format but does not state availability, markets or access for other DSPs. [74] | supported with qualification | Chapter 3 says what was announced and what was not stated. | Announcement read as deployment. | Ch. 3 | | X13 | Dianomi received a unanimously recommended offer. | Verifier B re-read the Rule 2.7 announcement. | The board intends to recommend unanimously; the contingent 24p depends on publishers moving to Taboola standard terms; conditions include shareholder approval and no Phase 2 review. [60, 31] | supported with qualification | Chapter 3 states the terms and conditions. | Terms were compressed. | Ch. 3 | | X14 | Google Network fell 1.9%. | Verifier A read the MD&A. | Network combines AdSense, AdMob and Ad Manager; the decline came from AdSense, partly offset by AdMob. | supported with qualification | Chapter 2 says the web piece fell by more than 1.9%. | Line mixes web and app. | Ch. 2 | | X15 | Life360 completed the Nativo acquisition on 5 January 2026. | SWOT drafting for Nativo re-read the 8-K. | The 8-K filed 5 January says the deal completed on 2 January 2026. [59] | contradicted | Chapters 3 and 10 use 2 January and cite the 8-K. | Filing date read as completion date. | Ch. 3, 10 | | X16 | MGID, Taboola and Teads product pages state no spending floor. | SWOT drafting for MGID re-read the help centre. | MGID sets a $100 minimum deposit (recommends $650) and a $50 minimum campaign budget. [269, 270] | contradicted | Chapter 10 separates MGID's documented minimums from the others. | Overgeneralised. | Ch. 10 | | X17 | Score rationales: MGID has no public advertiser API and is TCF-registered; Dianomi's publisher count fell from 341 to under 300. | SWOT drafting cross-checked the scores against the profiles. | MGID's help centre has an API section whose scope was not read; its TCF ID is from a Prebid listing, not the truncated vendor list; Dianomi's counts use different bases. [423, 587, 263] | supported with qualification | Rationales rewritten; scores unchanged. | Rationales overstated the record. | Matrix: MGID, Dianomi | | X18 | Source register: one entry per URL with query strings ignored. | SWOT drafting found search pages for different companies merged into one source. | The URL normaliser dropped query strings, so different EDGAR and Google News searches collapsed. | contradicted | Query strings are kept (tracking parameters stripped); the register was rebuilt. | Build bug. | Register; reference list | | X19 | Methodology: no composite ranking is published. | Author instruction after scoring. | The Native Quotient was added, built like the author's Contextual Quotient: group means, 45/25/30, shrinkage toward the field mean, bands set from the field mean, and a four-weighting sensitivity test. | supported with qualification | Chapter 9 adds NQ with its limits; chapter 12 discloses when and why it was added; no dimension score changed. | The shrinkage and group means answer the original objection. | Ch. 9, 12; matrix; NQ figure | | X20 | TripleLift and Sharethrough profiles unscored. | Profiles landed after the first scoring pass. | Both scored on the same anchors; sellers.json files graded C as vendor-published documents. [505, 526] | supported as stated | Two matrix rows added with rationales and sources. | Completes the scored set. | Matrix; ch. 9 | | X21 | Dianomi documents no conversion tracking, publisher controls or placement-level reporting. | Dianomi profile verifier read the help centre. | The help centre documents an image pixel, server-to-server events and AppsFlyer/Adjust; DoubleVerify and IAS integrations; publisher-level reporting; Supply Manager exclusions; named SSPs. [465, 466, 620, 623] | contradicted | Four Dianomi scores raised (conversion data n/e to 4, measurement 2 to 3, transparency 2 to 3, interoperability 3 to 4); SWOT and chapter 9 rewritten. | The first pass had not read the help centre. | Matrix: Dianomi; SWOT; ch. 9 | | X22 | Taboola has no agent-protocol implementation. | Taboola profile verifier searched Taboola developer documentation. | Taboola operates an official Realize MCP server with read and write campaign management, documented in July 2026; no AdCP membership found. [402, 403] | contradicted | Interoperability score 3 to 4; chapter 11 and the abstract say Taboola runs its own agent server. | Absence claim was wrong. | Matrix: Taboola; ch. 11; abstract | | X23 | MGID sellers.json lists 508 sellers; Google is its only intermediary; its audience claim is unchanged since 2018. | MGID profile verifier downloaded the full file. | 2,571 sellers across 2,522 domains and six intermediaries; audience claims range from 850 million to over 1 billion over time. [429] | contradicted | Score rationales and SWOT corrected; scores unchanged. | Truncated fetch. | Matrix: MGID; SWOT | | X24 | Revcontent sellers.json lists 502 sellers; no TAG participation. | Revcontent profile verifier parsed the raw file. | 436 entries across 399 domains; TAG Registered since 2018, current seal unverified. [455, 453] | contradicted | Rationale and SWOT corrected. | Wrong count. | Matrix: Revcontent; SWOT | | X25 | Nativo's $63 million 2025 revenue is trade-press reporting. | Nativo profile verifier read the earnings-call transcript. | The figure is an unaudited statement by Life360's CFO on the March 2026 call. [502] | supported with qualification | SWOT corrected; the margin threat removed because traffic acquisition cost was a small part of the margin change. | Mis-attributed source. | SWOT: Nativo | | X26 | Amazon has not confirmed TripleLift as a supply source. | TripleLift profile verifier found Amazon's own announcement. | Amazon named TripleLift a third-party native supply source in December 2023. [510] | contradicted | Score rationale and SWOT corrected. | Absence claim was wrong. | Matrix: TripleLift; SWOT | | X27 | The only randomised native evidence is Aribarg and Schwartz. | Verifier D. | Sahni and Nair (Marketing Science 2020) ran randomised field experiments on native search ads with more than 200,000 users. [114] | contradicted | Chapter 5 cites both and separates search from the open-web widget. | Superlative was wrong. | Ch. 5 | | X28 | Realize does not document its auction rule. | Verifier E read the Realize help centre. | Enhanced CPC reduces to second price; Creative Expedition charges second price; Fixed Bid charges the bid; programmatic clears first-price in private marketplaces and second-price on the open exchange. [147, 148, 149] | contradicted | Chapter 6 states the documented rules by strategy and route. | Absence claim was wrong. | Ch. 6 | | X29 | Neither large platform documents a pre-campaign inventory list. | Verifier G. | Outbrain DSP, Teads' programmatic product, documents publisher include lists at setup. [264] | supported with qualification | Chapter 10 and the decision tool limit the statement to the native self-serve platforms. | Scope was too broad. | Ch. 10; decision tool | | X30 | ANA: made-for-advertising spend 6.2% in 2024; cohort split of 2.1% against 0.9% in Q1 2026. | Verifier G opened the Q1 2026 benchmark. | The ANA restated 2024 as 1.1%; Q1 2026 exposure was roughly flat across cohorts; values across its measures are not comparable. [288, 289] | contradicted | Chapter 10 reports the series with the restatement and the comparability warning. | Secondary report misread. | Ch. 10 | | X31 | Teads' filings do not mention retail media. | Verifier H read the full 10-K. | The 10-K names retail media as a planned growth environment and a risk, with no revenue, product or partner. [46] | contradicted | Chapter 8 corrected. | Partial read. | Ch. 8 | | X32 | The AI Omnibus recorded no change to the Article 50 date. | Verifier K. | The Omnibus gave generative systems already on the market until 2 December 2026 for the Article 50(2) marking duty. [334, 335] | contradicted | Chapter 10 corrected. | Missed deferral. | Ch. 10 | | X33 | Nineteen states have comprehensive privacy laws. | Verifier K. | At least 21 by the cutoff, adding Oklahoma and Alabama. [327, 328] | supported with qualification | Chapter 10 adds the 2026 enactments. | Stale count. | Ch. 10 | | X34 | No AdCP owner page names native as a format. | Verifier J. | The AdCP docs refer to native formats and its catalogue carries native entries; a canonical native format is deferred to 3.2. [388, 389] | contradicted | Chapter 11 corrected. | Absence claim was wrong. | Ch. 11 | | X35 | LiveIntent agreed to be acquired; Unity native unconfirmed; Engageya and plista status unknown. | Verifier U. | Zeta closed LiveIntent on 21 October 2024; LevelPlay documents native ads; Engageya discontinued on 30 September 2024; plista was reported ceasing in March 2024. [731, 707, 370, 371] | contradicted | Universe rows and the chapter 11 table corrected. | Page observations taken as status. | Universe; ch. 11 | | X36 | Teads offers no attention guarantee. | Teads profile verifier. | Teads' blog described a pilot of attention-to-outcome guarantees built on Adelaide scores in May 2025, with no terms or status published. [309] | supported with qualification | Chapter 10 adds the pilot beside the Vevo guarantee. | Omission. | Ch. 10 | | X37 | The resumed research run renumbered verifier D's sources. | Build review. | Citations in chapters 5 and 8 and one SWOT line pointed at the wrong papers after the re-run. [124, 250, 104, 105] | contradicted | Citations remapped by URL. | Process error caught before release. | Ch. 5, 8; SWOT: Taboola | | F01 | Abstract, finding 4; Ch. 6 'From clicks to conversions, and who holds the risk', para 2; Ch. 10 closing callout: "The machine chases conversions, and the advertiser still pays for clicks." | Final argument review (P0). Both firms' own filings describe campaigns priced per acquisition, and Teads prices on 'cost-per-incremental action'. Pay-per-acquisition moves financial risk to the platform. That contradicts headline 4, and it contradicts Ch. 6's 'None of the product documentation reviewed offers one' ('one' being a contract that guarantees the outcome or charges per conversion). Ch. 10's inshort already says 'Both do describe camp | data/claim-ledger.json I-F16 review_note quotes Taboola's 10-K ('Generally, our charges are based on a CPC, CPM or CPA basis ... For campaigns priced on a performance-based CPA basis, the Company generates revenue when a user makes an acquisition') and Teads' 10-K ('cost-per-incremental action (CPA)', 'variable consideration based on ... measured outcomes'). The I-F03 review_note says the same. I ran EDGAR full-text | supported with qualification | Headline 4: 'The machine chases conversions, and most buying is still billed per click.' Ch. 6: 'Both filings also describe CPA-priced campaigns, and Teads books revenue on cost-per-incremental action. Neither discloses what share of revenue these carry or how incremental is measured.' Add to the Ch. 10 buyer callout: ask for CPA or incremental-action pricing and the measurement definition behind it. Mention Teads' i | Accepted and applied. | Abstract, finding 4; Ch. 6 'From clicks to conversions, and who holds the risk', | | F02 | Abstract, finding 3; Ch. 5 inshort: "In the studies opened, 7% to 37% of those tested saw article-style and in-feed native ads as advertising." | Final argument review (P1). The statement is false as written. The FTC lab study was opened, its roughly 47% to 68% is plotted in the paper's own recognition figure, and it falls outside the range. Hyman's per-ad results run to 72%, and the 'Paid Ad' bar reached 56%. The two lowest figures (7% and 9%) are single sponsored articles (advertorials), not the recommendation widgets that Taboola and Teads mainly sell. Only the FTC study tested a widg | claim-ledger D-F08 (FTC: 'from ... approximately 47% to ... approximately 68%'; native conditions +23 points); figures/fig-recognition.json caption includes 'the FTC's 47% to 68%'; D-F05 (per ad 21% to 72%); D-F07; D-F01 and D-F14 stimuli are one sponsored article each. | contradicted | 'In three studies of sponsored articles and in-feed ads, 7% to 37% recognised them as ads, against 81% for standard ads. In the FTC's lab, about 47% did, and 68% with its recommended labels. Only the FTC study tested a recommendation widget.' Make the same edit in the Ch. 5 inshort. | Accepted and applied. | Abstract, finding 3; Ch. 5 inshort | | F03 | Abstract, finding 5 and the inshort; Ch. 7 inshort: "In large Facebook experiments, observational comparisons like that, even with statistical adjustment, overstated lift by roughly three to thirteen times" | Final argument review (P1). Placed next to the vendor evidence, the 3x-13x range reads as the size of the bias in that evidence. The chapter itself says (fix R58) that no multiplier transfers. The designs also differ. Gordon et al. measured conversion outcomes from platform logs with double machine learning and matching. The native vendor evidence is a Kantar attitudinal brand-lift survey with matched control groups. Gordon et al. also found ob | claim-ledger F-F04 review_note (563 experiments, 663 pairs, Nov 2019 to Mar 2020; 'comparatively better for prospecting campaigns rather than for remarketing'; Meta co-author); F-F21 (Kantar survey, exposed versus matched control); corrections-ledger R58 (fix scoped to Ch. 7 only); Ch. 7 standfirst 'can overstate'. | supported with qualification | Abstract: 'Vendor evidence compares those who saw the ads with those who did not. In Facebook conversion experiments, such designs overstated lift, by about 3 to 13 times at the median. No one has measured the bias for native brand-lift surveys.' Inshort: 'designs that can overstate lift'. Ch. 7: '663 treatment-control pairs from 563 experiments'. | Accepted and applied. | Abstract, finding 5 and the inshort; Ch. 7 inshort | | F04 | Abstract, finding 5; Ch. 7 callout 'The absence, stated precisely'; Ch. 5 'Clicks, attention and trust': "The causal evidence is missing, not negative." | Final argument review (P1). The paper uses adjacent-market evidence only in one direction. It imports Facebook RCTs to discount vendor designs, but never reports that the same RCTs found positive median incremental lift for ads the paper itself calls native by format (Ch. 1: 'A social in-feed unit is native by format'). It cites Sahni and Nair only for disclosure. Yet that randomised field test of native search ads (>200,000 users) tracked conv | claim-ledger F-F04 (RCT median lifts 29%, 18% and 5%); D-F17 review_note ('tracked clicks and conversions ... found incremental conversions came from users who saw ads and then chose organic listings'); source-register V-D-S01; evidence/verify/D.json V-D-S08 (LaBrecque). | supported with qualification | Add to Ch. 7, and a clause to abstract finding 5: 'Randomised tests of native-format ads elsewhere, in Facebook campaigns and native search ads, found positive incremental effects. The gap is specific to the open-web widget.' Add LaBrecque 2024 to Ch. 12's unscreened leads. | Accepted and applied. | Abstract, finding 5; Ch. 7 callout 'The absence, stated precisely'; Ch. 5 'Click | | F05 | Abstract, finding 6; Ch. 9 'What survives every weighting'; Ch. 11 closing read: "The growth is moving beyond the widget." | Final argument review (P1). This is stated as a finding, but no filing splits revenue by format. Ch. 3 admits 'It does not disclose how much of its growth each line adds' (corrections R28). The one split the filings do allow points the other way at Taboola. In Q2 2026, revenue billed through Yahoo rose about $32.8M while total revenue rose about $11.3M, so all of the quarter's growth, and more, came through the largest native supply partner, an | claim-ledger C-F07 (Yahoo-billed revenue $79.267M in Q2 2026 vs $46.455M; 16.6%); L-F09 review_note (Q2 revenue $476.8M vs $465.474M); L-F08 review_note (Explore More, Greater China publisher exits); C-F18 (CTV 13%); C-F09 review_note (average revenue per scaled advertiser 'in part driven by the testing of ad formats with Yahoo'). | supported with qualification | 'The platforms are building beyond the widget: Realize display, DeeperDive, and Teads CTV at 13% of revenue. Neither discloses how much growth those lines add. In Q2 2026, revenue billed through Yahoo grew faster than Taboola's total.' Replace 'leans on' with 'is growing'. Soften Ch. 9's 'argument in one row' and Ch. 11's 'Growth is more and more in lines beyond the widget'. | Accepted and applied. | Abstract, finding 6; Ch. 9 'What survives every weighting'; Ch. 11 closing read | | F06 | Abstract inshort; Ch. 3 inshort: "The two listed platforms depend more on a few supply partners than on any advertiser." | Final argument review (P1). The supply-rights pillar of the thesis rests on one firm's disclosure. Every concentration figure is Taboola's. For Teads, the research file lists partner concentration as a gap it could not fill, and what Teads does disclose points to diffuse supply and a strong demand side: about 10,000 media owners, top-20 tenure of 7 years, 93% of revenue direct from advertisers, and about 50 JBPs averaging $4M. The Ch. 3 inshort | evidence/research/B.json gaps entry 'Teads Holding Co. 10%+ customer/publisher concentration disclosures'; claim-ledger B-F13, I-F03; B-F03 review_note ('The two thresholds use different denominators'); the C-S04 passage in evidence/research/C.json ('approximately 34% of our gross revenues generated from Advertisers on digital properties'). | supported with qualification | Abstract: 'Taboola depends more on two supply partners than on any advertiser. Teads does not disclose partner concentration.' Ch. 3 inshort: 'Yahoo and Microsoft are about 34% of revenue from advertisers on its partners' properties.' | Accepted and applied. | Abstract inshort; Ch. 3 inshort | | F07 | Ch. 3 heading and para 'Supply is concentrated. Demand is not.'; Ch. 3 inshort: "Supply is concentrated. Demand is not." | Final argument review (P1). The paper's own Ch. 4 data contradicts this. Revenue billed through Yahoo, with Yahoo as the billing entity for advertiser spend, was 10.5% of 2025 revenue, more than Taboola's ten largest advertisers combined (<10%). It was 16.6% in Q2 2026. One counterparty sits on both sides of the business. That strengthens the Yahoo-dependence reading, but it removes 'demand is not concentrated' and 'Their edge is not demand' as | claim-ledger C-F07 ('Yahoo's advertiser spend on the Company's network, for which Yahoo is the billing entity': $201.6M, 10.5% in 2025; 16.6% in Q2 2026); B-F03 (top ten advertisers <10%, none >3%). | supported with qualification | Retitle to 'Supply is concentrated, and so is one demand channel'. Add: 'Outside Yahoo, demand is diffuse. Revenue billed through Yahoo was 10.5% of 2025 revenue and 16.6% in Q2 2026, more than the ten largest advertisers combined.' | Accepted and applied. | Ch. 3 heading and para 'Supply is concentrated. Demand is not.'; Ch. 3 inshort | | F08 | Abstract inshort; Ch. 9 inshort: "Measurement is the weakest dimension in the field. No vendor scores above 3 on it. The reason: none documents holdout or incrementality tools run with an independent partner." | Final argument review (P1). The paper's own matrix does not support 'weakest'. Measurement has the lowest maximum score, but not the lowest average: commerce, optimisation and publisher controls all average lower. The stated reason is also wrong for Taboola. Its measurement cell cites an independent incrementality partner (Measured), and Ch. 7 says so. What holds the cell at 3 is the evidence-grade cap ('Vendor marketing can earn at most a 3'), | data/capability-matrix.json, computed over scored cells: measurement_support mean 2.67 (n=9, max 3); commerce_capabilities 2.25 (n=12); optimisation 2.45 (n=11); publisher_controls 2.64 (n=14). Taboola measurement rationale: 'a third-party integration marketed by Measured for incrementality' (src F-S33). Ch. 9: 'Vendor marketing can earn at most a 3'. | contradicted | 'No vendor scores above 3 on measurement. Incrementality tooling is documented only on marketing pages, including a third-party integration for Taboola, and none with published tests. On average, commerce and optimisation score lower.' Make the same edit in the abstract inshort. | Accepted and applied. | Abstract inshort; Ch. 9 inshort | | F09 | Ch. 2 'What the word covers in each source' and 'Answering question two'; Ch. 2 inshort: "It also counts in-app rewarded video and sponsored outstream video. It excludes banners and pop-ups" | Final argument review (P1). The paper presents this as the scope of the EMARKETER forecast. The verifier found that the list comes from a generic, AI-assisted EMARKETER FAQ, not from a forecast scope note, and that no page opened says whether the forecast counts rewarded or outstream video. The paper then builds on it in its answer to question two: 'The one native series counts outstream and rewarded video'. This correction was recorded in the | claim-ledger A-F09 review_note ('The five-format list appears under the generic FAQ heading ... closes with We prepared this article with the assistance of generative AI tools ... Whether the forecast itself counts rewarded or outstream video is not stated on any page opened') and its final_wording; A-F14 (eMarketer's video line counts native outstream video, unresolved). | supported with qualification | 'An AI-assisted EMARKETER FAQ lists in-feed, sponsored content, widgets, rewarded and outstream video as native formats. The forecast's own scope is not published beyond display ads that reflect the form and function of their surroundings.' Delete 'counts outstream and rewarded video' from the inshort and from 'Answering question two'. The social-share argument, which rests on the 2017-2019 vintages, stands. | Accepted and applied. | Ch. 2 'What the word covers in each source' and 'Answering question two'; Ch. 2 | | F10 | Ch. 11 scenario callouts and the closing read; Abstract 'What would change these findings': "Ruled out if two things happen. Publisher sessions level off. And the platforms' ex-TAC grows with no new lines outside native." | Final argument review (P1). As written, the scenarios cannot fail. Scenario 2's second condition can never be observed. Non-native lines (Realize display, DeeperDive) are already live, and no format split is disclosed. Scenario 1 is ruled out only if 'Widget revenue is disclosed', which Taboola does not do, so without that disclosure it survives by default. Scenario 3 counts 'A native vendor on an AdCP or AAMP implementer list with a reporting | Ch. 11 text; Ch. 3 'It does not disclose how much of its growth each line adds'; claim-ledger L-F03 review_note ('the release does not map segments to legacy Outbrain products'); indicator row 'the key disconfirming signal' names no scenario. | contradicted | Tie each disconfirmer to a series that is already disclosed. S1 is ruled out if Taboola's FY2026 revenue grows faster than ex-TAC with no new non-native line disclosed. S2 is ruled out if People Inc. core sessions are within 5% year on year for two quarters and Taboola's guarantee share of TAC rises. S3 counts only a holdout study with a named partner. Drop the protocol-listing trigger from S3, and say which scenario | Accepted and applied. | Ch. 11 scenario callouts and the closing read; Abstract 'What would change these | | F11 | Ch. 8 closing callout 'Where the commerce argument lands': "The largest independent platform in retail media, the budget next door, grew 2% last year." | Final argument review (P1). This is a weak comparator used to cool the commerce thesis. One firm's segment, which fell 17% in Q4 because of scope changes with two clients, stands in for a budget that the paper's own Ch. 2 shows growing 18%. The callout implies retail media is flat. The accurate point is narrower: native vendors capture and report none of it. | claim-ledger H-F16 (Criteo retail media +2%; Q4 down 17% 'on scope changes with two clients'); Ch. 2 citing A-S03: 'Commerce media ... reached $63.4 billion, up 18%'. | supported with qualification | 'Commerce media grew about 18% in 2025 (IAB/PwC). The largest independent retail-media platform grew 2% because of client-scope changes. The native vendors report none of it.' | Accepted and applied. | Ch. 8 closing callout 'Where the commerce argument lands' | | F12 | Ch. 4 'Who pays the cost' (final para); Ch. 10 inshort: "made-for-advertising spend fell from 15% in 2023 to a median of 0.4% in the third quarter of 2025. It then rose to 1.1% in the first quarter of 2026 as 'AI slop' sites appeared" | Final argument review (P2). This compares a 2023 average for 21 advertisers with a 2025 median for a self-selected panel. The ANA says those measures cannot be compared, and Ch. 10 criticises the playbook for exactly this move. 'Rose ... as AI slop sites appeared' also implies a cause. The panel grew from 54 to 86 marketers that quarter, and the ANA only named AI slop as an emerging sub-type. | claim-ledger G-F07 review_note (waterfall values 'can't be directly compared to benchmark medians'; the 2023 median was 10%); I-F20 limitations ('Not like-for-like'); G-F11 review_note (participants 'from 54 to 86'; ANA 'identifying AI slop as an emerging subtype'); Ch. 10 on the playbook's line. | supported with qualification | 'The median fell from 10% in 2023 to 0.4% in Q3 2025. It was 1.1% in Q1 2026, a quarter in which the panel grew from 54 to 86 marketers and the ANA named AI slop as an emerging sub-type.' Make the same change in the Ch. 10 inshort. | Accepted and applied. | Ch. 4 'Who pays the cost' (final para); Ch. 10 inshort | | F13 | Ch. 13 Definitions ('AdCP and AAMP', 'Contextual targeting'); Ch. 6 closing callout: "Neither names native as a format. No native vendor implements either at the cutoff." | Final argument review (P2). Three corrections reached the chapter bodies but not the summary text or the glossary. Ch. 11 says 'AdCP can describe native. Its 3.1.24 docs refer to native formats' (X34), and Kargo is an AdCP member (R89). The contextual definition says it 'does not use the reader's identity or history', which contradicts Ch. 1's corrected text (R15). Ch. 6's callout says learning rules 'slow or exclude low-volume campaigns', but | corrections-ledger X34, R89, R15, R50, R51; Ch. 1 'not a promise that no personal data is used'; claim-ledger E-F26 review_note (modes without a target 'have no such entry thresholds'). | supported with qualification | Glossary: 'AdCP's docs refer to native formats; a canonical native format is deferred to 3.2. No large native platform was found implementing either.' Contextual: 'uses page or session signals as its basis; it does not by itself exclude personal data.' Ch. 6 callout: 'slow low-volume campaigns and keep them out of target-CPA bidding'. | Accepted and applied. | Ch. 13 Definitions ('AdCP and AAMP', 'Contextual targeting'); Ch. 6 closing call | | F14 | Ch. 3 'Two companies, two supply strategies' (Yahoo para); Ch. 11 protocol para and indicator table: "A thirty-year exclusive needed three amendments in its third year." | Final argument review (P2). The dates are wrong. The deal closed on 17 January 2023, so Amendment 7 (1 March 2026) falls in the fourth year, and Amendment 4 (1 February 2025), which the verifier found, falls in the third. The paper also says 'at least three times' in Ch. 3 but 'seven' in Ch. 11. Separately, Ch. 3 dates OpenRTB Native 1.2 to March 2017 and Ch. 11 to July 2017. | claim-ledger B-F02 review_note ('The exhibit also cites Amendment No. 4 dated February 1, 2025'); B-F23 review_note (change log: '1.2 March 2017'); J-F26 (standards-index release date July 2017, unresolved). | contradicted | 'Three amendments took effect between February 2025 and March 2026, Nos. 4 to 7 among them.' Use one count throughout ('at least seven, numbered'). Use March 2017 in Ch. 11, or say 'finalised March 2017, indexed July 2017'. | Accepted and applied. | Ch. 3 'Two companies, two supply strategies' (Yahoo para); Ch. 11 protocol para | | F15 | Ch. 5 'What the units actually carry', final para: "No one outside that platform has measured whether it is still true." | Final argument review (P2). The same paragraph cites a February 2026 independent preprint, crawled over five months of 2025-26, that attributes problematic advertorial ads mainly to Taboola and Outbrain. The closing sentence contradicts it. | Ch. 5 citing V-D-S02; evidence/verify/D.json notes Papadogiannakis et al. 2026. | supported with qualification | 'One independent 2026 preprint, with a different codebook, suggests it still is. No peer-reviewed replication exists.' | Accepted and applied. | Ch. 5 'What the units actually carry', final para | | F16 | Ch. 4 'The arbitrage loop, and who pays for it': "That is why Forbes ran the subdomain for seven years." | Final argument review (P2). This assigns a motive from a single case. The source records Forbes calling the subdomain insignificant. The same kind of motive claim was removed from Ch. 11 under R91. | Ch. 4 citing G-S08/G-S09 ('It called the subdomain an insignificant part of its business'); corrections-ledger R91. | supported with qualification | 'Each party was paid, which may explain how the subdomain ran for seven years. The record does not show why Forbes kept it.' | Accepted and applied. | Ch. 4 'The arbitrage loop, and who pays for it' | | F17 | Ch. 9 inshort and bands; Abstract 'How it was done' and inshort: "puts one vendor in the front rank, 12 in Strong and 14 in Mid" | Final argument review (P2). The Strong band starts at 57, one point above the field NQ of 56, the score a vendor with average evidence gets. Six rows with only four scored dimensions therefore land in Strong: AppLovin, Google AM, Google AdSense, Meta AN, Microsoft, and StackAdapt, whose measurement cell reads 'Not opened'. Several rank above fully scored Readpeak and Dianomi, so the band counts partly reflect the prior. The abstract says 'The r | data/capability-matrix.json rows (coverage 4, NQ 57-58, band Strong); nq.bands_rule; data/stats.json nq_field 56, nq_strong_min 57; Ch. 12 'added after scoring, at the author's instruction'. | supported with qualification | Report band counts only for vendors with coverage of at least 8, or mark thin rows as 'unbanded'. In the abstract, add 'The NQ summary was added after scoring, using thresholds borrowed from the author's contextual index.' | Accepted and applied. | Ch. 9 inshort and bands; Abstract 'How it was done' and inshort | | F18 | Abstract 'Interests to declare' and finding 8; Ch. 11 scenario three: "None of these companies is scored in this paper." | Final argument review (P2). The disclosure leaves out that the author's employer, Samba TV, appears on AdCP's member logos. Scenario 3 then treats a native vendor joining an AdCP or AAMP implementer list as a sign that proof is arriving, which promotes adoption of the body the author co-leads. Headline 8 is also the only headline claim the verifier never checked, and its source notes that the full 141-member roster was not enumerated. | claim-ledger J-F08: passage 'AdCP member logos: Yahoo, PubMatic, Scope3, Samba TV, LG Ads, Kargo ...'; review_outcome 'unresolved', checked_independently false; confidence_reason 'the full 141-member roster was not enumerated'. | supported with qualification | Add 'Samba TV is an AdCP member.' Remove protocol listing as a Scenario 3 trigger. Verify J-F08 against the full roster, or qualify headline 8 as 'on the member pages opened'. | Accepted and applied. | Abstract 'Interests to declare' and finding 8; Ch. 11 scenario three | | F19 | Ch. 10 inshort and 'Who buys, and through what': "A significant majority of Taboola's revenue comes from about 2,200 advertisers who work with it directly" | Final argument review (P2). This merges two separate 10-K sentences. The 2,200 Scaled Advertisers work with Taboola 'directly, or through advertising agencies'. The 'significant majority' of revenue comes from the subset that works directly. The verifier flagged exactly this conflation. | claim-ledger I-F15 review_note ('The file's passage truncates the count sentence before or through advertising agencies, which reverses its meaning'). | supported with qualification | 'About 2,200 Scaled Advertisers, direct or through agencies, drove 84% of Q4 2025 revenue. Taboola says a significant majority of revenue came from Scaled Advertisers working with it directly.' | Accepted and applied. | Ch. 10 inshort and 'Who buys, and through what' | | F20 | Ch4, 'The arbitrage loop' last paragraph; Ch10 In short; Ch10 'What a buyer can inspect', second paragraph: "fell from 15% in 2023 to a median of 0.4% ... It then rose to 1.1% in the first quarter of 2026 as "AI slop" sites appeared" | Final fact review (P1). Chapter 4 runs three measures into one trend: 15% is the 2023 study's average across 21 advertisers, 0.4% is the Q3 2025 median, and 1.1% is the Q1 2026 waterfall figure. The ANA says these cannot be compared. The Chapter 10 In short repeats the mix ('15% in 2023 to a median under 1% in 2025'), even though Chapter 10's body quotes the ANA warning. The words 'as AI slop sites appeared' assert a cause, but the uptick c | claim-ledger G-F07 final_wording: the median fell from 10% (2023) to 0.39% (Q3 2025), and 'The ANA says waterfall, average and median values are not comparable'. G-F11 review_note: 0.4-0.6% through 2025 is the waterfall; 'participating marketers rose from 54 to 86, with active contributors reaching 66'; 'a change in participant mix is a stronger alternative explanation for the uptick'. I-F20 note: the all-environment | contradicted | Use one measure on each line: either the median (10% in 2023 to 0.39% in Q3 2025) or the average/waterfall (15% to 0.4-0.6% in 2025 to 1.1% in Q1 2026). Replace 'as AI slop sites appeared' with 'while the participant base nearly doubled; the ANA named AI slop as an emerging sub-type'. In Chapter 10 change '21 to 39 marketers' to '21 to 66 active marketers, depending on the quarter'. | Accepted and applied. | Ch4, 'The arbitrage loop' last paragraph; Ch10 In short; Ch10 'What a buyer can | | F21 | Ch2 In short; Ch2 'What the word covers in each source', paragraph 3; 'The definitional gulf'; 'Answering question two';: "It also counts in-app rewarded video and sponsored outstream video. It excludes banners and pop-ups" | Final fact review (P1). The paper presents the format list from an AI-assisted eMarketer FAQ as the definition behind the $147.98B forecast. The ledger's final wording says the page does not establish that the forecast counts rewarded or outstream video. The claim that outstream and rewarded video sit inside the native figure is a central step in the 'fence' argument. | claim-ledger A-F09 final_wording: 'An AI-assisted EMARKETER FAQ describes native formats as including ... Whether the forecast itself counts rewarded or outstream video is not stated on any page opened.' review_note: the list sits under a generic FAQ heading; the page says 'We prepared this article with the assistance of generative AI tools'; the only published forecast definition is the Jan 2023 chart note ('reflect | supported with qualification | Attribute the format list to the FAQ and say it may not be the forecast's scope. Keep social in-feed, which the vintage splits support. Drop or hedge outstream and rewarded video in the In short, 'The definitional gulf', 'Answering question two' and the inclusion-matrix caption. | Accepted and applied. | Ch2 In short; Ch2 'What the word covers in each source', paragraph 3; 'The defin | | F22 | Ch3 'Two companies, two supply strategies', paragraph 1: "Historically, most of its publisher deals were exclusive." | Final fact review (P1). This overstates the filing. The 10-K says most agreements required exclusivity OR other preferred-usage incentives, so it does not show that most deals were exclusive. The sentence carries the 'supply lock-in' thesis, and the Taboola supply_access rationale in the matrix repeats it. | claim-ledger B-F05 final_wording and review_note: 'Historically, the majority of Taboola's agreements with digital properties have typically required them to provide it exclusivity or other incentives based on preferred usage' ... 'does not establish that a majority of contracts are exclusive ... This matters for any supply lock-in headline.' Also evidence/verify/B.json. The capability-matrix.json Taboola supply_acce | supported with qualification | 'Historically, most of its publisher deals required exclusivity or other preferred-usage incentives.' Make the same change in the matrix rationale. | Accepted and applied. | Ch3 'Two companies, two supply strategies', paragraph 1 | | F23 | Abstract, In short: "The two listed platforms depend more on a few supply partners than on any advertiser." | Final fact review (P1). Only Taboola's record supports this. Teads Holding Co. discloses no supply concentration. Its 10-K says no media partner reached 10% of its traffic acquisition cost and no marketer reached 10% of revenue. The supply-dependence claim therefore rests on one company, while the abstract states it for both. | claim-ledger C-F19 passage: 'none of the Company's media partners accounted for 10% of its total traffic acquisition costs'; 'No single marketer accounted for 10% or more of the Company's total revenue'. Taboola-only support: evidence/research/C.json risk factor, Yahoo and Microsoft about 34%. A-F20 note also flags the unqualified phrase 'the two listed ... platforms' (Dianomi is listed too). | supported with qualification | 'The larger listed platform, Taboola, depends more on two supply partners than on any advertiser; both large US-listed platforms lock supply with long contracts and guarantees.' | Accepted and applied. | Abstract, In short | | F24 | Ch4 'The arbitrage loop', Forbes paragraph and 'Who pays'; Ch10 'The publisher's decision': "It drew more than 70% of its readers from paid Taboola and Outbrain placements." | Final fact review (P1). This drops part of the source's wording and so overstates the widgets' share in the paper's best-documented case of widget-fed arbitrage. The Chapter 10 line 'It carried 200 ads a session' turns one observed 52-slide slideshow session into a general rate. | Checked the live source (adalytics.io/blog/ads-observed-on-www3-forbes-subdomain): 'appears to source more than 70% of its readership through paid display ads on Taboola, Outbrain, and other paid traffic acquisition sources'; 'One consumer was shown ... 201+ ads total while viewing a 52 slide slideshow'. claim-ledger G-F15 review_note flags the same overstatement: 'drops and other paid traffic acquisition sources, so | supported with qualification | Ch4: 'more than 70% of its readers, per Similarweb estimates, from paid display ads on Taboola, Outbrain and other paid traffic sources'. Ch10: 'one observed session carried more than 200 ads'. | Accepted and applied. | Ch4 'The arbitrage loop', Forbes paragraph and 'Who pays'; Ch10 'The publisher's | | F25 | Ch1 callout 'Three contested boundaries', sponsored answers paragraph: "Chapter 8 records one live product and one retired one. It records no research at all on whether consumers recognise them as ads." | Final fact review (P1). Chapter 8 contradicts both sentences. It records several live products: ChatGPT ads, Microsoft Copilot ads, Amazon's sponsored prompts (generally available March 2026), ads in Google AI Overviews and AI Mode, and Taboola's DeeperDive ad engine, with Perplexity as the retired one. It also cites user research on recognising sponsored content in AI answers. | Ch8 'The newest surface': 'A 2024 study of users found they often missed product placements in generated search answers (V-D-S04)'. Ch5 callout cites the March 2026 AAA experiment on sponsored content in AI overviews (AUD-S42). corrections-ledger R72 and R73 added these products to chapter 8 but did not update chapter 1. | supported with qualification | 'Chapter 8 records several live products, one retirement and one trial, and only thin, non-representative research on whether users recognise them as ads.' | Accepted and applied. | Ch1 callout 'Three contested boundaries', sponsored answers paragraph | | F26 | Ch2 'The budget destinations, on their own terms': "It fell 1.9% to $29.8 billion, its second decline in a row." | Final fact review (P2). The count is wrong. Google Network also fell in 2023 ($32,780M in 2022 to $31,312M), so 2025 was its third annual decline in a row. | Checked the live source, Alphabet's FY2023 10-K on SEC EDGAR, revenue-by-type table: Google Network 31,701 (2021), 32,780 (2022), 31,312 (2023). claim-ledger A-F16 gives 31,312 / 30,359 / 29,792 for 2023-2025. | contradicted | 'its third annual decline in a row'. | Accepted and applied. | Ch2 'The budget destinations, on their own terms' | | F27 | Ch8 'The newest surface', Google paragraph: "It also announced a Direct Offers pilot." | Final fact review (P2). The pilot is dated wrongly. Google introduced Direct Offers on 11 January 2026, and Google Marketing Live 2026 in May expanded an existing pilot. The ledger's own passage says 'expanded'. | Checked the live source, blog.google/products/ads-commerce/agentic-commerce-ai-tools-protocol-retailers-platforms/ (11 Jan 2026): 'we're now introducing Direct Offers. This new Google Ads pilot...'. claim-ledger J-F15 passage for GML 2026: 'Direct Offers pilot expanded'. | supported with qualification | 'It also expanded the Direct Offers pilot it had introduced in January 2026.' | Accepted and applied. | Ch8 'The newest surface', Google paragraph | | F28 | Ch11 'The traffic shock, measured where it can be', paragraph 1: "Its digital revenue still grew 6% in the second quarter of 2026, on higher programmatic rates." | Final fact review (P2). This names the wrong driver. People Inc.'s advertising revenue was flat ($174.2M). Higher programmatic rates only offset lower volumes. The 6% growth came from performance marketing (+13%) and licensing (+23%). | claim-ledger L-F17 passage: 'Advertising: $174.2M (flat)'; 'Higher open programmatic advertising revenue due to higher rates, partially offset by lower impression volumes'; performance marketing $68.8M (+13%); licensing $47.0M (+23%) including the Meta content partnership. | contradicted | 'Its digital revenue still grew 6%, driven by performance marketing and a Meta content licence; advertising held flat as higher programmatic rates offset lower volumes.' | Accepted and applied. | Ch11 'The traffic shock, measured where it can be', paragraph 1 | | F29 | Ch8 'Content-to-commerce at the publisher': "It fell 22% in the second, as page views fell 13%." | Final fact review (P2). This ties a full-year figure to one half. The 13% drop in unique page views is for the full year. Future also names consumer confidence, FX and closures, not audience alone. | claim-ledger H-F20 review_note: 'the 13% UPV decline is a full-year product-waterfall driver, not an H2 figure, and Future names consumer confidence, FX and closures alongside audience ... should be softened'. | supported with qualification | 'It fell 22% in the second half; for the full year Future cites a 13% fall in unique page views, weaker consumer confidence, FX and closures.' | Accepted and applied. | Ch8 'Content-to-commerce at the publisher' | | F30 | Ch10 In short and 'Who buys, and through what', paragraph 2: "A significant majority of Taboola's revenue comes from about 2,200 advertisers who work with it directly" | Final fact review (P2). Two 10-K sentences are merged. The 2,200 Scaled Advertisers work with Taboola 'directly, or through advertising agencies'. The filing says separately that a significant majority of revenue came from the Scaled Advertisers who work with it directly. The body sentence 'These deal with it direct, not through an agency' repeats the merge. | claim-ledger I-F15 review_note: 'The file's passage truncates the count sentence before "or through advertising agencies", which reverses its meaning. The stream summary ... conflates the two sentences.' | supported with qualification | 'Taboola had about 2,200 Scaled Advertisers, working directly or through agencies; a significant majority of its revenue came from those working with it directly.' | Accepted and applied. | Ch10 In short and 'Who buys, and through what', paragraph 2 | | F31 | Ch3 'Two companies', Yahoo paragraph: "A thirty-year exclusive needed three amendments in its third year." | Final fact review (P2). Both counts are off. The exhibit also cites Amendment No. 4, dated 1 February 2025, so the text should say at least four amendments (Nos. 4-7) between February 2025 and March 2026. Amendment 7 took effect on 1 March 2026, in the deal's fourth year (it closed 17 January 2023). Chapter 11's indicator table says 'seven', so 'at least three times' also understates. | claim-ledger B-F02 review_note: 'The exhibit also cites Amendment No. 4 dated February 1, 2025 ... the summary's "amended at least three times in 2025-2026" should read "at least four times"'. | supported with qualification | 'The contract has been amended at least seven times; Amendments 4 to 7 took effect between February 2025 and March 2026.' Delete 'in its third year'. | Accepted and applied. | Ch3 'Two companies', Yahoo paragraph | | F32 | Ch11 retirement table (Nativo row) and agents paragraph; Ch3 In short: "\| 2026-01-05 \|" | Final fact review (P2). There are three internal inconsistencies. (1) The Nativo row says completed 2026-01-05, but Chapters 3 and 10 say 2 January, as corrections-ledger X15 requires. (2) Chapter 11 says OpenRTB Native 'has been at version 1.2 since July 2017', but Chapters 1 and 3 say 1.2 was finalised in March 2017. (3) 'Ten days before this paper's cutoff' is wrong: 18 to 27 September is nine days. | corrections-ledger X15: 'Chapters 3 and 10 use 2 January and cite the 8-K'. B-F23 note: 'Native spec change log ... 1.2 March 2017'. J-F26 passage: the Tech Lab index lists 'Release Date: July 2017'. B-F20: announcement dated 2026-09-18. | contradicted | Set the Nativo completion to 2026-01-02 (reported 5 January). Use 'finalised March 2017 (the Tech Lab index lists July 2017)' in all three places. Change 'Ten days' to 'Nine days'. | Accepted and applied. | Ch11 retirement table (Nativo row) and agents paragraph; Ch3 In short | | F33 | Ch10 'The rules, in three jurisdictions', CCPA paragraph: "Since 29 March 2023 the rules have made a compliant browser signal count as an opt-out." | Final fact review (P2). The start date is wrong. The original CCPA regulations, effective August 2020, already required businesses to treat user-enabled global privacy controls as a valid opt-out of sale; the AG enforced this against Sephora in 2022. The March 2023 regulations extended the duty to 'sharing' and to linked profiles. The corrections-ledger R78 fix brought in the 2023 start date. | Checked live: the text of 11 CCR §999.315(c) (Aug 2020 CCPA regulations), as quoted by Baker McKenzie and Hunton: businesses 'shall treat user-enabled global privacy controls ... as a valid request' to opt out of sale. AUD-S43 dates only the 2023 package. | contradicted | 'Since August 2020 the rules have made a browser signal count as an opt-out of sale; since 29 March 2023 that extends to sharing and to any linked profile.' | Accepted and applied. | Ch10 'The rules, in three jurisdictions', CCPA paragraph | | F34 | Ch8 'What the adjacent systems show' and closing callout; Ch3 buying-routes heading and paragraph 1; Ch4 'Post-acquisiti: "The largest independent retail-media platform tells a flatter story." | Final fact review (P2). The paper states superlatives the record does not establish, including ones the verifiers flagged: (a) Criteo as 'the largest independent retail-media platform'; (b) 'Every major buying and serving platform supports the format' and the heading 'every platform supports native'; (c) 'Connected TV grew fastest of its lines'; (d) 'the largest native deal of the decade', for an outstream-video/CTV acquisition, with no com | H-F16 review_note: 'Criteo's 10-K claims only "the world's largest independent Retail Media API program". Drop the superlative or attribute it.' B-F25 note: 'unqualified superlative ("Every major buying and serving platform") ... does not cover every major platform (e.g. Amazon DSP buy-side, Yahoo DSP, Criteo)'. C-F18 note: 'Fastest-growing line is the researcher's characterisation, not company wording.' | supported with qualification | 'Criteo, which calls its program the world's largest independent Retail Media API program'; 'The major platforms checked (DV360, GAM, AdSense, Microsoft Advertising, The Trade Desk, APS) support the format'; 'Connected TV grew 67%'; delete 'of the decade' or name the comparison set. | Accepted and applied. | Ch8 'What the adjacent systems show' and closing callout; Ch3 buying-routes head | | F35 | Ch7 In short: "The IAB's own November 2025 guidance rates such designs as weak proxies" | Final fact review (P2). There are two overreaches. (1) The IAB guidance is about commerce-media sales incrementality and never mentions brand-lift surveys, so putting the Kantar and Nielsen designs in its 'weak' class is the paper's reading, not the IAB's rating. A matched exposed/control design could arguably sit in the IAB's 'model-based counterfactual' row. (2) 'some platforms shortened view-through attribution windows' rests on one plat | F-F17 review_note: 'the IAB document is about commerce-media sales incrementality and never mentions brand-lift surveys. A Kantar matched exposed/control design could arguably sit in the model-based counterfactual row.' F-F13 final_wording and note: 'deprecated ... in its Ads Insights API (reporting) ... neither opened source says Ads Manager campaign attribution settings changed'. | supported with qualification | 'By this paper's reading, such designs fall in the IAB's weak, proxy class.' And: 'In January 2026 Meta removed 7- and 28-day view windows from its reporting API.' | Accepted and applied. | Ch7 In short | | F36 | Ch11 In short and 'Agents and protocols', platform-layer paragraph: "Each automated buying system whose docs were opened keeps some human approval step." | Final fact review (P2). The paper's own records contradict this universal claim. Realize+'s Decision Engine 'automatically moves budget in real time'. SpendGuard caps spend automatically and is on by default. Taboola's official MCP server can change campaigns. The body concedes Realize+ and Apostra act 'on their own within rules', yet the In short and the body still say each system keeps some approval step. | E-F21 passage: Decision Engine 'automatically moves budget in real time to the highest-performing campaigns'. G-F22 note: SpendGuard is on 'by default'. J-F24 is unresolved and was not selected for verification. | contradicted | 'Most systems whose docs were opened keep a human approval step for launch; Realize+ reallocates budget automatically within buyer-set rules.' | Accepted and applied. | Ch11 In short and 'Agents and protocols', platform-layer paragraph | | F37 | Ch11 'Agents and protocols', paragraph 2: "It was listed as released on 26 February 2026 at the umbrella level." | Final fact review (P2). AAMP's status changed before the cutoff and the text misses it. IAB Tech Lab announced AAMP 3.0 on 22 September 2026, with its OpenProposal spec in public comment until 22 October 2026. The chapter describes AAMP only as of February, although its 'change each week' contrast relies on the ledger record that cites AAMP 3.0. The ledger also says 'Agentic Audiences' has no release listed, which is not the same as 'proof | claim-ledger J-F06 final_wording: 'AAMP 3.0 was announced on 2026-09-22, and its new OpenProposal specification is in public comment until 2026-10-22 ... Agentic Audiences has no release listed.' J-F26 final_wording also cites 'AAMP 3.0'. | supported with qualification | Add: 'On 22 September 2026 Tech Lab announced AAMP 3.0; its OpenProposal spec is in public comment until 22 October.' Make the audiences wording match the record. | Accepted and applied. | Ch11 'Agents and protocols', paragraph 2 | | F38 | Ch9 'The wider universe' and 'What survives every weighting'; Ch12 method: "at least four of the twelve dimensions had evidence" | Final fact review (P2). Two scored vendors break the stated inclusion rule. JW Player/JWX has 3 scored cells and no n/a cells. Amazon APS Native has 3 scored cells. Chapter 9 itself says the large platforms have 'three or four scored dimensions'. Chapter 12 also says Meta is not scored, yet Meta Audience Network is. | data/capability-matrix.json rows: amazon-aps-native coverage 3 (NQ 56), jwx coverage 3 (NQ 55). scoring-rules.json scored_set (c): 'at least four of the twelve dimensions can be evidenced'. | supported with qualification | Remove the two rows (27 vendors becomes 25, and the dependent stats need recomputing), or restate the rule as 'at least three', or say that n/a counts as evidenced. Clarify that Meta Audience Network is scored as an intermediary. | Accepted and applied. | Ch9 'The wider universe' and 'What survives every weighting'; Ch12 method | | F39 | Ch10 'The rules, in three jurisdictions', US paragraph: "There were at least ten fake-review matters between August 2023 and July 2026" | Final fact review (P2). This overstates a record the verifier contradicted. The Legal Library shows 'roughly ten review-related matters'. The listed dates are last-updated dates. Two of the matters (Iconic Hearts, B.E.S.T. GDR) concern reviews only loosely. The Rytr order was set aside in December 2025. | claim-ledger K-F09 (contradicted) final_wording: 'roughly ten review-related matters dated 2023-2026 (dates are last updated dates). The Rytr order ... was reopened and set aside by a 2-0 Commission vote on 22 December 2025'. | supported with qualification | 'The FTC's Legal Library lists roughly ten review-related matters dated 2023-2026 (by last-updated date); one, the Rytr order, was set aside in December 2025.' | Accepted and applied. | Ch10 'The rules, in three jurisdictions', US paragraph | | F40 | Subtitle/abstract (line 3) and Ch.1 'Three scope groups, one counting rule' (line 83); Ch.3-4 economics: "Open-web native is a supply-rights business sold as a performance product." | Final completeness review (P0). Ch.1 places sponsored editorial (branded content/advertorials) in the core market, and most Ch.5 recognition evidence tests sponsored articles, yet no chapter measures who sells sponsored editorial, what it earns or how it has fared. Publisher studios sell it direct with no platform supply rights, so the headline describes recommendation/in-feed networks, not 'open-web native'. The collapse of branded-content-led mod | chapters/01-defining.md line 29 ('It covers recommendation units, in-feed native on open-web pages and sponsored editorial'). data/source-register.json holds no publisher branded-content studio or sponsored-content revenue source; its only 'T Brand' hits are false matches on 'Outbrain True Engagement Brand Lift'. evidence/research/*.json hold no branded-content revenue record. The only seller of sponsored content tha | supported with qualification | Either narrow the subtitle, abstract findings 1-2 and the Ch.3 opener to 'open-web recommendation and in-feed native networks' and add a Limits line saying sponsored-editorial economics were not measured, or add a half-page on sponsored editorial using publisher filings already in the register (the BuzzFeed content line) plus the Nativo profile. | Accepted and applied. | Subtitle/abstract (line 3) and Ch.1 'Three scope groups, one counting rule' (lin | | F41 | Ch.10 box 'For the buyer, in three sentences' (line 1182); decision tool row 1 (line 1127): "Run your own holdout or geo test, and expect it to be noisy." | Final completeness review (P0). The central buyer recommendation gives no inputs for judging whether a test is feasible. There is no US native CPC, CPM or conversion-rate band, no minimum-detectable-effect or sample-size example, and no fallback for buyers too small to test. The paper's own numbers suggest many native buyers cannot run an informative test, which changes the recommendation for them. | Ch.7 line 455 and study table line 1853: Lewis & Rao's 25 tests spent $2.8M in total, about $112K each (derived), and still had median ROI intervals more than 100pp wide, with informative tests needing more than 10M person-weeks. Ch.3 line 212: Taboola's Scaled Advertisers, the top of its demand, average about $204,000. Ch.6 lines 416-425: learning-phase floors. Ch.10 line 1085: spend ranges 'come from third-party bl | supported with qualification | Add a worked sizing box (assumed CVR, CPC band, MDE, required users/weeks), state the spend level below which a holdout is uninformative, and give the fallback: a longer geo test, pooling across flights, or capping native spend as unmeasured. Make the three-sentence advice conditional on scale. | Accepted. The final-review version gave a budget comparison but no baseline rate, effect size, power, significance level, randomisation unit or sample size, so it was only partly applied. Round 2 (R2-05) replaced it with a worked power calculation. | Ch.10 box 'For the buyer, in three sentences' (line 1182); decision tool row 1 ( | | F42 | Ch.5 'Clicks, attention and trust...' (line 359); Ch.7 'Two facts...' (line 465) and 'The absence, stated precisely' (li: "They found no evidence that typical disclosure formats fooled users. But that is search, not the open-web widget [112]." | Final completeness review (P1). The same randomised experiment (more than 200,000 users) also measured the ads' causal effect. The native ads benefited advertisers, and most incremental conversions came from exposure (users later clicking the advertiser's organic listing), not from clicks on the ad. It is the only randomised incrementality result on a native format in the source set. It is positive, and it bears directly on Ch.6-7: CPC billing, cli | I opened the [112] abstract at https://ideas.repec.org/a/inm/ormksc/v39y2020i1p5-32.html. It says: 'we find that native advertising benefits advertisers'; 'incremental conversions due to advertising are not driven by users clicking on the native ads'; 'mere exposure seems sufficient to produce most of the incremental effect'. The reference list gives [112] as 'Access: not recorded (added by verifier)'. No claim-ledge | supported with qualification | Add a paragraph to Ch.7 and a row to the study table reporting the lift result with its scope (mobile restaurant search, not open-web widgets). Qualify abstract finding 5: no open-web study exists, but the one randomised native-format test found (in search) was positive and worked through exposure, not clicks. Link this to Ch.6's point that CPC billing prices the click rather than the exposure. | Accepted and applied. | Ch.5 'Clicks, attention and trust...' (line 359); Ch.7 'Two facts...' (line 465) | | F43 | Ch.7 'What this means...' (line 500); Ch.10 contract paragraph (line 1136); decision tool row 2 (line 1128): "Both should be disclosed before the campaign, as the retail-media guidelines require [181]." | Final completeness review (P1). The IAB/MRC retail-media guidelines do not govern native platforms, so 'require' overstates the basis for the contract ask. The cross-media standard that does cover outcome claims is missing: the MRC Outcomes and Data Quality Standard (Sept 2022), backed by the ANA, 4A's and ACA, which covers attribution, MMM and experiments. The paper also never says whether any native vendor's outcome reporting is MRC-accredited un | I opened the MRC Outcomes and Data Quality Standard (Final) PDF on mediaratingcouncil.org. Section 2.3.4.5 says lookback windows 'must be disclosed up front, in advance of campaign execution and measurement' and that 'Measurement providers must establish empirically supported limits to the length of a lookback period'. source-register.json has 0 hits for 'Outcomes and Data Quality'. | supported with qualification | Cite this standard as the cross-media benchmark for disclosing windows and lag. Reword the retail-media reference as an analogy ('as the retail-media guidelines require of retail media networks'). Add the standard to the contract asks and the decision tool. | Accepted and applied. | Ch.7 'What this means...' (line 500); Ch.10 contract paragraph (line 1136); deci | | F44 | Ch.1 box 'Three contested boundaries, and where this paper draws them' (lines 89-96): "The three cases that matter most are set out below." | Final completeness review (P1). Creator and influencer sponsored content (and podcast host-reads) is native under the FTC's resemblance test and is the main target of the Endorsement Guides the paper cites. The paper's own evidence depends on it: Eisend pools influencer posts, and Lord & Taylor turned on about 50 influencer posts. Yet the scope groups never classify it as core, adjacent or excluded, and no size is given. A reader asking how big nat | I opened IAB's release of 20 Nov 2025: US creator ad spend was $29.5B in 2024 and $37B in 2025, measured as 'sponsored content, amplified sponsored content, and planned creator adjacencies'. In the paper: line 1150 (Lord & Taylor, about 50 influencer posts), line 1839 (Eisend formats include influencer), line 1166 (Endorsement Guides row). The source register has no creator-economy source. | supported with qualification | Boundary ruling added to chapter 1; the IAB creator-spend figure was not added because its source was not opened here. | Accepted and partly applied. | Ch.1 box 'Three contested boundaries, and where this paper draws them' (lines 89 | | F45 | Ch.3 'The second tier has fewer owners...' (line 224) and consolidation figure caption (line 222); Ch.10 'The rules, in : "Between June 2024 and September 2026, the independent tier of open-web native fell into fewer hands." | Final completeness review (P1). The 'platform power' chapter and the rules chapter both leave out the competition-law record. First, the October 2019 Taboola–Outbrain merger between today's 'two large US-listed platforms' went through a DOJ second request, was cleared around July 2020 and was terminated in September 2020. Second, there is the E.D. Va. ad-tech antitrust ruling against Google, which Teads cites in its 3 Aug 2026 damages suit and Equa | I opened the Davis Polk matter page: DOJ second request, approval by about 22 July 2020, deal terminated in September 2020. Search only: the CNBC 8 Sep 2020 report 'Taboola and Outbrain end talks to merge' (403 when fetched). References [742] and [743] are Teads' 8-K of 3 Aug 2026 'Following Federal Court Antitrust Ruling'; evidence/profiles/teads.json has the passage 'following the E.D. Va. ad tech antitrust ruling' | supported with qualification | Named as not covered in chapter 12; the Google suits are cited in chapter 4; the 2019 to 2020 merger review is not added. | Accepted and partly applied. | Ch.3 'The second tier has fewer owners...' (line 224) and consolidation figure c | | F46 | Abstract, 'Interests to declare' (line 48): "None of these companies is scored in this paper." | Final completeness review (P1). The statement is accurate but incomplete. The author's current employer, Samba TV, appears in the paper as one of Teads' claimed measurement-partner integrations and as an AdCP member. Teads is scored second on NQ (64), and its CTV business, where Samba TV operates, is a headline growth line (abstract finding 6). A reviewer would expect the declaration to name that relationship. | Line 720, Teads measurement row: '30+ partner integrations claimed (Kantar, IAS, DoubleVerify, Lumen, Samba TV)' [406]. Line 1268: AdCP members include Samba TV. Line 30: 'Teads Holding Co. leans on connected TV.' | supported with qualification | Add one sentence saying that Samba TV is named in Teads' materials as a measurement partner and is an AdCP member, and that Teads' CTV business, which the paper discusses, overlaps Samba TV's market. | Accepted and applied. | Abstract, 'Interests to declare' (line 48) | | F47 | Ch.8 'Content-to-commerce at the publisher' (lines 509, 531, 539): "Publisher commerce lines are set by affiliate terms and traffic, not by ad tech." | Final completeness review (P2). The chapter leaves out Google's site reputation abuse policy (announced March 2024 and tightened in November 2024). The policy targets third-party commerce and coupon sections on publisher domains, and it explicitly addresses 'advertorial' and 'native advertising' pages. It is a third, platform-policy driver of publisher commerce and sponsored-content performance, and a de facto standard for how sponsored content mus | I opened developers.google.com/search/docs/essentials/spam-policies. It exempts 'advertorial' or 'native advertising' type pages whose purpose is to share content directly with readers 'rather than hosting the content to manipulate search rankings', and it gives as a violation 'sponsored reviews of payday loans written by a third-party'. Search only: trade reports naming Forbes Advisor, CNN Underscored and WSJ Buysid | supported with qualification | Add a paragraph to Ch.8 naming search-platform policy as a driver alongside affiliate terms and traffic. Note its explicit carve-out for native and advertorial pages that are integrated with the host site. | Accepted and applied. | Ch.8 'Content-to-commerce at the publisher' (lines 509, 531, 539) | | F48 | Ch.11 'Retirements: brands, products and the difference' (lines 1215, 1231); Ch.6 (line 412); Sharethrough profile (line: "This project could open just one documented case of a publisher dropping a widget. It is nine years old." | Final completeness review (P2). The failure record misses the clearest product retirements. Yahoo shut its own Gemini native marketplace and moved native demand to Taboola in 2023, yet Ch.6 still says 'On Taboola, Outbrain, MGID and Yahoo, the advertiser still pays per click' and describes the Yahoo system as persisting. Yahoo also closed its SSP in February 2023. Microsoft shut its Invest (Xandr) DSP on 28 Feb 2026, but the Sharethrough profile st | evidence/research/E.json lists 'Yahoo (Gemini native, historical)' with 2026 status unknown. Search only: Axios, 4 Oct 2023, on the Taboola–Yahoo deal going live (Gemini shut); VideoWeek, 10 Feb 2023, 'Yahoo Closes its SSP'; Digiday, 'Microsoft Advertising is closing the Xandr DSP' (shutdown by 28 Feb 2026); Nieman Lab, Oct 2016, 'Slate, The New Yorker dump those terrible Around the Web links' (403 when fetched). sou | supported with qualification | Named as not covered in chapter 12; the Sharethrough best-fit list no longer names Xandr; the other retirements are not added. | Accepted and partly applied. | Ch.11 'Retirements: brands, products and the difference' (lines 1215, 1231); Ch. | | F49 | Ch.4 'Post-acquisition losses, impairment and financing pressure' (line 299); Ch.11 indicator table (lines 1308-1319): "The filings neither show a liquidity event nor rule one out." | Final completeness review (P2). Two material 2026 facts appear only in the Teads profile, not in the chapter a buyer reads for vendor stability or in the watch list. The first is a second Nasdaq minimum-bid-price deficiency notice (11 Aug 2026), with a reverse split contemplated and an 8 Feb 2027 compliance deadline. The second is the 3 Aug 2026 lawsuit against Google, with its disclosed retaliation risk. References [740], [742] and [743] are in th | Teads profile lines 700, 707 and 710 ('Second Nasdaq minimum-bid-price deficiency (notice 2026-08-11); reverse split contemplated'; 'Litigation against Google (filed 2026-08-03)'). The reference list has [740] (regained compliance, June 2026) and [742]/[743] (8-K on the Google suit). grep finds them only on reference-list lines. | supported with qualification | Add one sentence to Ch.4 on the listing deficiency and the litigation. Add an indicator row to Ch.11: Nasdaq $1.00 bid-price compliance by 8 Feb 2027. | Accepted and applied. | Ch.4 'Post-acquisition losses, impairment and financing pressure' (line 299); Ch | | F50 | Ch.10 'The rules, in three jurisdictions' (line 1150) and US rows of the instruments table (lines 1164-1171): "Enforcement since 2023 has gone to endorsements and reviews." | Final completeness review (P2). The paper does acknowledge that it ran no defined enforcement-case search. Three US instruments that are not enforcement cases are still missing. The first is the FTC's September 2023 staff perspective on blurred ('stealth') advertising to children, the Commission's most recent native-specific policy statement, which says disclosures are unlikely to work for young children. The second is the COPPA Rule amendments (ef | Search only: FTC, 'Protecting Kids from Stealth Advertising in Digital Media: A FTC Staff Perspective' (Sept 2023, approved 3-0); Federal Register, COPPA Rule, 22 Apr 2025 (effective 23 Jun 2025, compliance 22 Apr 2026); BBB National Programs, whose fast-track SWIFT process accepts native-advertising disclosure cases. source-register.json has 0 hits for 'Stealth Advertising'; its only COPPA hit is a Prebid adapter li | supported with qualification | Named as not covered in chapter 12. | Accepted and partly applied. | Ch.10 'The rules, in three jurisdictions' (line 1150) and US rows of the instrum | | F51 | Ch.10 'The rules, in three jurisdictions' (line 1150ff); Dianomi and TripleLift profiles: "The legal floor under all of this asks for much the same things." | Final completeness review (P2). The rules chapter is general-purpose, but native demand is concentrated in regulated verticals. The paper itself names insurance and mortgage advertorials and investment pitches (line 367) and a pharma format (TripleLift, line 930), and Dianomi is a financial-content specialist earning 78% of its revenue in the US (line 303). Sector rules go unmentioned: FINRA Rule 2210 guidance on native advertising (Regulatory Noti | Search only: FINRA Regulatory Notice 17-18 (April 2017), which says native advertising must comply with Rule 2210 and prominently disclose the firm's name; the ABA Business Law Today article on CIPA pixel litigation (Aug 2024). source-register.json has 0 hits for FINRA. In the paper: lines 303, 367, 545 and 930. | supported with qualification | Named as not covered in chapter 12. | Accepted and partly applied. | Ch.10 'The rules, in three jurisdictions' (line 1150ff); Dianomi and TripleLift | | F52 | Taboola profile, measurement row (line 682); Ch.9 'Standing is where the field is thin' (line 639): "No named third-party verification partner (IAS, DoubleVerify, MOAT, TAG), no viewability standard and no incrementality/holdout product were found" | Final completeness review (P2). Industry certification status is recorded for some vendors but not others. MGID's TAG certification and Revcontent's lack of TAG participation (X24) are recorded, but TAG's 2025 certified list also names Taboola, Outbrain, Teads, TripleLift and Sharethrough, and their status is not recorded. That makes Standing comparisons uneven. The paper also lacks any statement of which certifications and accreditations (TAG seal | I opened TAG's release of 6 Mar 2025 on PR Newswire: Taboola, Outbrain Inc., Teads, MGID, TripleLift and Sharethrough are listed among 2025 certified companies, and Revcontent is not; seals per company are in the TAG Registry, which I did not check. Scoring rules (data/scoring-rules.json) count 'third-party verification' and 'independent audit' in the measurement and transparency anchors. | supported with qualification | Named as not covered in chapter 12; Revcontent's TAG registration is recorded. | Accepted and partly applied. | Taboola profile, measurement row (line 682); Ch.9 'Standing is where the field i | | F53 | Abstract finding 6 (line 30); Ch.1 (lines 67, 104); Ch.2 (line 139); universe rows (lines 1043-1045) vs matrix (lines 59: "Teads Holding Co. leans on connected TV." | Final completeness review (P2). The paper treats Teads' CTV growth as 'beyond native', but much of it is home-screen OEM tiles ('3D CTV homescreen takeovers'), which sit in the TV interface's content rows. Ch.1's list of placement surfaces has no TV surface and no boundary ruling. In-app is labelled three ways: 'outside this paper's core' in Ch.2, 'core native (in-app platform)' in the universe table, and 'adjacent' in the matrix. | Lines 702 and 715 (CTV HomeScreen via LG, VIDAA, Samsung, TiVo; homescreen takeovers); line 1242 (CTV HomeScreen placements in Teads Ad Manager); line 139 ('in-app ad space outside this paper's core'); lines 1043-1045 vs 590 and 597. | supported with qualification | Boundary rulings added to chapter 1 and finding 6 reworded; universe and matrix labels for in-app were not relabelled. | Accepted and partly applied. | Abstract finding 6 (line 30); Ch.1 (lines 67, 104); Ch.2 (line 139); universe ro | | F54 | Ch.1 format and surface axes (line 67); universe LiveIntent row (line 1066): "It can also be a retailer's site, an AI assistant's answer or a newsletter." | Final completeness review (P2). Mail and newsletter inventory is inside the largest supply deals (Yahoo Mail, Microsoft Outlook and Microsoft 365, Realize 'mail' environments) and is sold by Teads, Revcontent and Dianomi. Yet it is never analysed as a surface: its share, how readers recognise ads there, and the rules that apply. The dedicated email/newsletter native vendor (LiveIntent, now Zeta) is listed but not scored, and newsletter sponsorship | Line 200 (the move into Outlook and Microsoft 365); line 1004 (Realize: 'mail, mobile, and premium editorial environments'); line 714 (Teads 'newsletter Platforms'); line 1008 (Revcontent email monetisation); line 816 (Dianomi email); line 1066 (LiveIntent 'core native (unseeded; email/newsletter native)', not in the matrix). | supported with qualification | Named in the chapter 1 boundary rulings and in the limits as documented only in passing. | Accepted and partly applied. | Ch.1 format and surface axes (line 67); universe LiveIntent row (line 1066) | | F55 | Abstract (line 10); Ch.10 heading 'The rules, in three jurisdictions' (line 1148); Limits (line 38): "with the United States as the main market" | Final completeness review (P2). The paper is partly upfront (it says US primary), but it does not name the Asia-Pacific gap as a limit. It scores Asia-Pacific vendors (Dable, popIn) with no Asia-Pacific market or rules context, notes that Taboola is leaving publishers 'largely in Greater China', and never mentions Japan's industry native-ad guidelines (JIAA). For the scaled platforms the economics are mostly non-US: Teads earns 58.7% of revenue in | Line 687 (Taboola revenue by geography); line 725 (Teads EMEA 58.7%, US 27.0%); line 1199 (Greater China exit); matrix lines 611-612. Search only: CyberAgent's advertising guideline page references 'JIAA's Guidelines for Native Advertisements'. | supported with qualification | Add a Limits line naming the Asia-Pacific markets and rule sets not covered. In Ch.9, note that the Dable and popIn scores lack regional context. State that the buyer guidance is US-specific while the vendor economics are global. | Accepted and applied. | Abstract (line 10); Ch.10 heading 'The rules, in three jurisdictions' (line 1148 | | F56 | Ch.10 intro (line 1081) and 'The publisher's decision' (lines 1144-1146): "They are the ones a buyer, an agency and a publisher face." | Final completeness review (P2). The chapter promises choices for publishers and agencies but gives the buyer a decision tool and the publisher one paragraph, sourced mostly from 2017-2018 reporting. There is no current revenue share, RPM or widget share of publisher revenue, and no publisher checklist, even though the paper already holds usable data. Later trade reporting on the retreat of guarantees (2023) is not cited. | Sources [79] (2018), [302] and [365] (2017) carry the publisher section. The paper's own payout data: Taboola TAC 63.5%, Teads 59.3%, Dianomi 72.9%; guarantee cost 15% of TAC; guarantees priced per thousand page views (lines 263-303). Search only: Digiday, March 2023, 'publishers can no longer count on content-recommendation guarantee checks'. | supported with qualification | Add a short publisher checklist built from data already in the paper (payout proxies, guarantee structure and trend, exclusivity terms, the YouGov publisher-credibility penalty, MFA exposure), label what is unmeasured, and either add a note on agency choices or drop 'an agency' from the promise. | Accepted and applied. | Ch.10 intro (line 1081) and 'The publisher's decision' (lines 1144-1146) | | F57 | Appendix 'Definitions' (lines 1919 and 1945); agentic map AdCP row (line 1268): "It does not use the reader's identity or history." | Final completeness review (P2). Two fixes the corrections ledger records were made in the chapters but not carried into the appendix. The contextual definition contradicts Ch.1's corrected text (R15: contextual describes the basis for selection, not an absence of personal data). The AdCP/AAMP definition ('Neither names native as a format. No native vendor implements either') contradicts Ch.11's correction (X34: AdCP docs and catalogue refer to nati | Ch.1 line 75: 'It describes the basis for the choice, not a promise that no personal data is used.' Ch.11 line 1287: 'AdCP can describe native. Its 3.1.24 docs refer to native formats'. data/corrections-ledger.json entries R15 and X34. | supported with qualification | Rewrite the two definitions to match the corrected chapter text. Change the agentic map field to 'standard native format: deferred to 3.2'. | Accepted and applied. | Appendix 'Definitions' (lines 1919 and 1945); agentic map AdCP row (line 1268) | | F58 | Ch. 9 'In short' and 'A quotient, and why not a sum' item 3; fig-nq caption; data/capability-matrix.json nq.bands: "A thin record then neither helps nor hurts." | Final scoring review (P0). The Strong band is mostly an artefact of how the index is built. Strong starts at 57, which is the field NQ (56) plus 1. With 4 pseudo-observations and k=4 scored cells, shrinkage keeps half of the raw score. So any row whose few scored cells are 3s lands in Strong, and 3 is the anchor for 'real but thinly evidenced, narrow or in beta'. Six of the 12 'Strong' vendors have only four scored cells, all grade C: AppLovin | data/capability-matrix.json: rows[].coverage/nq/band; nq.field_nq=56; nq.bands Strong=57. Recomputed: a vendor with all 12 cells at 3 gets round(100*((12*3+4*2.805)/16)/5)=59, and a vendor with 4 cells at 3 gets 58. Both are Strong. | contradicted | Withhold the band for rows below a coverage floor (for example fewer than 6 scored cells, or no Depth group) and label them 'thin record'. Report band counts only for rows above that floor. Either rename Strong/Mid as 'above/below field mean' or raise the Strong threshold so that an all-3 profile does not qualify. Restate the ch. 9 'In short' counts and the fig-nq caption to match. | Accepted and applied. | Ch. 9 'In short' and 'A quotient, and why not a sum' item 3; fig-nq caption; dat | | F59 | Ch. 9 'In short' and callout 'The honest limit of this number'; data/capability-matrix.json nq.sensitivity: "The bands are the finding. The order is not." | Final scoring review (P1). The reweighting test never checks band membership. It only records whether the front rank keeps Taboola (front_kept, plus max and mean position shift). Recomputing from rows[].nq_alt, 6 of 27 vendors change band under the paper's own four weightings. TripleLift drops to 56 (Mid) under Standing-led. Dianomi drops to 56 and 55 (Mid) under Depth-led and Standing-led. Readpeak drops to 56 (Mid) under Depth-led. Meta Audi | data/capability-matrix.json nq.sensitivity (fields max_shift, mean_shift, front_kept, front_size only); rows[].nq_alt; nq.bands 68/57/48. Recomputation reproduces every published NQ exactly. | supported with qualification | Add a per-vendor band-stability column (bands held under 5 of 5 weightings). State that only Taboola's front rank and Teads/MGID/Nativo's Strong are robust to the weights. Qualify 'The bands are the finding' accordingly. | Accepted and applied. | Ch. 9 'In short' and callout 'The honest limit of this number'; data/capability- | | F60 | Abstract 'In short'; Ch. 9 'In short'; Ch. 9 'What survives every weighting': "Measurement is the weakest capability across the field." | Final scoring review (P1). The matrix's own numbers do not support this, and the reason given in ch. 9 conflicts with a scored cell. By mean score, measurement (2.67, n=9) is the fourth-lowest dimension. Commerce (2.25, n=12), optimisation (2.45, n=11) and publisher controls (2.64, n=14) are all lower, and the same holds within the nine profiled vendors (2.25 / 2.44 / 2.44). What is true is that measurement has the lowest ceiling: no cell scor | Per-dimension means computed from data/capability-matrix.json; data/scores.json taboola.measurement_support.why (cites F-S33); evidence/research/F.json F-F23 passage; WebFetch of https://www.measured.com/integration/taboola/. | supported with qualification | Replace with: 'Measurement has the lowest ceiling: no vendor scores above 3. No vendor documents holdout tooling of its own. The one independent-partner route found, Measured's page for Taboola, gives no design, so it cannot lift the cell above the marketing cap.' Change the abstract line to match. | Accepted and applied. | Abstract 'In short'; Ch. 9 'In short'; Ch. 9 'What survives every weighting' | | F61 | Ch. 9 'The wider universe' and 'The large platforms' native products...'; Ch. 12 'Sampling and scoring, stated plainly';: "Second, at least four of the twelve dimensions had evidence." | Final scoring review (P1). Two scored rows fail the paper's own entry test. Amazon Publisher Services Native has 3 scored cells, and JWX has 3, all grade D from one homepage (U-S29). Ch. 9 concedes this in passing ('three or four scored dimensions'). Kargo, which ch. 3 names among the private native players, was left unscored on similar one-homepage evidence. The universe data also disagrees with the matrix. 36 universe rows carry scored:true | data/capability-matrix.json amazon-aps-native.coverage=3 and jwx.coverage=3; data/scoring-rules.json selection_rules.scored_set (c); data/vendor-universe.json scored flags (36 true; the 'Connatix / JWP Connatix / JWX' row false); scripts/check.mjs line 48. | contradicted | Either drop Amazon APS and JWX (leaving 25 scored vendors), or change the rule and text to three dimensions and re-examine Kargo. Reconcile the vendor-universe scored flags with the matrix. Add a coverage assertion to check.mjs. | Accepted and applied. | Ch. 9 'The wider universe' and 'The large platforms' native products...'; Ch. 12 | | F62 | Ch. 9 'How to read the matrix' and 'The exchanges hand the outcome to the buyer'; exchange rows vs. large-platform rows: "If the dimension does not fit the business, the cell says n/a. One case is conversion data at a pure exchange." | Final scoring review (P1). One fact, that the outcome capability sits in another buying tool, is coded three different ways. (1) TripleLift and Sharethrough conversion data is coded n/e, not n/a as the text says. (2) Their optimisation is scored 1/D ('Outcome bidding sits with the buyer's DSP'), which pulls their Depth down. (3) For Google AdSense ('Optimisation sits in Google Ads'), Meta Audience Network ('Meta Ads optimisation'), AppLovin MA | data/capability-matrix.json: triplelift/sharethrough optimisation (v=1, grade D) and conversion_data_access (k=ne); google-adsense-native, meta-audience-network, applovin-max-native and amazon-aps-native demand-side cells (k=na); amazon-aps-native.commerce_capabilities.why; data/scoring-rules.json cell_rules.na. | supported with qualification | Adopt one rule. Either code 'capability sits with a third-party buying tool' as n/a for the exchanges too, or score (or n/e) the large platforms' demand-side dimensions from the product the unit is bought through. Do not credit commerce from demand-side while coding the other demand-side cells n/a. Correct the ch. 9 example, or recode exchange conversion data as n/a. | Accepted and applied. | Ch. 9 'How to read the matrix' and 'The exchanges hand the outcome to the buyer' | | F63 | Ch. 9 matrix, Teads Holding Co. row, Formats = 4: "Vendor marketing can earn at most a 3, and vendor documentation at most a 4." | Final scoring review (P1). Teads format_support is scored 4 with evidence grade D. Its own rationale says the formats are 'described, but on product pages rather than in specifications or filings'. It is the only cell in the matrix above its grade cap. Neither the corrections ledger nor scripts/check.mjs catches it. | data/scores.json teads.format_support {v:4, grade:'D'}; data/scores.json _rules ('marketing-only evidence caps at 3'); scripts/check.mjs line 48 has no grade-cap assertion. | supported with qualification | Score 3 (Teads NQ 64→63, still second), or cite a specification-grade source. Add cap assertions to check.mjs (grade D ≤3, grade C ≤4). | Accepted and applied. | Ch. 9 matrix, Teads Holding Co. row, Formats = 4 | | F64 | Ch. 9 matrix, Teads row, Conv. data = 3 (rationale in data/scores.json): "a server-side endpoint is mentioned but not documented; no conversion-value or CRM identity matching is described" | Final scoring review (P1). Both the Teads verifier and the paper's own ch. 8 contradict this rationale. The verifier found a documented S2S conversion postback URL keyed on ob_click_id (V-teads-S06). Ch. 8 states that the postback takes optional 'Order value, currency and order ID' fields and that 'Teads also publishes a server-side conversion API template for Google Tag Manager'. With a pixel, an S2S postback carrying value, CSV offline impor | evidence/verify/teads.json TEADS-F21 correction and additional_sources V-teads-S06; chapters/08-commerce.md line 43 (V-H-S01); data/scores.json mgid.conversion_data_access. | contradicted | Rewrite the rationale. Score 4 for parity with MGID (Teads NQ 64→65), or state what difference justifies 3. | Accepted and applied. | Ch. 9 matrix, Teads row, Conv. data = 3 (rationale in data/scores.json) | | F65 | Ch. 9 matrix, Optimise / Workflow / Creative columns (Dianomi, Nativo, Revcontent, Teads, Readpeak, MGID, Taboola): "The same standard applies to each vendor." | Final scoring review (P1). Comparable evidence gets different scores in the Depth group. Optimisation: Dianomi is 2/A ('maximise engagement, clicks, views'; no conversion objective), Nativo is 2/D (engagement-based audience expansion; no conversion objective) and Revcontent is 1/C (CPC and vCPM only). All three meet the same anchor, '1 = click or impression objectives only'. The Dianomi verifier's own source shows delivery ranking is a functio | data/scores.json optimisation for dianomi, nativo and revcontent; advertiser_workflow for dianomi and revcontent; creative_tools for teads, readpeak, mgid, taboola and nativo; evidence/verify/dianomi.json V-dianomi-S12; evidence/verify/nativo.json V-nativo-S04; data/scoring-rules.json optimisation and creative_tools anchors. NQ recomputed. | supported with qualification | Optimisation rescored for Dianomi and Nativo; workflow and creative alignment left for the author's scoring review. | Accepted and partly applied. | Ch. 9 matrix, Optimise / Workflow / Creative columns (Dianomi, Nativo, Revconten | | F66 | Ch. 9 matrix, Pub ctrl: Dianomi 1 (unconfirmed), Readpeak 2 (unconfirmed); scoring-rules publisher_controls anchor; Ch. : "If the record gives no basis to score, the cell says n/e. That is not a zero." | Final scoring review (P1). Missing evidence is scored low for two profiled vendors but coded n/e for thin rows. Dianomi's 1 and Readpeak's 2 rest on publisher-side controls being 'not documented'. The publisher_controls anchor itself defines '1 = undocumented', which contradicts the global n/e rule, while Microsoft's pubCenter access error is coded n/e. Dianomi's rationale also lists central advertiser vetting and real-time publisher reporting | data/scores.json publisher_controls for dianomi, readpeak and taboola; teads.commerce_capabilities; dianomi.creative_tools; data/scoring-rules.json dimensions[publisher_controls].anchors, cell_rules.ne and cell_rules.availability; chapters/12-method.md line 88. NQ recomputed. | contradicted | Delete '1 = undocumented' from the anchor. Recode not-documented cells as n/e, or score Dianomi on the same basis as Taboola. Do not score cells labelled 'unconfirmed'. Correct the ch. 12 sentence. | Accepted and applied. | Ch. 9 matrix, Pub ctrl: Dianomi 1 (unconfirmed), Readpeak 2 (unconfirmed); scori | | F67 | Ch. 9 matrix, Supply and Interop columns (MGID, TripleLift, Sharethrough, Nativo, Readpeak): "A large help centre earns grade C, not a higher score." | Final scoring review (P1). Reach cells are graded and scored unevenly on similar evidence. Supply: TripleLift (1,268 distinct publisher domains) and Sharethrough (1,932) get 4/C from their sellers.json files (X20 says 'sellers.json files graded C'). MGID's sellers.json lists 2,522 domains and names MSN and Xiaomi, yet MGID supply is 3/D. Nativo supply is 4/A from a Life360 investor deck that names no publishers (anchor 3), while Nativo's own S | data/scores.json supply_access for mgid, triplelift, sharethrough and nativo; interoperability for nativo and readpeak; data/corrections-ledger.json X20; evidence/swot/nativo.json notes; evidence/verify/nativo.json NATIVO-F19 correction and overall_reliability; data/scoring-rules.json interoperability anchors. | supported with qualification | MGID, Nativo and Readpeak cells regraded or rescored; other Reach cells unchanged. | Accepted and partly applied. | Ch. 9 matrix, Supply and Interop columns (MGID, TripleLift, Sharethrough, Nativo | | F68 | Ch. 9 'Deep profiles and SWOT'; Ch. 12 'What to distrust in this paper'; evidence/swot/*.json basis labels: "Strengths and weaknesses rest on the record." | Final scoring review (P1). Vendor help-centre documentation is labelled 'record' for Taboola, Revcontent, Readpeak, TripleLift and Sharethrough, but 'vendor_claim' for MGID, Teads and Nativo. There are direct pairs: Taboola S3 (help-centre conversion bidding, record) vs Teads S4 (help-page conversion bidding rules, vendor_claim), and Taboola W3 vs Teads W4 (both help-page site-control and reporting lines). All four MGID strengths are vendor_cl | evidence/swot/taboola.json strengths[2], weaknesses[2], notes item 4; evidence/swot/teads.json strengths[3], weaknesses[3]; evidence/swot/mgid.json strengths[0-3]; evidence/swot/readpeak.json notes ('labelled record as official product documents'); ch. 9 definitions of Record and Vendor claim. | supported with qualification | Adopt one rule, either a third label ('vendor documentation') or vendor_claim everywhere for help-centre evidence. Relabel all nine SWOTs and rewrite the ch. 12 sentence. | Accepted and applied. | Ch. 9 'Deep profiles and SWOT'; Ch. 12 'What to distrust in this paper'; evidenc | | F69 | Ch. 12 'What the process is not'; Abstract 'How it was done' and 'Limits': "Scoring was done once. It was a single AI-assisted pass under the author's rules, after the verifier records were read." | Final scoring review (P1). The paper's own ledger records at least two scoring passes, the first made before some verifier records were read. X20 says 'Profiles landed after the first scoring pass'. X21 says 'The first pass had not read the help centre' and records four Dianomi scores raised. X22 records Taboola interoperability raised from 3 to 4. Ch. 9 itself says Dianomi's 'first scores were low'. The abstract's claim that the rules 'were w | data/corrections-ledger.json X20, X21, X22; data/stats.json corrections_scores=8; data/scoring-rules.json published_before_scoring; `git status` in the nofluffadvisory repo shows research/native-2026/ untracked. | supported with qualification | Text corrected to describe a first pass and a revision; no independent timestamp exists for the rules file, and the text says so. | Accepted and partly applied. | Ch. 12 'What the process is not'; Abstract 'How it was done' and 'Limits' | | F70 | Ch. 9 'What survives every weighting' (Standing paragraph); matrix Revcontent row, Measure = 1 and Pub ctrl = 2: "Revcontent has the lowest Standing of the profiled vendors. It names no third-party verifier." | Final scoring review (P2). The verifier found three things missing from Revcontent's cells. Revcontent has been TAG Registered since May 2018, and its current sellers.json carries the TAG-ID. A c.2018 vendor post claims TAG Certified Against Fraud/Malware. Its Publisher Agreement offers a Revenue/RPM Guarantee. None of these appears in the measurement or publisher-controls rationale; X24 fixed only the sellers.json count. MGID's measurement 3 | evidence/verify/revcontent.json REVCONTENT-F19 (verdict contradicted), omissions 2-3, V-revcontent-S05 and V-revcontent-S06; data/scores.json revcontent.measurement_support, mgid.measurement_support, taboola.publisher_controls. | supported with qualification | TAG registration added to the rationale; the revenue guarantee was not added. | Accepted and partly applied. | Ch. 9 'What survives every weighting' (Standing paragraph); matrix Revcontent ro | | F71 | TripleLift SWOT weakness 2; matrix TripleLift row, Transp. = 2 (grade B): "Adalytics found its ads on sampled made-for-advertising sites in January 2024 despite a stated block." | Final scoring review (P2). The SWOT line and the cell single out TripleLift from sources that named many firms and hedged their findings. The ANA passage lists five SSPs that 'declined participation or could not provide data in the required timeframe': FreeWheel, Google AdX, PubMatic, TripleLift and Yahoo. The TripleLift profile records that Adalytics found 'the same pattern across Google, Magnite, OpenX, PubMatic, Index Exchange and Criteo'. | evidence/profiles/triplelift.json TRIPLELIFT-F26 passage and TRIPLELIFT-F27 counterevidence; evidence/verify/triplelift.json F27 correction and omissions; data/source-aliases.json TRIPLELIFT-S43→TABOOLA-S30; evidence/profiles/taboola.json TABOOLA-F30. | supported with qualification | Add the peer firms and the 'potentially' hedge to W2 and the cell rationale. Then apply the report to Taboola and Teads the same way, or to no vendor. | Accepted and applied. | TripleLift SWOT weakness 2; matrix TripleLift row, Transp. = 2 (grade B) | | F72 | MGID SWOT (threat 3; weaknesses); Teads measurement cell; Abstract disclosure paragraph: "since MGID is absent from AdCP's member logos as of September 2026." | Final scoring review (P2). The MGID SWOT leaves out independent evidence the verifier found. On the negative side: Aos Fatos 2024 (misleading MGID-served ads on Veja) and a peer-reviewed WWW 2023 paper naming MGID as an organiser of conflict clickbait campaigns. On the positive side: TAG Brand Safety certification after audit in 2022, with MGID on TAG's 2026 list. Taboola's SWOT carries comparable adverse crawls (W4). Threat 3 treats AdCP memb | evidence/verify/mgid.json MGID-F32, MGID-F22, omissions 1-2; evidence/swot/mgid.json threats[2] and notes item 5; evidence/profiles/teads.json capabilities.measurement_support.observed (lists Samba TV); chapters/00-abstract.md disclosure paragraph. | supported with qualification | Add one MGID weakness citing the independent misleading-ad findings, and credit the TAG Brand Safety seal in its measurement cell. Replace threat 3 or disclose the author's AdCP role in the line. Add to the disclosure that Samba TV is listed as a Teads measurement partner. | Accepted and applied. | MGID SWOT (threat 3; weaknesses); Teads measurement cell; Abstract disclosure pa | | F73 | Ch. 9 matrix, Geo / Formats / Interop cells for large-platform and thin rows: "Global." | Final scoring review (P2). The same evidence earns different scores in the thin rows. Geo: 'Global.' earns 3 for Meta Audience Network and AppLovin, 'Global product' earns 3 for both Google rows, and Microsoft's 'markets not listed on the page opened' earns 3. For Media.net, EX.CO and Primis, 'Described as global; no detail' earns 2. The rules say absence of geographic evidence is n/e. Format anchors are misapplied: NewsBreak (seven documented | data/capability-matrix.json geographic_reach cells for meta-audience-network, applovin-max-native, google-adsense-native, google-ad-manager-native, microsoft-native, medianet, exco and primis; format_support cells for newsbreak, stackadapt, ezoic, primis, amazon-aps-native and jwx; stackadapt.interoperability; data/scoring-rules.json anchors. | supported with qualification | Geography and format cells made consistent; StackAdapt interoperability left as scored. | Accepted and partly applied. | Ch. 9 matrix, Geo / Formats / Interop cells for large-platform and thin rows | | F74 | Dianomi SWOT weakness 4; Nativo SWOT strength 1: "conflicts with 39.1bn audited impressions for FY2025" | Final scoring review (P2). Two SWOT lines go beyond their sources. The 39.1bn figure is an operating KPI in Dianomi's annual report. BDO audits the financial statements, and nothing in the cited passage says the impression count is audited. Nativo S1 ('so the unit's scale appears in SEC filings') conflicts with Nativo W3 ('Standalone Nativo revenue appears in no Life360 filing reviewed'). Life360's advertising line also includes its pre-acquis | evidence/swot/dianomi.json weaknesses[3]; evidence/profiles/dianomi.json DIANOMI-F26 passage ('Annual report KPI: Impressions (millions) 39,129'); evidence/swot/nativo.json strengths[0] and weaknesses[2]; evidence/verify/nativo.json NATIVO-F13 correction. | supported with qualification | Dianomi: '39.1bn impressions reported in the FY2025 annual report'. Nativo S1: 'Life360 now reports advertising as its own line ($22.0M in Q2 2026, including Life360's own in-app ads); Nativo is not reported separately.' | Accepted and applied. | Dianomi SWOT weakness 4; Nativo SWOT strength 1 | | F75 | Abstract, finding 2; Ch4 In short and 'The price of exclusivity'; fig-guarantee-share title and caption; Ch11 Scenario t: "Its cost of guarantees to publishers peaked at about 18% of that cost in 2024" | Final numerical and citation review (P0). The paper's own data cannot support 'peaked in 2024'. The FY2023 value was never captured. The caption and prose also leave out the FY2020 value (13%) that the chart itself plots. They measure the rise from the 2021 low ('double the 2021 share') and then tell a 'guarantees faded, then came back' story that the 2020 figure undercuts. | data/figure-data.json guarantee_share: FY2020 pct 13, FY2021 9, FY2022 10, FY2023 pct null with src [], FY2024 18, FY2025 15. The sidecar figures/fig-guarantee-share.json scope reads '(2023 not captured in the opened filings)'. evidence/research/C.json open_questions lists the guarantee cost for FY2023 as not opened. Claim C-F04 in the ledger includes 13% in 2020. I ran an EDGAR full-text search (via WebFetch) for 'c | supported with qualification | Take the FY2023 percentage from the FY2023 10-K MD&A and redraw the series. Until then, drop 'peaked' and 'double the 2021 share' from the abstract, Ch4, the figure title and Ch11. Suggested wording: 'about 18% of TAC in 2024 and 15% in 2025, up from 9–10% in 2021–22 (13% in 2020; 2023 not captured)'. | Accepted and applied. | Abstract, finding 2; Ch4 In short and 'The price of exclusivity'; fig-guarantee- | | F76 | Ch9 deep profiles (rendered from data/profiles.json): Dianomi 'Poor fit', 'Limits' and the Conversion-data and Interoper: "Advertisers needing documented conversion-event integration, CPA/ROAS bidding, incrementality measurement or DSP-native programmatic buying at scale." | Final numerical and citation review (P0). The profile bodies were not regenerated after the verifier and scoring corrections. On the same page they now contradict the SWOT and the matrix. Dianomi's buyer guidance and capability table say it documents no conversion pixel or API, no named verification vendors, no named SSPs and no advertiser exclusion tools, yet the matrix scores its conversion data at 4. The MGID profile keeps '508 sellers', 'Google LLC is th | data/profiles.json dianomi.capabilities.conversion_data_access = {availability: 'unconfirmed', grade: 'E', observed: 'No documentation of a conversion pixel...'}; data/scores.json dianomi.conversion_data_access v=4, grade C ('help centre documents an image conversion pixel, server-to-server events... AppsFlyer and Adjust'). evidence/verify/dianomi.json, DIANOMI-F25 and the capabilities check, both contradicted: the v | contradicted | Rebuild data/profiles.json from the corrected verifier records. For Dianomi: rewrite Poor fit bullets 1 and 3, Limits bullets 3–5, and the Conversion-data, Interoperability and Supply-transparency rows. For MGID: replace 508 and 'sole intermediary' with 2,571 sellers across 2,522 domains and six intermediaries, and replace 'unchanged since 2018' with 'moved between 850M and over 1B'. For Taboola: add the MCP server t | Accepted and applied. | Ch9 deep profiles (rendered from data/profiles.json): Dianomi 'Poor fit', 'Limit | | F77 | Ch9 In short and callout 'The honest limit of this number'; fig-nq title; Ch12 'Sampling and scoring': "The bands are the finding. The order is not." | Final numerical and citation review (P1). The sensitivity test checks only that the front rank holds (front_kept). It never tests whether vendors keep their Strong, Mid or Narrow band. They often do not. Because Strong starts one point above the field NQ, a vendor with four cells scored 3 lands in 'Strong' by default. As a result, 6 of the 12 Strong vendors have only 4 of 12 dimensions scored, and they outrank full-record vendors such as Sharethrough (52) an | I reproduced every NQ in data/capability-matrix.json exactly, using field mean 2.805 and bands 68/57/48. Using rows[].nq_alt, 6 of 27 vendors change band under at least one of the four weightings: Dianomi (Strong to Mid under Depth-led 56 and Standing-led 55), Readpeak (Strong to Mid under Depth-led 56), TripleLift (Strong to Mid under Standing-led 56), Revcontent (Mid to Narrow under Standing-led 47), Meta Audience | supported with qualification | Report band stability for each vendor and name the vendors that change band. Do not band vendors with 5 or fewer scored dimensions; label them 'record too thin to band'. Alternatively, set the Strong threshold clearly above the field mean. Soften 'bands are the finding' to 'only the front rank is robust'. | Accepted and applied. | Ch9 In short and callout 'The honest limit of this number'; fig-nq title; Ch12 ' | | F78 | Abstract In short; Ch9 In short: "Measurement is the weakest capability across the field." | Final numerical and citation review (P1). This is true only as a ceiling: measurement is the only dimension on which no vendor scores above 3. By mean score it is not the weakest. Commerce, optimisation and publisher controls all average lower, both across all 27 vendors and among the nine profiled ones. | From data/capability-matrix.json, counting scored cells only, mean scores are: measurement 2.67 (n=9, max 3), commerce 2.25 (n=12), optimisation 2.45 (n=11), publisher controls 2.64 (n=14). Among the nine profiled vendors: optimisation 2.44, publisher controls 2.44, commerce 2.25, measurement 2.67. Every other dimension reaches at least 4. | supported with qualification | Change to 'Measurement is the only dimension on which no vendor scores above 3, because none documents independent holdout or incrementality tooling.' Apply the change in both the abstract and Ch9. | Accepted and applied. | Abstract In short; Ch9 In short | | F79 | Ch4 closing paragraph; Ch10 In short; Ch10 'What a buyer can inspect': "made-for-advertising spend fell from 15% in 2023 to a median of 0.4% in the third quarter of 2025" | Final numerical and citation review (P1). This compares two measures that cannot be compared: a 2023 average from a one-off study of 21 advertisers against a Q3 2025 median from the opt-in benchmark. The ANA itself warns against this, and Ch10 criticises the playbook for doing exactly this. On a like-for-like basis the median fell from 10% to 0.39%. The follow-on sentence ('rose to 1.1% ... as AI slop sites appeared') compares a Q1 2026 panel of 86 participa | Ledger G-F07 final_wording: 'MFA share of spend fell from 15% (2023 study)... The median fell from 10% (2023) to ... 0.39% (Q3 2025). The ANA says waterfall, average and median values are not comparable.' The G-F11 verifier note: Q1 2026 had 86 participating and 66 active marketers, and 'a change in participant mix is a stronger alternative explanation for the uptick'. I-F20: 39 participating marketers in Q3 2025. | contradicted | Use one measure throughout: 'the median fell from 10% in 2023 to 0.39% in Q3 2025 among participants'. Report the Q1 2026 1.1% as a cost-waterfall value from a larger, different panel, with no causal 'as'. Correct the participant counts to 21 (2023 study), 39 (Q3 2025) and 86 (Q1 2026). | Accepted and applied. | Ch4 closing paragraph; Ch10 In short; Ch10 'What a buyer can inspect' | | F80 | Ch2, fig-spend-vintages (the 2019 bar in the 2019-03 vintage): "social 95.6%" | Final numerical and citation review (P1). The figure labels the 2019 bar 'social 95.6%', as though social feeds were 95.6% of native. The 95.6% is actually native's share of social display: the inversion Ch2 itself calls out in a trade report ('the same two numbers the wrong way round'). The 2019 vintage's own split is 77% in 2018 falling to 74% in 2020. | figures/fig-spend-vintages.svg, text element at x≈353.6. scripts/make-figures.mjs line 222 takes the first percentage in social_share. data/figure-data.json native_spend_vintages for the 2019 row: social_share 'native = 95.6% of social display; native = 30.8% of nonsocial display'. Ledger A-F12: social share 77% in 2018 and 74% in 2020. | contradicted | Relabel the bar 'social ~75% (77%→74%, 2018–20)' or leave it blank. Stop the label regex from reading 'native = X% of social' as a social share. | Accepted and applied. | Ch2, fig-spend-vintages (the 2019 bar in the 2019-03 vintage) | | F81 | Ch12 In short and 'How it was built'; fig-pipeline; published claim ledger (Ch12 instrument and data/claim-ledger.json): "So a verifier re-checked 311 of 642 ledger claims. Claims outside that sample are marked unresolved in the ledger." | Final numerical and citation review (P1). The ledger build did not join 23 verifier checks because their IDs are compound or annotated (for example 'MGID-F07 (top finding 2, ...)' or 'DIANOMI-F04/F05/F06/F07'). Ten of those checks were verdicts of 'contradicted'. As a result, several claims the verifier refuted are published as 'unresolved — Not selected for the independent verification sample', still in their original wording, and the counts disagree: 311 c | evidence/verify/*.json: 23 checked[].finding_id values match no ledger id, including MGID-F06, F07, F13 and F32, DIANOMI-F25 and REVCONTENT-F33, all contradicted. In data/claim-ledger.json those ids show review_outcome 'unresolved' with the note 'Not selected for the independent verification sample'. data/stats.json: verifier_checked 311 and claims_checked 288; verifier_contradicted 28 and claims_contradicted 18. D-F | supported with qualification | Normalise the verifier finding_ids (split compound ids and strip annotations), rebuild the ledger, and write each contradicted claim's corrected wording into final_wording. Then quote a single consistent count in Ch12 and in the pipeline figure. Change the wording to 'up to seventeen claims per file'. | Accepted and applied. | Ch12 In short and 'How it was built'; fig-pipeline; published claim ledger (Ch12 | | F82 | Ch12 fig-pipeline, 'Reviewed' box: "8 scores changed" | Final numerical and citation review (P2). The label counts corrections-ledger entries of kind 'scoring', not scores that changed. Five of those eight entries changed no score, and one added two matrix rows. The actual changes were five cells across two entries. | scripts/stats.mjs sets corrections_scores = corrections.filter(kind==='scoring').length. In the corrections ledger, X17 says 'scores unchanged', X19 'no dimension score changed', X23 'scores unchanged', X24 and X26 corrected rationales only, and X20 added rows. X21 raised four Dianomi scores and X22 raised one Taboola score. | supported with qualification | Relabel the box '8 scoring corrections (5 cell scores changed)', or count the cells whose score changed. | Accepted and applied. | Ch12 fig-pipeline, 'Reviewed' box | | F83 | Ch2 'What the industry's own accounts show'; fig-iab-formats caption: "Video's was 18.6%" | Final numerical and citation review (P2). The prose and the caption disagree with the canonical derived figure: (78.0/39.5)^0.25 − 1 = 18.54%. | data/derived-checks.json iab_video_cagr_2021_2025 value 18.5; data/figure-data.json iab_cagr.video 18.5. My recomputation gives 18.54%. | supported with qualification | Change to 18.5% in Ch2 and in the fig-iab-formats caption. | Accepted and applied. | Ch2 'What the industry's own accounts show'; fig-iab-formats caption | | F84 | Ch12 'Sampling and scoring'; Ch9 'The wider universe': "Each could also be evidenced on at least four of the twelve dimensions." | Final numerical and citation review (P2). Two scored vendors break the inclusion rule that was fixed in advance. JWX has 3 dimensions scored and 9 n/e. Amazon APS Native has 3 scored (plus 5 n/a). Ch9 itself says Amazon has 'three or four scored dimensions'. | In data/capability-matrix.json, jwx has 3 score, 0 na and 9 ne; amazon-aps-native has 3 score, 5 na and 4 ne. data/scoring-rules.json inclusion rule (c) requires 'at least four of the twelve dimensions can be evidenced'. | supported with qualification | Drop JWX, or state that n/a counts as evidenced and justify its inclusion. Correct the Ch12 sentence to match. | Accepted and applied. | Ch12 'Sampling and scoring'; Ch9 'The wider universe' | | F85 | Ch11 retirements table (Nativo row); fig-consolidation-timeline; Ch13 glossary ('AdCP and AAMP', 'Contextual targeting'): "Neither names native as a format." | Final numerical and citation review (P2). Several logged corrections reached the chapter text but not the surfaces around it. The Nativo completion still reads 2026-01-05 in the Ch11 table and in the timeline figure, while Chs 3 and 10 give 2 January (X15). The glossary says AdCP does not name native, which Ch11 and J-F04 refute (X34). The glossary says contextual targeting 'does not use the reader's identity or history', which Ch1 corrected (R15). The abstr | Corrections ledger X15, X34, R15 and R31; ledger J-F04 (contradicted) and J-F21 final_wording; data/figure-data.json timeline entry '2026-01-05 close Life360, Inc. / Nativo'; ledger B-F27 '(closed 2026-01-05)'. | supported with qualification | Use 2 January 2026 in the table, the timeline and B-F27. Glossary: 'AdCP docs reference native formats; a canonical native format is deferred to 3.2' and 'contextual describes the selection basis, not an absence of personal data'. Use 'two large US-listed' in the abstract and figure title. Use 'sponsored follow-up questions' in Ch8 In short. | Accepted and applied. | Ch11 retirements table (Nativo row); fig-consolidation-timeline; Ch13 glossary ( | | F86 | Ch3 In short and the Yahoo paragraph; Ch3 buying routes and timeline vs Ch11 'Agents and protocols': "A thirty-year exclusive needed three amendments in its third year." | Final numerical and citation review (P2). There are three date and arithmetic slips. (1) The deal closed on 17 January 2023, so March 2026 (Amendment 7) falls in the fourth year, not the third. Ch3's 'at least three' amendments also sits oddly beside Ch11's indicator of 'seven'. (2) 'Ten days before this paper's cutoff': 18 to 27 September is nine days. (3) OpenRTB Native 1.2 is dated March 2017 in Ch3, the timeline and B-F23, but July 2017 in Ch11 and J-F26 | Ledger B-F02 gives amendment effective dates of 2025-06-01, 2025-08-11 and 2026-03-01. Ledger B-F20 dates the offer 2026-09-18. The B-F23 verifier note has the change log '1.2 March 2017', while J-F26 quotes the index 'Release Date: July 2017'. | supported with qualification | Change to 'three amendments in nine months (June 2025–March 2026), seven in all' and to 'Nine days before the cutoff'. Use the spec's own date (March 2017) throughout, noting that the Tech Lab index lists July 2017. | Accepted and applied. | Ch3 In short and the Yahoo paragraph; Ch3 buying routes and timeline vs Ch11 'Ag | | F87 | Ch7 In short: "In 2026, some platforms shortened view-through attribution windows. That cuts attributed conversions." | Final numerical and citation review (P2). This overstates the evidence, which the body correctly narrows. Only one platform is documented (Meta), and the change was a deprecation of 7-day and 28-day view windows in the Ads Insights reporting API. The sources opened do not show that campaign attribution settings changed. | Ledger F-F13 verifier correction: 'Ads Insights API (reporting) effective 12 January 2026, not a general removal; neither opened source says Ads Manager campaign attribution settings changed'. The Ch7 body says the same. | supported with qualification | Change to 'In January 2026 Meta dropped 7- and 28-day view windows from its reporting API; shorter reported windows cut attributed conversions, not what the ads caused.' | Accepted and applied. | Ch7 In short | | F88 | Ch10 In short and 'Who buys, and through what': "A significant majority of Taboola's revenue comes from about 2,200 advertisers who work with it directly" | Final numerical and citation review (P2). This merges two separate 10-K statements. The roughly 2,200 Scaled Advertisers work with Taboola 'directly, or through advertising agencies'. The 'significant majority of revenue' comes from those Scaled Advertisers that work with it directly. The body sentence 'These deal with it direct, not through an agency' repeats the error. The verifier qualified this claim. | Ledger I-F15 review_note quotes the 10-K: 'approximately 2,200 Scaled Advertisers working with us directly, or through advertising agencies' and 'a significant majority of our Revenues came from Scaled Advertisers working with us directly, rather than via an agency'. | contradicted | Change to 'A significant majority of Taboola's revenue comes from Scaled Advertisers (about 2,200, spending over $100K a year) that buy directly rather than through an agency.' | Accepted and applied. | Ch10 In short and 'Who buys, and through what' | | F89 | Ch12 'The challenge passes' and 'What could not be reached': "Five reviews were then run on the final text. Their results are in the corrections ledger below" | Final numerical and citation review (P2). The corrections ledger contains no entry from any of the five final-text reviews. Its 137 entries are X01–X37 (verification, scoring, process) and R01–R100 (the interim-draft review). The access disclosure also understates what is unknown. It says '90 were not read in full', but access is 'not recorded' for a further 137 sources that the verifiers added. | data/corrections-ledger.json kinds: review 100, verification 23, scoring 8, research-lead 2, author-review 2, process 2; none cites a fact, argument, completeness or numerical review of the final text. In data/source-register.json, access_limits is 'not recorded (added by verifier)' for 137 sources. | supported with qualification | Log the outcomes of this and the other final-text reviews (including 'no change' results) before release, or reword the sentence. Change the access sentence to '90 recorded as not read in full; access not recorded for a further 137 verifier-added sources'. | Accepted and applied. | Ch12 'The challenge passes' and 'What could not be reached' | | F90 | Ch12 claim-ledger instrument (the chapter column in the rendered table and in data/claim-ledger.json): "\| D-F18 \| ... \| 7 \| independent measurement \| contradicted \|" | Final numerical and citation review (P2). The published ledger's chapter column points readers to the wrong chapter for most of the claims the text cites. It appears to carry the research stream's chapter_hint from the old numbering, which is the drift R16 asked to be rechecked. | Of the 255 claims cited in chapters/*.md, 175 have a ledger chapter value that is not a chapter citing them. Examples: A-F05 through A-F17 are marked '3' but used in Ch2; C-F01 is marked '6' but used in Ch4; D-F18 is marked '7' but used in Ch5; G-F11 is marked '8' but used in Chs 4 and 10. | contradicted | Derive the chapter column from where each claim ID actually appears in the chapters, and list several chapters where a claim is used more than once. | Accepted and applied. | Ch12 claim-ledger instrument (the chapter column in the rendered table and in da | | F91 | data/figure-data.json comparators (published download behind fig-comparators): ""fy2024": 3, "yoy": "Q4: 23"" | Final numerical and citation review (P2). The Amazon advertising row's FY2024 value in the published dataset is a parser artefact ($3 million), not the FY2024 figure. The chart does not draw it, but the file ships to readers as data. | data/figure-data.json comparators, Amazon.com row: fy2025 68635 and fy2024 3. Ledger A-F17 gives the quarterly values; the FY2024 total is not captured. | contradicted | Set fy2024 to null (or to the sum of the four FY2024 quarters, if opened) and set yoy to 'Q4 2025: +23%'. | Accepted and applied. | data/figure-data.json comparators (published download behind fig-comparators) | | R01 | Opening, L3; Chapters 5, 7 and 10: Three sweeping propositions are presented as settled: most readers do not recognise native; its incremental value has never been tested publicly; growth now comes from non-native formats. | Review of the interim draft, finding 1 (P0, overstatement). Required: Limit recognition findings to the studied formats and samples. A public practitioner account describes a Taboola geo test, although it is not independently reproducible and does not isolate recommendation widgets. Add DeeperDive, a directly relevant expansion into publisher AI discovery. Rewrite the opening a | confirmed: states a Taboola geo test in Feb 2024 fed into an MMM; no advertiser, format, design or results given. confirmed: generative AI answer engine on publisher sites; ads inserted in the AI results page; monetisation engine opened to other AI companies; tens of millions of answers a month, 7M+ users (vendor figures). [172, 51] | supported with qualification | Rewrite subtitle and thesis: recognition findings limited to studied formats and samples; 'no public study with design and results' instead of 'never tested'; growth diversification rather than 'comes from non-native'; add DeeperDive. | Accepted after checking the cited source where one could be reached, or on the text. | Rewrite subtitle and thesis | | R02 | Chapter 6, L371; E-F25/E-F26: Automated conversion bidding under CPC is described as transferring conversion risk to the platform. | Review of the interim draft, finding 2 (P0, error). Required: Automation transfers bid-setting authority. If the advertiser pays per click, it still pays for clicks that do not convert. Financial risk transfers only under an applicable guarantee or different contractual charging arrangement. MGID explicitly says target CPA is not guaranteed. | confirmed: 'a directional goal rather than a guaranteed cost per conversion'; 'You continue to pay per click'. | contradicted | Ch6 and delivery figure: automation transfers bid authority, not financial risk; advertiser still pays per click; MGID calls target CPA a directional goal. | Accepted after checking the cited source where one could be reached, or on the text. | Ch6 and delivery figure | | R03 | Chapter 7, L457; Chapter 9 measurement table and conclusions: Native platforms are said not to disclose conversion windows or definitions. | Review of the interim draft, finding 3 (P0, error). Required: Taboola documents conversion configuration and click/view attribution windows. Separate those documented settings from missing lag distributions, independent outcomes and log-level access. Rewrite the measurement table rather than marking all fields unavailable. | confirmed: click-through window 1-30 days, default 30; view-through 1-24 hours, default 24; URL- and event-based conversions with values. [181] | contradicted | Ch7/Ch10 and decision tool: Taboola documents click 1-30 days (default 30) and view 1-24h (default 24h); Revcontent 30-day pixel; Nativo configurable; separate documented windows from missing lag distributions and log access. | Accepted after checking the cited source where one could be reached, or on the text. | Ch7/Ch10 and decision tool | | R04 | Chapter 9, L598; K-F05: Lord & Taylor is described as the only FTC case framed around native advertising. | Review of the interim draft, finding 4 (P0, error). Required: Add the FTC's Creaxion and Inside Publications proceedings, with final orders in February 2019. Remove the enforcement-trend inference built on the purported absence of later cases. | confirmed: final consent orders against Creaxion and Inside Publications; Inside Publications published paid ads disguised as articles; complaint announced Nov 2018; vote 5-0. | contradicted | Ch10 rules section: add Creaxion and Inside Publications final orders (Feb 2019); drop the enforcement-trend inference. | Accepted after checking the cited source where one could be reached, or on the text. | Ch10 rules section | | R05 | Chapter 6, L390; E-F22: The 0.76% versus 0.65% GenAI CTR comparison is presented as controlled within advertiser/campaign/day; conversion outcomes supposedly were not reported. | Review of the interim draft, finding 5 (P0, error / omission). Required: The release distinguishes the raw CTR comparison from tighter controlled comparisons and discusses downstream conversion performance. Identify raw versus adjusted results, obtain the academic paper, and avoid claiming a causal 16.9% advantage from raw rates. The paper has now been identified, but its full tex | confirmed: raw 0.76% vs 0.65%; 'performed comparably' under tightest controls; qualitative line on downstream conversions. opened: 4,633 ads in 1,186 quasi-experiments; no significant effect of AI images on conversion rate for purchase-objective ads; authors say they cannot isolate conversion effects. Sample totals differ from the SSRN abstract (369M impressions, 2.5M clicks) and the press release (500M+, 3M): version reconciliation required. SSRN page itself returned 403. [160] | contradicted | Raw-vs-controlled already fixed; now add the paper (no significant conversion-rate effect; version samples differ) and state that the release's conversion line is qualitative. | Accepted in part; the change records what was and was not adopted. | | | R06 | Chapters 8 and 10; AI/product map: Several shipped or documented products are marked unconfirmed, missing or given inconsistent dates. | Review of the interim draft, finding 6 (P0, errors / omissions). Required: Correct MGID creative generation, Copilot ads, Amazon prompts/Rufus interactions, Realize+ chronology and Teads conversational-ad API. Add Taboola DeeperDive. Distinguish documented capability, public beta, general availability, announcement and independently demonstrated performance.– | confirmed: text-to-image, image-to-image, title/description generation in 29 languages, performance prediction. confirmed: Sponsored Products/Brands prompts, open beta Nov 2025, US GA 25 Mar 2026; prompts may open a dialog in Rufus. [154] | contradicted | Update agentic map and Ch6/Ch8/Ch11: MGID generative tools (2024), Copilot ads live with vendor metrics, Amazon prompts GA US Mar 2026 opening Rufus, Realize+ Apr 2026 chronology, Teads conversational SDK beta with MCP forthcoming, DeeperDive. | Accepted after checking the cited source where one could be reached, or on the text. | Update agentic map and Ch6/Ch8/Ch11 | | R07 | Chapter 3, L169; C-F09: The $204,000 average spend and 84% revenue share appear under annual 2025 wording. | Review of the interim draft, finding 7 (P1, period error). Required: These refer to Q4 2025. The definition of a Scaled Advertiser uses spending over four quarters; the reported average and revenue share use the quarter. State both periods explicitly. | Judged on the text of the interim draft. | contradicted | Ch3: Scaled Advertiser count, $204K average and 84% share refer to Q4 2025; qualification uses trailing four quarters. | Accepted after checking the cited source where one could be reached, or on the text. | Ch3 | | R08 | Chapter 2, L70, L125, L129: $1.24 billion ex-TAC is described as simply the amount remaining after publisher payments; competing platforms supposedly report net. | Review of the interim draft, finding 8 (P1, accounting error). Required: Combined reported ex-TAC measures total $1.2432 billion; combined revenue minus TAC is $1.2268 billion. The difference includes Taboola's $16.4 million amortisation adjustment. Google generally reports relevant advertising revenue gross and partner payments as costs. Label each metric separately. | Judged on the text of the interim draft. | contradicted | Ch2 and comparators figure: combined ex-TAC $1,243.2M vs revenue less TAC $1,226.8M; Google reports network revenue gross with TAC in cost of revenues; drop 'platforms report net'. | Accepted after checking the cited source where one could be reached, or on the text. | Ch2 and comparators figure | | R09 | Chapter 4 summary; Chapter 10 indicator, L755: $637.5 million issued notes and $605.1 million carrying value are treated as current interchangeable debt amounts. | Review of the interim draft, finding 9 (P1, date/measure error). Required: Identify issue principal, outstanding face value and balance-sheet carrying value separately. The $605.1 million long-term carrying amount belongs to December 2025; June 2026 long-term debt was $607.386 million, plus $7.081 million short-term debt. | confirmed: LT debt $607.386M + ST $7.081M at 30 Jun 2026; LT $605.113M + ST $17.595M at 31 Dec 2025; cash $88.0M; equity $7.4M. | contradicted | Ch4 and Ch11 indicator: issued principal $637.5M; $628.2M principal outstanding at end 2025; total debt carrying value $614.5M at 30 Jun 2026. | Accepted after checking the cited source where one could be reached, or on the text. | Ch4 and Ch11 indicator | | R10 | Chapter 9, L539; I-F28: 120 opportunities × 20% is said to produce 23 deals; the stated deal values do not produce the revenue range. | Review of the interim draft, finding 10 (P1, arithmetic error). Required: 120 × 20% = 24 deals. At $144,000–230,000 ACV, 24 deals imply $3.456–5.520 million annual contract value. At 23 deals, the range is $3.312–5.290 million. Explain any separate ramp, proration or revenue-recognition assumption. | Judged on the text of the interim draft. | contradicted | Ch10 model box: 120 x 20% = 24; flag to author that the playbook itself says 23. | Accepted after checking the cited source where one could be reached, or on the text. | Ch10 model box | | R11 | L14, common definition: FTC's resemblance-based definition, IAB's format taxonomy and OpenRTB's asset specification are not equivalent. | Review of the interim draft, finding 11 (P1, overstatement). An editorially produced sponsored article need not be assembled from ad assets. Use: “Native describes advertising designed to fit its surrounding content or experience; programmatic native commonly achieves this through publisher-rendered assets.” | Judged on the text of the interim draft. | supported with qualification | Ch1 definition sentence replaced with the proposed wording. | Accepted after checking the cited source where one could be reached, or on the text. | | | R12 | Six-axis framework: A campaign can use multiple surfaces, formats, targeting methods and buying routes. | Review of the interim draft, finding 12 (P1, conceptual). Apply the axes to a placement or line item, or allow multiple values. Do not require a campaign to have exactly one value on every axis. | Judged on the text of the interim draft. | supported with qualification | Ch1 and category map: axes apply to a placement or line item; a campaign can carry several values. | Accepted after checking the cited source where one could be reached, or on the text. | Ch1 and category map | | R13 | L44, emerging surfaces: Agent-mediated buying is a buying method, not itself an advertising surface. | Review of the interim draft, finding 13 (P1, scope). Put agents on the buying-route axis; keep AI search and conversational interfaces on the surface axis. | Judged on the text of the interim draft. | supported with qualification | Ch1 and scope figure: agent-mediated buying moves to the buying-route axis. | Accepted after checking the cited source where one could be reached, or on the text. | Ch1 and scope figure | | R14 | L46, affiliate exclusions: An editorially chosen affiliate link can still involve a material commercial connection. | Review of the interim draft, finding 14 (P1, overstatement). Distinguish excluded from this paper's spend perimeter from not advertising or outside disclosure rules. Reconcile this paragraph with Chapter 8's affiliate-disclosure discussion. | Judged on the text of the interim draft. | supported with qualification | Ch1: affiliate links excluded from the spend perimeter, not from disclosure rules. | Accepted after checking the cited source where one could be reached, or on the text. | Ch1 | | R15 | Contextual versus native: Contextual describes the basis for selection, not a guarantee that a system uses no personal data. | Review of the interim draft, finding 15 (P2, precision). Keep format, placement, selection signals and data use separate. The native/contextual distinction is useful once this qualification is added. | Judged on the text of the interim draft. | supported with qualification | Ch1: contextual describes selection basis, not absence of personal data. | Accepted after checking the cited source where one could be reached, or on the text. | Ch1 | | R16 | L44, L46, L61 and later references: There is no Chapter 11. | Review of the interim draft, finding 16 (P2, navigation). References to that chapter should point to Chapter 10 or be removed. Some references to Chapter 10's rules and measurement decision tool belong to Chapter 9. Recheck every cross-reference after editing. | Judged on the text of the interim draft. | supported with qualification | References were written against final numbering (chapter 9 = instruments); the interim build shifted them. Recheck every cross-reference after the final build. | Accepted in part; the change records what was and was not adopted. | | | R17 | “Only” public native estimate / no published size: A bounded search cannot establish that exactly one public estimate exists, or that no estimate exists anywhere. | Review of the interim draft, finding 17 (P1, overstatement). Say “the only explicit native-display estimate in the source set reviewed” and describe excluded market-research estimates and why they are not comparable. | Judged on the text of the interim draft. | supported with qualification | Ch2: 'only explicit native-display estimate in the source set reviewed'. | Accepted after checking the cited source where one could be reached, or on the text. | Ch2 | | R18 | A-F09/A-F12; historical social shares: The 74–84% range describes historical vintages, not a measured 2026 split. | Review of the interim draft, finding 18 (P1, historical scope). Do not use it to estimate today's open-web remainder. Non-social native can also include in-app and other inventory outside the paper's core scope. | Judged on the text of the interim draft. | supported with qualification | Ch2: historical social shares are vintage-specific, not a 2026 split; non-social includes in-app. | Accepted after checking the cited source where one could be reached, or on the text. | Ch2 | | R19 | L70; fig-iab-formats: “Slowest-growing format” needs qualification. | Review of the interim draft, finding 19 (P1, error). Display's 9.8% is below the other major named formats, but the report's “other” category grows about 6.8% from rounded totals. Say “the slowest-growing of the four major named formats.” | Judged on the text of the interim draft. | contradicted | Ch2 and IAB figure: 'slowest of the four major named formats'; 'other' grew about 6.9%. | Accepted after checking the cited source where one could be reached, or on the text. | Ch2 and IAB figure | | R20 | L135 / interpretation of IAB categories: The paper correctly says social crosses formats, then treats the accounts as excluding social when discussing native's container. | Review of the interim draft, finding 20 (P1, internal contradiction). Display growth cannot isolate recommendation-widget growth or prove core native contraction. | Judged on the text of the interim draft. | supported with qualification | Ch2: display growth cannot isolate recommendation widgets. | Accepted after checking the cited source where one could be reached, or on the text. | Ch2 | | R21 | L129, 1.1% comparison: Global company revenue divided by US advertising revenue is not a market share. | Review of the interim draft, finding 21 (P1, metric mismatch). Remove the 1.1% ratio from the market-size argument. A caveat does not make mismatched numerator and denominator analytically informative. | Judged on the text of the interim draft. | supported with qualification | Ch2: remove the 1.1% global/US ratio. | Accepted after checking the cited source where one could be reached, or on the text. | Ch2 | | R22 | L138, market size and capturable spend: Two companies' revenues are neither a complete native-market census nor native-only revenue. | Review of the interim draft, finding 22 (P1, scope). Annual revenue guidance is a company forecast, not an estimate of theoretically capturable spend. Rename these as company scale benchmarks and management revenue guidance. | Judged on the text of the interim draft. | supported with qualification | Ch2 callout: 'company scale benchmarks' and 'management revenue guidance'. | Accepted after checking the cited source where one could be reached, or on the text. | Ch2 callout | | R23 | Inclusion matrix, including L108: “Not reported” and “excluded” are different. | Review of the interim draft, finding 23 (P1, classification error). Paid search is listed as excluded while also described as mentioned in dentsu's scope. Use separate statuses: included, excluded, unclear and not separately reported. | Judged on the text of the interim draft. | contradicted | Inclusion matrix: four statuses; fix parser that coded 'not mentioned' as N (dentsu paid search). | Accepted after checking the cited source where one could be reached, or on the text. | Inclusion matrix | | R24 | Meta growth decomposition: A 12% impression increase and 9% price increase imply 22.08% combined growth, not 21%. | Review of the interim draft, finding 24 (P2, reproducibility). Any 57%/43% attribution of growth requires an explicit method for allocating the 1.08 percentage-point interaction. It is not a unique causal decomposition. | Judged on the text of the interim draft. | supported with qualification | Ch2: drop the 57/43 split; state 12% volume and 9% price compound to about 22%. | Accepted after checking the cited source where one could be reached, or on the text. | Ch2 | | R25 | 2026 EMARKETER estimate: Retain $147.98 billion and 13.1% as the cited forecast, dated to its release/vintage. | Review of the interim draft, finding 25 (P2, currentness). Do not silently turn a forecast available early in 2026 into a September outturn or newly refreshed forecast. | Judged on the text of the interim draft. | supported with qualification | Ch2: label $147.98B as the December 2025 vintage forecast. | Accepted after checking the cited source where one could be reached, or on the text. | Ch2 | | R26 | L171, “can lose any advertiser / cannot lose either”: Revenue concentration establishes relative exposure, not binary survivability. | Review of the interim draft, finding 26 (P1, overstatement). Replace with “Loss of a major supply partner could have a much larger revenue effect than loss of any single disclosed advertiser.” Account for substitution, margins, contract terms and transition periods. | Judged on the text of the interim draft. | supported with qualification | Ch3: relative-exposure wording. | Accepted after checking the cited source where one could be reached, or on the text. | Ch3 | | R27 | L173, partner count decline: Growth in revenue alongside a lower partner count does not prove that lost partners were small. | Review of the interim draft, finding 27 (P1, invalid inference). Pricing, mix, growth at retained partners and definition changes could offset larger losses. State the two observations; mark the cause of the count change unknown. | Judged on the text of the interim draft. | supported with qualification | Ch3: partner-count change cause unknown. | Accepted after checking the cited source where one could be reached, or on the text. | Ch3 | | R28 | L173, all growth is non-native: Realize's display expansion supports diversification, but does not allocate revenue growth by format. | Review of the interim draft, finding 28 (P1, overstatement). Add DeeperDive and distinguish product launches from measured growth contributions. No disclosed format bridge means the contribution is unknown. | confirmed: generative AI answer engine on publisher sites; ads inserted in the AI results page; monetisation engine opened to other AI companies; tens of millions of answers a month, 7M+ users (vendor figures). [51] | supported with qualification | Ch3: no format bridge disclosed; add DeeperDive. | Accepted after checking the cited source where one could be reached, or on the text. | Ch3 | | R29 | Dianomi offer / distressed-multiple inference: £19 million upfront equity value divided by £27.4 million revenue is about 0.69×. | Review of the interim draft, finding 29 (P1, valuation interpretation). It is not enterprise value/revenue without cash/debt adjustments, nor proof of distress. Separate upfront consideration, up to £8 million contingent value, premium, balance sheet and closing conditions. The offer was not completed at the cutoff. | Judged on the text of the interim draft. | supported with qualification | Ch3 callout: upfront equity value about 0.69x revenue, not EV/revenue, not proof of distress; contingent value and conditions stated. | Accepted after checking the cited source where one could be reached, or on the text. | Ch3 callout | | R30 | Dianomi publisher-count discussion: An exact historical count of 341 and a later statement of “over 250” do not establish a decline. | Review of the interim draft, finding 30 (P1, non-comparable counts). The latter is a lower bound. Request like-for-like dated counts or remove the claimed change. | Judged on the text of the interim draft. | supported with qualification | Ch3: counts not like-for-like; drop claimed decline. | Accepted after checking the cited source where one could be reached, or on the text. | Ch3 | | R31 | “Two listed platforms”; buying-route heading: Say “the two large US-listed platforms” where appropriate: Dianomi is also listed in the paper. | Review of the interim draft, finding 31 (P2, scope). “Nobody says how much is bought” is too absolute; Dianomi discloses a programmatic revenue figure. A market-wide route split remains a separate unresolved question. | Judged on the text of the interim draft. | supported with qualification | Ch3: 'two large US-listed platforms'; Dianomi discloses programmatic revenue. | Accepted after checking the cited source where one could be reached, or on the text. | Ch3 | | R32 | Chapter opening; fig-platform-economics: TAC is an accounting expense, not uniformly a same-period cash payment. | Review of the interim draft, finding 32 (P1, accounting terminology). Taboola's TAC includes non-cash amortisation. Replace “go straight back out” with “are recorded as traffic acquisition costs” and provide the reconciliation. | Judged on the text of the interim draft. | supported with qualification | Ch4 and figure: 'recorded as traffic acquisition cost', non-cash amortisation noted. | Accepted after checking the cited source where one could be reached, or on the text. | Ch4 and figure | | R33 | Q2 2026 add-back; C-F03 and Chapter 10: The recomputed 4.7% is mathematically sound when the $12.169 million write-off is retained in the metric. | Review of the interim draft, finding 33 (P1, interpretation). It is not automatically organic or underlying growth: the write-off also affected the reported accounting result. Label this a sensitivity to the adjustment convention, alongside reported 11.8% growth. | Judged on the text of the interim draft. | supported with qualification | Ch4/Ch11: 4.7% labelled an adjustment-convention sensitivity beside reported 11.8%. | Accepted after checking the cited source where one could be reached, or on the text. | Ch4/Ch11 | | R34 | L210, L238–240; C-F04: 9% in 2021 to 15% in 2025 is about 1.67×. | Review of the interim draft, finding 34 (P1, trend overstatement). The 18% peak in 2024 was twice 2021; the share subsequently fell to 15% and approximately 13% in Q2 2026. “Doubled” should identify the peak year, not imply continuous growth. | Judged on the text of the interim draft. | supported with qualification | Ch4 and guarantee figure: peak 2024 doubled 2021; 2025 about 1.67x; later decline stated. | Accepted after checking the cited source where one could be reached, or on the text. | Ch4 and guarantee figure | | R35 | Guarantee discussion: A temporal increase after the Yahoo deal does not establish the deal caused the entire increase. | Review of the interim draft, finding 35 (P1, causal/contract scope). A guaranteed rate per thousand impressions also does not necessarily guarantee total annual income. Explain whether the contract guarantees a rate, minimum payment or volume. | Judged on the text of the interim draft. | supported with qualification | Ch4: no causal attribution to Yahoo; guarantee is a rate per thousand page views, not total income. | Accepted after checking the cited source where one could be reached, or on the text. | Ch4 | | R36 | “A merger that has not yet paid for itself”: Goodwill impairment and losses establish financial strain, not a completed investment-return calculation. | Review of the interim draft, finding 36 (P1, unsupported conclusion). Replace with “Post-acquisition losses, impairment and financing pressure.” A payback conclusion needs acquisition cash flows, financing, integration costs and an explicit horizon. | Judged on the text of the interim draft. | supported with qualification | Ch4 heading: 'Post-acquisition losses, impairment and financing pressure'. | Accepted after checking the cited source where one could be reached, or on the text. | Ch4 heading | | R37 | Debt and liquidity commentary: Positive operating cash flow alone cannot clear liquidity risk; adjusted EBITDA is not cash available for interest. | Review of the interim draft, finding 37 (P1, financial inference). Consider cash, working capital, debt service, capital spending and covenants without asserting that a liquidity event is either inevitable or ruled out. | confirmed: LT debt $607.386M + ST $7.081M at 30 Jun 2026; LT $605.113M + ST $17.595M at 31 Dec 2025; cash $88.0M; equity $7.4M. | supported with qualification | Ch4: liquidity sentence reworded without ruling in or out. | Accepted after checking the cited source where one could be reached, or on the text. | Ch4 | | R38 | Dianomi Apple/open-web CPC: The stated 22.5p versus 110.5p values imply a 4.91× ratio. | Review of the interim draft, finding 38 (P1, attribution of difference). This is descriptive, not evidence that placement alone causes the difference. Advertiser, audience, geography, creative and buying mix can differ. The release's inconsistent currency typography should be reconciled against the annual report before publication. | Judged on the text of the interim draft. | supported with qualification | Ch4: RPC ratio descriptive, not placement-caused. | Accepted after checking the cited source where one could be reached, or on the text. | Ch4 | | R39 | Dianomi 2025 impression decline: The rounded Apple/non-Apple figures in the results release imply a decline near 15%, while the narrative uses roughly 14%. | Review of the interim draft, finding 39 (P1, open reconciliation). Reconcile the precise series, coverage and denominator. Do not silently “correct” one series to the other without the underlying table. | Judged on the text of the interim draft. | supported with qualification | Open item: Dianomi components sum to 44.5bn vs 45.5bn total; keep company's -14.1% and note. | Accepted after checking the cited source where one could be reached, or on the text. | Open item | | R40 | Made-for-advertising series: Retain the manuscript's warning that ANA participants are self-selected. | Review of the interim draft, finding 40 (P2, evidence labelling). Do not splice a market-wide estimate, participant median and high-spend subgroup into one population trend. Definitions, denominator and cohort should appear beside every figure. | Judged on the text of the interim draft. | supported with qualification | Keep cohort warnings beside every MFA figure. | Accepted after checking the cited source where one could be reached, or on the text. | | | R41 | Opening and recognition summary: Historical recognition experiments do not establish the percentage of all 2026 native readers who identify advertising. | Review of the interim draft, finding 41 (P0, generalisation). The evidence covers different units, tasks and populations. Use “Several studies found low recognition under particular native formats and disclosure conditions.” | Judged on the text of the interim draft. | supported with qualification | Ch5 opening: 'several studies found low recognition under particular formats and disclosure conditions'. | Accepted after checking the cited source where one could be reached, or on the text. | Ch5 opening | | R42 | Recognition range and figure: Do not select only the 7–37% low-recognition results while presenting a category-wide summary. | Review of the interim draft, finding 42 (P1, denominator/scope). Other conditions in the cited evidence are substantially higher. Hyman's 21–72% per-item range includes a Facebook stimulus outside this paper's core perimeter. FTC results also combine search/native conditions. Display study and stimulus strata explicitly. | Judged on the text of the interim draft. | supported with qualification | Recognition figure and text: stratify by task, stimulus and surface; note Facebook stimulus and FTC mixed conditions. | Accepted after checking the cited source where one could be reached, or on the text. | Recognition figure and text | | R43 | “Better labels never reach parity” and mechanism claims: “Never” exceeds the experiments. | Review of the interim draft, finding 43 (P1, overstatement). Recognition, clicks, attention, trust and persuasion are distinct outcomes. The assertion that native's click advantage and recognition failure are the same mechanism requires direct causal mediation evidence; resemblance alone does not prove it. | Judged on the text of the interim draft. | supported with qualification | Ch5: 'did not reach parity in the studies reviewed'; mechanism framed as hypothesis. | Accepted after checking the cited source where one could be reached, or on the text. | Ch5 | | R44 | Eisend meta-analysis; D-F13 and related discussion: A small, statistically non-significant average behavioural-intention association is not proof of no effect on actual sales. | Review of the interim draft, finding 44 (P1, statistical interpretation). Distinguish attitudes, intentions and observed behaviour; report uncertainty and heterogeneity. The counts of studies/effects cited in the manuscript are supported. | Judged on the text of the interim draft. | supported with qualification | Ch5: non-significant intention association is not evidence of no sales effect. | Accepted after checking the cited source where one could be reached, or on the text. | Ch5 | | R45 | Comparisons of recognisers with non-recognisers: Recognition can be affected by treatment and participant characteristics. | Review of the interim draft, finding 45 (P1, causal inference). An outcome difference between recognisers and non-recognisers is not automatically the causal effect of recognising the ad. Use randomised disclosure contrasts for causal claims, with the appropriate design limitations. | Judged on the text of the interim draft. | supported with qualification | Ch5: recogniser vs non-recogniser differences are associational. | Accepted after checking the cited source where one could be reached, or on the text. | Ch5 | | R46 | L324; D-F20: The 2020 problematic-ad findings are a dated coded sample, not a current fraud rate or a census of all native ads. | Review of the interim draft, finding 46 (P1, time/scope). Preserve the codebook's meaning of “problematic.” The source also warrants a denominator check because its narrative and display-ad table do not line up transparently. Do not independently label every coded category fraudulent. | Judged on the text of the interim draft. | supported with qualification | Ch5: dated coded sample, codebook meaning preserved. | Accepted after checking the cited source where one could be reached, or on the text. | Ch5 | | R47 | L330 onward, inference that most units fail FTC rules: FTC guidance does not set a universal percentage-recognition pass mark. | Review of the interim draft, finding 47 (P1, legal inference). Old study recognition rates cannot determine the legality of most present-day units. Evaluate particular disclosures, placement and overall impression; separate empirical concern from legal conclusion. | Judged on the text of the interim draft. | supported with qualification | Ch5: remove 'most native units would not pass'; FTC sets no numeric pass mark. | Accepted after checking the cited source where one could be reached, or on the text. | Ch5 | | R48 | Claims that recognition research is absent/current evidence exhausted: Add the 2025 disclosure/detection study and the newer CARE-model work to the research screen. | Review of the interim draft, finding 48 (P1, missing literature). A 2026 university-listed conference contribution also addresses recognition in GenAI overviews. These are research leads, not full-text-cleared findings here, and do not establish a representative 2026 recognition rate. | Judged on the text of the interim draft. [106, 107] | supported with qualification | 2022/2024 studies already added after verifier D; add the 2025 and 2026 GenAI-overview recognition leads as unscreened. | Accepted in part; the change records what was and was not adopted. | | | R49 | L320 and L365, vendor work “not usable”: Vendor funding is a conflict and affects evidential weight; it does not automatically make an experiment unusable. | Review of the interim draft, finding 49 (P2, evidence hierarchy). Distinguish independently replicated work, transparent internal experiments, survey comparisons and unsupported marketing claims. Assess design before deciding the strength of an inference. | Judged on the text of the interim draft. | supported with qualification | Ch5: vendor work weighted by design, not dismissed. | Accepted after checking the cited source where one could be reached, or on the text. | Ch5 | | R50 | L373; E-F26: The 50-conversion rule is for setting a target CPA in the cited Realize documentation, not a ban on all automated optimisation below that threshold. | Review of the interim draft, finding 50 (P0, eligibility error). Separate hard eligibility rules, recommendations and learning guidance. MGID's 10–30 conversions over a learning period also does not imply five to ten conversions every day. | confirmed: 'a directional goal rather than a guaranteed cost per conversion'; 'You continue to pay per click'. confirmed: 50 conversions in 7 days only for setting a target CPA. | contradicted | Ch6: 50 conversions only for target CPA; MGID 10-30 over the learning period, not per day. | Accepted after checking the cited source where one could be reached, or on the text. | Ch6 | | R51 | Low-volume/mid-market discussion: “Most mid-market advertisers cannot meet the threshold” needs advertiser-volume data, not inference from the threshold alone. | Review of the interim draft, finding 51 (P1, unsupported market claim). Present an illustrative budget/conversion scenario instead of a market prevalence assertion. | Judged on the text of the interim draft. | supported with qualification | Ch6: illustrative scenario instead of 'most mid-market advertisers'. | Accepted after checking the cited source where one could be reached, or on the text. | Ch6 | | R52 | Generative creative subsection and product map: MGID's documented image-generation tools predate the cutoff. | Review of the interim draft, finding 52 (P1, product omission). Its CTR Guard also concerns creative generation/testing. Treat pre-launch scoring, creative generation and bid optimisation as separate capabilities. | confirmed: text-to-image, image-to-image, title/description generation in 29 languages, performance prediction. [154] | supported with qualification | Ch6: MGID generative tools since Feb 2024; scoring, generation and bidding separated. | Accepted after checking the cited source where one could be reached, or on the text. | Ch6 | | R53 | E-F22, study “not located”: The academic work is identifiable as “AI in Disguise—Quasi-Experimental Analysis of a Large-Scale Deployment of AI-Generated Ads.” Link it and record the version examined. | Review of the interim draft, finding 53 (P1, research traceability). The press release and later paper metadata use different sample totals, which require a version reconciliation rather than a single blended sample description. | confirmed: raw 0.76% vs 0.65%; 'performed comparably' under tightest controls; qualitative line on downstream conversions. opened: 4,633 ads in 1,186 quasi-experiments; no significant effect of AI images on conversion rate for purchase-objective ads; authors say they cannot isolate conversion effects. Sample totals differ from the SSRN abstract (369M impressions, 2.5M clicks) and the press release (500M+, 3M): version reconciliation required. SSRN page itself returned 403. [160] | supported with qualification | Ch6: paper identified and linked; sample totals differ by version. | Accepted after checking the cited source where one could be reached, or on the text. | Ch6 | | R54 | L394, all heavy modelling before request: Offline training does not imply that runtime is only a lookup. | Review of the interim draft, finding 54 (P1, architecture overreach). Candidate retrieval, feature computation and model inference can occur at serving time. State only the stages and latency measurements actually described by each dated technical source. | Judged on the text of the interim draft. | supported with qualification | Ch6: drop 'request is a lookup'; state only dated stage descriptions. | Accepted after checking the cited source where one could be reached, or on the text. | Ch6 | | R55 | L384, auction rules: A fixed-bid help page, exchange transport field and historical secondary description do not establish one auction rule across all inventory. | Review of the interim draft, finding 55 (P2, auction scope). Record mechanism by product, route and date. A disclosure gap is a valid finding; a universal mechanism claim is not. | Judged on the text of the interim draft. | supported with qualification | Ch6: auction rule recorded per product and source; disclosure gap retained. | Accepted after checking the cited source where one could be reached, or on the text. | Ch6 | | R56 | L403–431 and L455: A public practitioner account describes a February 2024 Taboola geo experiment used to calibrate MMM. | Review of the interim draft, finding 56 (P0, missed counterevidence). It is not a reproducible, format-specific RCT, but it prevents a blanket “never tested in public” conclusion. Report the account and its missing design details. | confirmed: states a Taboola geo test in Feb 2024 fed into an MMM; no advertiser, format, design or results given. [172] | supported with qualification | Ch7: report the Violet Growth geo-test account and its missing detail; 'no public study with design and results'. | Accepted after checking the cited source where one could be reached, or on the text. | Ch7 | | R57 | L406, L420; fig-rct-vs-observational: From the displayed medians, DML/RCT ratios are 2.86×, 3.22× and 4.80×; propensity-score matching/RCT ratios are 5.97×, 9.78× and 12.80×. | Review of the interim draft, finding 57 (P1, arithmetic / statistical wording). “Three to ten” describes neither full set. Ratios of medians are not medians of experiment-level ratios. | Judged on the text of the interim draft. | supported with qualification | Ch7 and figure: 2.9-4.8x (DML) and 6.0-12.8x (matching), ratios of medians. | Accepted after checking the cited source where one could be reached, or on the text. | Ch7 and figure | | R58 | Application of Facebook results to native: The Facebook experiments demonstrate a serious identification problem; they do not estimate a correction factor for Taboola, Teads or MGID. | Review of the interim draft, finding 58 (P1, external validity). Remove statements that native observational lift is necessarily inflated by a particular multiple. | Judged on the text of the interim draft. | supported with qualification | Ch7: no native-specific multiplier implied. | Accepted after checking the cited source where one could be reached, or on the text. | Ch7 | | R59 | “Experiment-grade / unbiased” language and decision tool: Randomisation supports causal inference when assignment, exposure, outcome measurement, attrition and interference are handled appropriately. | Review of the interim draft, finding 59 (P1, experiment assumptions). A geo test needs credible comparable areas and power. A label alone does not guarantee an unbiased estimate. | Judged on the text of the interim draft. | supported with qualification | Ch7: randomisation conditional on execution. | Accepted after checking the cited source where one could be reached, or on the text. | Ch7 | | R60 | L435, “retargeting works on average”: The cited retailer study measures return visits in a particular setting. | Review of the interim draft, finding 60 (P1, overgeneralisation). It does not establish positive average incremental sales or profit across all retargeting. State the measured outcome, population and horizon. | Judged on the text of the interim draft. | supported with qualification | Ch7: retargeting result stated as return visits in one setting. | Accepted after checking the cited source where one could be reached, or on the text. | Ch7 | | R61 | L443, Meta windows then Taboola efficiency: Meta's reporting-window change does not show Taboola changed its reporting definition. | Review of the interim draft, finding 61 (P1, invalid causal link). Remove the implied explanation of Taboola's efficiency claim unless a relevant Taboola change is documented. “Shorter windows reduce attributed conversions” is a conditional measurement observation, not evidence of a particular vendor's change. | Judged on the text of the interim draft. | supported with qualification | Ch7: Meta window change not linked to Taboola's claim. | Accepted after checking the cited source where one could be reached, or on the text. | Ch7 | | R62 | L447–449 and measurement conclusions: An MMM using independently measured sales is not mechanically dependent on the platform's attribution window. | Review of the interim draft, finding 62 (P1, mmm precision). Window changes matter when they alter an input/outcome used by that model. Explain the model specification rather than assigning an inherent limitation to all MMM. | Judged on the text of the interim draft. | supported with qualification | Ch7: MMM dependency on windows only when attributed conversions are inputs. | Accepted after checking the cited source where one could be reached, or on the text. | Ch7 | | R63 | Creative split tests, holdouts and controls: Random allocation and correct inference matter more than exactly equal spend. | Review of the interim draft, finding 63 (P1, design advice). Adaptive allocation is not intrinsically invalid, but needs appropriate analysis. A PSA control may itself change behaviour; do not automatically treat it as a no-ad counterfactual. | Judged on the text of the interim draft. | supported with qualification | Decision tool: randomisation over equal spend; PSA control caveat. | Accepted after checking the cited source where one could be reached, or on the text. | Decision tool | | R64 | Ghost-ad efficiency / absence of adoption: Keep the cost-efficiency result tied to the original setting. | Review of the interim draft, finding 64 (P2, source scope). “No implementation located in this review” is defensible; “none exists” is not. Likewise, a new campaign may lack its own conversion history while still benefiting from network-trained models. | Judged on the text of the interim draft. | supported with qualification | Ch7: 'no implementation located in this review'. | Accepted after checking the cited source where one could be reached, or on the text. | Ch7 | | R65 | Skimlinks commercial terms: Public documentation states a typical 75% publisher / 25% Skimlinks commission split. | Review of the interim draft, finding 65 (P1, missed disclosure). Add it, labelled as the documented general model rather than a guarantee for every agreement. “No split disclosed” is incorrect. | support page 403; public FAQ says Skimlinks takes 'a small percentage'; the 75/25 figure appears in search results and the audit but was not opened here. [199] | supported with qualification | Ch8: FAQ opened says 'a small percentage'; 75/25 reported but not opened here; stated as such. | Accepted in part; the change records what was and was not adopted. | Ch8 | | R66 | Criteo retail-media comparison: $263.872 million is retail-media revenue, while the cited approximately 22% share uses group contribution ex-TAC. | Review of the interim draft, finding 66 (P1, mixed metrics). Use $259.684 million / $1,174.617 million ≈ 22.1% for the contribution share; revenue share is $263.872 million / $1,944.901 million ≈ 13.6%. | Judged on the text of the interim draft. | supported with qualification | Ch8: 22.1% is contribution share; revenue share 13.6%. | Accepted after checking the cited source where one could be reached, or on the text. | Ch8 | | R67 | L482–484, Amazon route and AMC: “The one documented route” should be limited to the paper's retrieved examples. | Review of the interim draft, finding 67 (P1, unsupported uniqueness / causal method). Access to a clean room does not make exposed/unexposed comparisons causal. Identify the actual experimental or identification design used through AMC. | Judged on the text of the interim draft. | supported with qualification | Ch8: 'one route documented in the sources retrieved'; clean room enables but does not make analyses causal. | Accepted after checking the cited source where one could be reached, or on the text. | Ch8 | | R68 | L502, Teads/Outbrain ingestion: Legacy Outbrain's 24-hour first-party matching description is not a universal attribution window for all current Teads products. | Review of the interim draft, finding 68 (P1, product lineage). A public Teads server-side conversion/GTM integration exists. Audit legacy pixel, server-to-server and current Teads CAPI separately; do not infer unsupported fields or deduplication guarantees from the integration's existence. | confirmed: server-side GTM template for a Teads conversion API with Consent Mode v2; endpoint and dedup not stated. [224] | supported with qualification | Ch8: separate legacy Outbrain pixel, CSV import, and Teads server-side CAPI template. | Accepted after checking the cited source where one could be reached, or on the text. | Ch8 | | R69 | L504 and buyer conclusions: A documentation search that does not find raw exports does not prove no customer can obtain them or preserve its own click IDs. | Review of the interim draft, finding 69 (P1, export inference). Separate public API documentation, advertiser-owned logs, contractual access and independent auditability. Taboola's documented S2S pathway also has a specific attribution scope; do not generalise it to every possible integration. | Judged on the text of the interim draft. | supported with qualification | Ch8/Ch10: 'no public documentation found' vs contractual access. | Accepted after checking the cited source where one could be reached, or on the text. | Ch8/Ch10 | | R70 | L514; J-F20/J-F21: The initial ads test included eligible Free and Go users, not only Free users. | Review of the interim draft, finding 70 (P1, chatgpt scope). The official description separates ads from organic answers. Use “labelled ads in the conversational interface,” not language implying paid insertion into the answer itself. | Judged on the text of the interim draft. | supported with qualification | Ch8: Free and Go tiers; labelled ads below answers, separate from answers. | Accepted after checking the cited source where one could be reached, or on the text. | Ch8 | | R71 | ChatGPT scale claims: OpenAI's August 2026 statement supports the vendor-reported annualised run-rate and rollout claims. | Review of the interim draft, finding 71 (P1, source and metric). Replace Wikipedia/secondary attribution with the primary announcement. Annualised run rate is not revenue already earned over a full year. Add the September Sponsored Agents pilot separately, without calling it general availability. | Judged on the text of the interim draft. | supported with qualification | Primary OpenAI pages 403; keep attributed reporting, note inaccessible primary; add Sponsored Agents as a reported pilot with no OpenAI documentation at 14 Sep. | Accepted in part; the change records what was and was not adopted. | | | R72 | AI assistant map: Microsoft's Copilot ad documentation and Amazon's prompts announcement contradict “unconfirmed.” Amazon's March 2026 US general availability relates to the described prompts product; some interactions can open Rufus. | Review of the interim draft, finding 72 (P1, missing products). Do not claim that every Rufus answer contains advertising. | partial: ads in Copilot with vendor first-party metrics (73% higher CTR vs search, Feb-May 2025); formats and availability not listed on that page. confirmed: Sponsored Products/Brands prompts, open beta Nov 2025, US GA 25 Mar 2026; prompts may open a dialog in Rufus. [254] | supported with qualification | Agentic map: Copilot ads live (vendor-reported metrics); Amazon prompts GA US opening Rufus; no claim that every Rufus answer carries ads. | Accepted after checking the cited source where one could be reached, or on the text. | Agentic map | | R73 | Chapter 8 and outlook: Add Taboola's June 2026 DeeperDive advertising platform and Teads' documented conversational-ad API beta. | Review of the interim draft, finding 73 (P1, missing native developments). Dianomi's 2026 results also describe its Dappier partnership. These directly affect the account of native distribution moving into AI interfaces. Keep vendor scale and effectiveness claims attributed. | confirmed: generative AI answer engine on publisher sites; ads inserted in the AI results page; monetisation engine opened to other AI companies; tens of millions of answers a month, 7M+ users (vendor figures). confirmed: conversational AI SDK public beta 12 Nov 2025; MCP 'coming soon'. [51, 252] | supported with qualification | Ch8/Ch11: DeeperDive, Teads conversational SDK, Dianomi-Dappier partnership (Mar 2026, conversations ongoing). | Accepted after checking the cited source where one could be reached, or on the text. | Ch8/Ch11 | | R74 | L536, “enterprise business held”: Renewed brand partnerships do not establish a stable enterprise segment P&L. | Review of the interim draft, finding 74 (P1, overstatement). Say the company reported renewals while citing direct-response/SMB weakness; enterprise revenue and profitability are not isolated by that evidence. | confirmed: LT debt $607.386M + ST $7.081M at 30 Jun 2026; LT $605.113M + ST $17.595M at 31 Dec 2025; cash $88.0M; equity $7.4M. | supported with qualification | Ch10: renewals reported alongside DR/SME weakness; segment P&L not isolated. | Accepted after checking the cited source where one could be reached, or on the text. | Ch10 | | R75 | Measurement table, L565 onward: Replace blanket “none available” entries with vendor-specific evidence. | Review of the interim draft, finding 75 (P1, invalid universal conclusion). At minimum Taboola's conversion windows and event configuration are documented. “No public documentation found” should identify the searched product and date; it is not equivalent to “cannot provide.” | confirmed: click-through window 1-30 days, default 30; view-through 1-24 hours, default 24; URL- and event-based conversions with values. [181] | supported with qualification | Ch10 table: vendor-specific documented settings; 'not found in public documentation (date)'. | Accepted after checking the cited source where one could be reached, or on the text. | Ch10 table | | R76 | L598, regulatory table: FTC guidance and endorsement guides interpret enforcement expectations; they should not be described as standalone statutes. | Review of the interim draft, finding 76 (P1, legal precision). Identify Section 5 and distinguish statutes, rules, guidance, codes and industry recommendations. | Judged on the text of the interim draft. | supported with qualification | Rules table: kind labels distinguish statute, rule, guidance, code, practice. | Accepted after checking the cited source where one could be reached, or on the text. | Rules table | | R77 | Fake reviews and AI-generated reviews: The rule addresses fake or false reviews/testimonials and specified related practices; it does not ban every AI-assisted review or every incentive. | Review of the interim draft, finding 77 (P1, legal scope). Describe the falsity, material connection and conditional-incentive issues accurately. | Judged on the text of the interim draft. | supported with qualification | Ch8/Ch10: rule covers fake or false reviews and specified practices. | Accepted after checking the cited source where one could be reached, or on the text. | Ch8/Ch10 | | R78 | L600, California OOPS: Do not imply California opt-out preference signal obligations first arise in 2026. | Review of the interim draft, finding 78 (P1, legal timing/scope). Distinguish the earlier obligation from amendments and implementation dates. A signal and a website opt-out link are not universally interchangeable alternatives. | search summary: OOPS obligations date from regulations finalised 29 Mar 2023; 1 Jan 2026 amendments require displaying whether the signal was processed. To be confirmed from the regulation text during edits. [323] | supported with qualification | Ch10: OOPS obligation since 2023 regulations; 2026 amendments. | Accepted after checking the cited source where one could be reached, or on the text. | Ch10 | | R79 | L602, DSA: Specify the covered service and applicable exemptions. | Review of the interim draft, finding 79 (P1, legal precision). The provisions concern profiling using special-category data and profiling-based ads to minors where the platform knows with reasonable certainty that the recipient is a minor; “all targeting to minors” is too broad. Recheck the statutory text and Article 19 exemptions in the final legal table. | opened this time: Art 19 micro/small exemption; Art 26(3) no ads on profiling with special-category data; Art 28(2) 'aware with reasonable certainty' minors. [332] | supported with qualification | Ch10: DSA scope with Art 19 exemption and 'reasonable certainty' for minors. | Accepted after checking the cited source where one could be reached, or on the text. | Ch10 | | R80 | L602, AI Act: The cited 2026 Omnibus change and delayed high-risk dates are supported by the Commission announcement. | Review of the interim draft, finding 80 (P1, legal precision). Keep those dates, while separating Article 50 transparency obligations from high-risk implementation. Include relevant public-interest text/human editorial review qualifications; do not imply all AI-assisted ad copy must carry one identical label. | Judged on the text of the interim draft. | supported with qualification | Ch10: Art 50 separated from high-risk dates; labelling scope qualified. | Accepted after checking the cited source where one could be reached, or on the text. | Ch10 | | R81 | US state privacy law count: The count depends on the date and whether Florida's narrower law is included. | Review of the interim draft, finding 81 (P2, counting convention). Do not “correct” 19 to 20 without defining the convention. Put the as-of date, included laws and scope beside the count. | Judged on the text of the interim draft. | supported with qualification | Ch10: count convention stated with as-of date. | Accepted after checking the cited source where one could be reached, or on the text. | Ch10 | | R82 | Claimed capability scores / enterprise matrix: The supplied body does not contain the promised scored capability matrix or adequate scoring rubric. | Review of the interim draft, finding 82 (P1, missing declared method). Add actual vendor rows, evidence dates, definitions, weights and treatment of unknowns, or remove promises of scores and deep profiles. | Judged on the text of the interim draft. | structural | Matrix, rubric and profiles exist in data; chapter 9 renders them. | Resolved by completing the chapters, instruments and package that the interim draft lacked. | | | R83 | L645, “cleanest natural experiment”: Two companies' different quarterly outcomes are a descriptive comparison. | Review of the interim draft, finding 83 (P0, method error). There is no identified exogenous assignment or credible counterfactual. Replace “natural experiment” with “a useful contrast in reported quarterly performance.” | Judged on the text of the interim draft. | contradicted | Ch11: 'a useful contrast in reported quarterly performance'. | Accepted after checking the cited source where one could be reached, or on the text. | Ch11 | | R84 | L647–651, revenue/profit guidance divergence: Gross-versus-net presentation can change reported revenue without changing ex-TAC economics. | Review of the interim draft, finding 84 (P1, accounting inference). Do not blend that issue with the new write-off adjustment or underlying demand. Present each possible mechanism separately and say management did not disclose a complete bridge. | Judged on the text of the interim draft. | supported with qualification | Ch11: mechanisms separated; no complete bridge disclosed. | Accepted after checking the cited source where one could be reached, or on the text. | Ch11 | | R85 | L657 and L751; L-F16: The sequence needs quarter labels: Q4 2025 −13%, Q1 2026 −17%, Q2 2026 −22%. | Review of the interim draft, finding 85 (P1, chronology error). “Fell 5% in 2025, then … three following quarters” misleadingly implies a Q3 2026 result, unavailable at the cutoff. Identify the period underlying −5% before retaining it in the series. | Judged on the text of the interim draft. | contradicted | Ch11: FY2025 -5%, Q4 2025 -13%, Q1 2026 -17%, Q2 2026 -22%. | Accepted after checking the cited source where one could be reached, or on the text. | Ch11 | | R86 | L659, “every one … is a page-view series”: Sessions, page views, ad requests and impressions are distinct. | Review of the interim draft, finding 86 (P1, metric error). Sessions can contain multiple pages; pages can produce multiple requests; requests may not fill or render. Describe each measured series and treat the implication for widget inventory as an inference. | Judged on the text of the interim draft. | contradicted | Ch11: sessions, ad requests and impressions named; widget implication an inference. | Accepted after checking the cited source where one could be reached, or on the text. | Ch11 | | R87 | Retirement table, Adyoulike: The table's “owner unconfirmed” conflicts with Chapter 3's documented OpenWeb acquisition. | Review of the interim draft, finding 87 (P1, internal contradiction). Say “acquired by OpenWeb in 2022; current product status not independently established” if no newer evidence is available. An ownership transaction is not itself product retirement. | Judged on the text of the interim draft. | supported with qualification | Ch11 table: Adyoulike acquired by OpenWeb in 2022; current product status not established. | Accepted after checking the cited source where one could be reached, or on the text. | Ch11 table | | R88 | AI/agent map; L726–732: Realize+ is discussed as April/Q2 beta elsewhere but dated September in the map. | Review of the interim draft, finding 88 (P1, dates and categories). The April release also announces Claude Skills. Reconcile dates and distinguish an announced integration from verified production adoption. Teads' AI API documentation calls its MCP integration forthcoming; do not mark that integration live. | confirmed: completed beta, phase 2 expansion in Q2 2026, first Claude Skill (campaign setup and optimisation) released. confirmed: conversational AI SDK public beta 12 Nov 2025; MCP 'coming soon'. [252] | supported with qualification | Agentic map dates reconciled; Teads MCP forthcoming. | Accepted after checking the cited source where one could be reached, or on the text. | | | R89 | L728, “No native vendor appears in any of it”: The paper itself mentions native-related participants such as Kargo. | Review of the interim draft, finding 89 (P1, overstatement). Limit the finding to the specifically searched incumbents and initiatives. Membership, standards compatibility, announced integration and deployed buying activity are different evidence categories. | Judged on the text of the interim draft. | supported with qualification | Ch11: Kargo is an AdCP member per our own J-F02; limit to searched incumbents. | Accepted after checking the cited source where one could be reached, or on the text. | Ch11 | | R90 | L730–734, assistants versus agents: A human approval gate does not rule out autonomous execution within authorised limits. | Review of the interim draft, finding 90 (P1, definition). Define the tested capabilities: planning, tool use, budget changes, placement changes, creative generation and monitoring. Marketing labels alone cannot classify autonomy. | Judged on the text of the interim draft. | supported with qualification | Ch11: define tested capabilities; approval gates do not rule out bounded autonomy. | Accepted after checking the cited source where one could be reached, or on the text. | Ch11 | | R91 | L762, proof absent because “absence has been cheaper than the answer”: This assigns motive without evidence. | Review of the interim draft, finding 91 (P0, unsupported motive). Remove it or explicitly identify it as a hypothesis and give plausible alternatives: private experiments, commercial confidentiality, publication incentives, measurement cost and format taxonomy. Absence of a public paper does not establish a vendor's reason for non-publication. | Judged on the text of the interim draft. | supported with qualification | Ch11: motive sentence removed; alternatives listed as hypotheses. | Accepted after checking the cited source where one could be reached, or on the text. | Ch11 | | R92 | Scenarios and failure modes: Separate observations, hypotheses and forecasts. | Review of the interim draft, finding 92 (P1, forecast discipline). Give each scenario measurable signals, an observation window and a disconfirmation condition. Public practitioner tests and AI-native launches already weaken two categorical premises; update scenarios before treating the next annual report as the sole resolving event. | Judged on the text of the interim draft. | supported with qualification | Ch11 scenarios: observation window and disconfirmation condition each. | Accepted after checking the cited source where one could be reached, or on the text. | Ch11 scenarios | | R93 | All 15 SVGs and 16 listed data downloads are absent from the supplied package.: Supply the actual files before chart and reproducibility sign-off. | Review of the interim draft, finding 93 (P1, open assets). Verify every caption value against its data row. | Judged on the text of the interim draft. | structural | Assets ship in the package. | Resolved by completing the chapters, instruments and package that the interim draft lacked. | | | R94 | The source/claim JSON ledgers and build instructions referenced by the manuscript are absent.: Deliver them or remove the claim that the reader can reproduce the output from the package. | Review of the interim draft, finding 94 (P1, open provenance). A manuscript reference is not the asset itself. | Judged on the text of the interim draft. | structural | Ledgers and build files ship in the package. | Resolved by completing the chapters, instruments and package that the interim draft lacked. | | | R95 | Bibliography access labels include “paywalled excerpt” for public regulatory, SEC and developer material.: Distinguish genuine paywall, access error, extraction failure, abstract-only access and full text. | Review of the interim draft, finding 95 (P1, citation strength). Retrieval failure is not a property of the source's business model. | Judged on the text of the interim draft. | supported with qualification | Register: relabel access limits (paywall vs partial extract vs access error vs abstract-only). | Accepted after checking the cited source where one could be reached, or on the text. | Register | | R96 | Several granular findings rely on abstract/metadata mirrors.: Obtain originals for numerical claims and methods. | Review of the interim draft, finding 96 (P1, claim-source granularity). Record page/table/section, sample, outcome and exact support; do not infer full-study detail from bibliographic metadata. | Judged on the text of the interim draft. | supported with qualification | Claims resting on abstract-only sources flagged in the ledger. | Accepted after checking the cited source where one could be reached, or on the text. | | | R97 | 76 bibliography entries are not numerically cited in the body. Reference IDs 258 and 260 are absent, but no body citation points to them.: Remove unused entries or label them as supplementary reading. | Review of the interim draft, finding 97 (P2, unused references). Renumber at final build; missing numbers are cosmetic, not broken citations in this version. | Judged on the text of the interim draft. | supported with qualification | Reference list limited to sources cited in the rendered paper; full register in data files. | Accepted after checking the cited source where one could be reached, or on the text. | | | R98 | Research questions are referenced by number without a complete question list/search protocol.: Add the question list, scope, databases/domains, query families, date limits, screening rules and treatment of inaccessible sources. | Review of the interim draft, finding 98 (P1, missing research method). This is essential to qualify absence claims. | Judged on the text of the interim draft. | supported with qualification | Method chapter: research questions, search protocol and absence-claim rule. | Accepted after checking the cited source where one could be reached, or on the text. | Method chapter | | R99 | The appendix describes disclosure conventions, but the reader needs an actual author/funding/conflict statement.: State the author's relevant commercial work, clients/relationships where appropriate, funding and standards/working-group participation. | Review of the interim draft, finding 99 (P1, disclosure). Do not imply a disclosure has been made merely by describing the intended practice. | Judged on the text of the interim draft. | structural | Abstract carries the interests-to-declare statement. | Resolved by completing the chapters, instruments and package that the interim draft lacked. | | | R100 | The static Markdown advertises filtering/sorting and scoring materials not present in the body.: Provide the functioning instrument or change the description to match a static table. | Review of the interim draft, finding 100 (P2, format promises). Remove internal claim tags from the reading edition, retaining them in the audit edition. | Judged on the text of the interim draft. | supported with qualification | Markdown reading edition: static wording, no internal claim tags. | Accepted after checking the cited source where one could be reached, or on the text. | Markdown reading edition | | R2-01 | Ch.9: 'A thin record then neither helps nor hurts.' | Round-2 review (P0). Omitting a low score can raise NQ by more than the added shrinkage removes; the reviewer showed MGID 62 to 66 with its commerce score of 1 set to n/e. | Reproduced in the build: every scored cell set to n/e one at a time (data/nq-missing-evidence.json). MGID commerce 1 removed: 61.7 to 65.5, rounded 62 to 66. Largest single-cell moves across the field: +7 and -6. | contradicted | Sentence removed. Chapter 9, the matrix footer and nq-method.csv now say shrinkage damps sparse records but does not make missing evidence neutral, and quote the MGID example and the field-wide range from data. | The claimed property is false for the implemented formula. | Ch.9; matrix footer; build-matrix.mjs; stats.json; nq-method.csv; nq-missing-evidence.csv | | R2-02 | Ch.9: 'This is the standard correction for small samples.' | Round-2 review (P1). Four pseudo-observations, the 45/25/30 weights and the band offsets are design choices; following the Contextual Quotient does not validate them. | scoring-rules.json composite_index; the dimensions are different capabilities, not repeated measures. | supported with qualification | Called an author-defined heuristic, with the reason for four (a third of twelve dimensions; three quarters of the weight stays on a complete record). The build now reruns NQ with two and eight pseudo-observations (max shift 3 and 2 places). NQ is stated to be uncalibrated to buyer outcomes. | Accepted. | Ch.9; scoring-rules.json; matrix footer | | R2-03 | Format-support cells for Taboola and MGID (4 each) cite display, vertical video, placement positions, push and app promotion alongside native units. | Round-2 review (P1). The anchor counts native formats; positions, objectives and non-native formats were mixed in. | Format audit of all 26 cells against Chapter 1's definition, with 18 help-centre and product sources added (FMT-S01 to FMT-S18); two spot-checked by the author. [738, 739, 740] | supported with qualification | Every format cell now lists the native formats counted and names what was excluded. Scores changed: Teads 3 to 4 (specification source), AdSense 3 to 4, Revcontent 4 to 3, Sharethrough 2 to 1, StackAdapt 4 to 3, Media.net, AdNow, Dable, popIn and EX.CO 2 to 1, Microsoft and AppLovin 3 to 1, GumGum 2 to n/e; grades lowered for Dianomi, StackAdapt, Ezoic and APS. Taboola and MGID stay at 4 on the native units alone. NQ and bands recomputed; GumGum becomes a thin record. | Accepted; rationale rewritten for every cell. Close-out: Primis and EX.CO had no qualifying paid native unit; both format cells became n/e and both vendors left the scored set. With them, 13 cells changed value or status in round 2 (12 scores, one to n/e); Dianomi, Ezoic and APS changed grade only. | data/scores.json; data/matrix-extra.json; data/sources-extra.json; NQ; Ch.9 | | R2-04 | Ch.9 and the matrix: bands are the finding, yet the default view and the NQ figure order all 26 vendors, name first and second place and call the front rank 'fully robust'. | Round-2 review (P1). Thin records sit in the same ordering; adjacent vendors with large n/a groups are not directly comparable. | Matrix default order, fig-nq, Ch.9 callout. | supported with qualification | Default matrix view grouped by peer group: banded vendors by NQ, thin records listed apart A to Z with no score shown (score kept in nq.csv and the tooltip). fig-nq plots banded vendors only, by peer group, and lists thin records. Coverage by group added to the tooltip and nq.csv. 'Fully robust' replaced by 'unchanged under the four tested weightings'; first/second-place wording removed. Robustness claims limited to weights, shrinkage and single cells. | Accepted. | instruments.mjs; make-figures.mjs; Ch.9; nq.csv | | R2-05 | Ch.10: a native flight below a Taboola Scaled Advertiser's ~$204,000 quarter 'will rarely separate a small effect from zero'. | Round-2 review (P0). Average spend in other experiments is not a minimum test budget, and a quarterly advertiser metric is not a design; F41 was marked applied without a power calculation. | Two-proportion sample sizes recomputed by the author: 327,922 / 163,095 / 80,682 users per group for 0.5% / 1% / 2% baselines at a 10% relative lift, alpha 0.05 two-sided, 80% power; they match the reviewer. | contradicted | Paragraph replaced by a worked power table with its assumptions and formula, the budget bridge inputs (reach, impressions per person, CTR, CPC), duration drivers, the geo-test caveat and a conditional recommendation. The buyer box now says to run the power calculation first. F41 reopened and closed. | Accepted; the inference was not established. | Ch.10 holdout paragraph and buyer box; F41 | | R2-06 | Ch.10: sharing 'requires one of two things': a Do Not Sell or Share link, or honouring an opt-out preference signal. | Round-2 review (P0). 11 CCR 7025(e) rejects that reading; R78 fixed only the timing. | Approved regulation text (K-S18), 7025(e): 'They do not give the business the choice between posting the above-referenced links or honoring opt-out preference signals'; 7025(f)-(g) set the frictionless conditions. Checked by an AI agent against the PDF and the consolidated text effective 1 January 2026. [319] | contradicted | Rewritten: a business that sells or shares must honour valid signals, generally must also post the link, and may omit the link only under the frictionless-processing conditions and disclosures. K-F12 wording and outcome updated. | Accepted. | Ch.10 privacy paragraph; K-F12 | | R2-07 | Ch.8: 'ChatGPT ads began as a test in January 2026' and 'Ads launched in the United States in February'. | Round-2 review (P1). January was the announcement; the test began on 9 February. | OpenAI, 'Testing ads in ChatGPT', originally published 9 February 2026: 'Today, we're beginning to test ads in ChatGPT in the U.S.'; 'Our approach to advertising...' (16 January 2026) announced plans. Opened by an AI agent with a browser user agent. [241, 240] | contradicted | Both passages now say: announced 16 January, test began 9 February for logged-in US adults on Free and Go. The separate February launch sentence was removed. J-F20 wording updated. | Accepted. | Ch.8 opening and AI assistants paragraph; J-F20 | | R2-08 | Ch.8: 'OpenAI had published no docs for it [Sponsored Agents] at that date.' | Round-2 review (P1). OpenAI published an official announcement on 16 September 2026; an access error does not establish absence. | OpenAI, 'Reimagining advertising with AI' (16 September 2026): 'We're testing Sponsored Agents' with select US advertisers; help-centre article: limited alpha, no early-access requests; no integration documentation. Opened by an AI agent. [244, 245] | contradicted | Paragraph rewritten to cite the announcement and help centre, label the product a limited test, and separate a product description from technical documentation. Agentic map row updated (status beta, primary sources). | Accepted. | Ch.8 AI assistants paragraph; agentic-map | | R2-09 | Ch.8: Skimlinks' 75/25 split 'is not confirmed here' after an access error. | Round-2 review (P2). The support page is reachable and states the split. | Skimlinks support, updated 7 September 2026: 'Skimlinks typically operates on a 75/25 revenue share'; the 75% is of the merchant commission. Opened by an AI agent. [200] | supported with qualification | Sentence now states the typical 75/25 split with the support page as source and notes that terms may differ. H-F05 wording updated. | Accepted. | Ch.8 commerce paragraph; H-F05 | | R2-10 | Ch.7: the Violet account 'names no advertiser, format, design or result'. | Round-2 review (P1). The account reports results (MMM and last-click CPA, higher spend, stable blended performance) but no geo-test lift, interval or design. | Violet playbook (10 February 2025): geo test February 2024; last-click CPA about £43 vs MMM CPA about £13 before the test; investment 'up nearly 300%' with stable blended metrics; no lift, interval or design published. Opened by an AI agent; the chart also shows last-click CPA spikes later in 2024. [172] | contradicted | Rewritten as limited practitioner evidence: what it reports, what it does not, and that its model-based CPA is not an experimental one. Ch.7 in-short and the Ch.11 scenario box aligned. | Accepted. | Ch.7; Ch.11 scenario three | | R2-11 | Universal absence claims: Ch.5 standfirst 'no one has published a representative replication since'; Ch.8 'No native vendor documents a clean room'; Ch.11 'only Taboola documents an interface for agents' and 'No native vendor documents...'. | Round-2 review (P1). Absence claims should state the vendors, sources and cutoff searched. | Nearby passages already used bounded wording; Ch.12 names two unscreened recognition leads. | supported with qualification | Each now says what was searched: 'this review found no...', 'the nine vendors profiled in chapter 9... in the public material this review opened'; the Ch.5 standfirst names the two unscreened leads. | Accepted. | Ch.5 standfirst; Ch.8 clean-room paragraph; Ch.11 agents paragraph | | R2-12 | Ch.8 conclusion: 'The native vendors do not report commerce revenue.' | Round-2 review (P1). Dianomi's profile records £0.5m of FY2025 affiliate revenue (DIANOMI-S14), an internal contradiction. | data/profiles.json, Dianomi financials: affiliate vertical £0.5m in FY2025, from its annual report. [263] | contradicted | Narrowed to the Taboola and Teads Holding Co. filings reviewed; Dianomi named as a small exception, with affiliate commission distinguished from retailer advertising. | Accepted. The amount rests on the annual report already in the register; it was not re-opened in round 2. | Ch.8 conclusion | | R2-13 | Ch.8: learning flowing back to the advertiser 'does not exist'. | Round-2 review (P1). This contradicts the later qualification that contracts may grant more and that aggregate reports come back. | Ch.8 data-flow paragraphs (H-F12). | contradicted | Replaced by: the public documentation reviewed describes aggregate reporting but does not establish a right to export learned models or reusable audience data; contracts may differ. The chapter conclusion now says the same. | Accepted. | Ch.8 data-flow opening and conclusion | | R2-14 | Abstract finding 4 'most buying is still billed per click'; Ch.6 standfirst 'The commercial unit is still a click.' | Round-2 review (P1). 'Most' is a market-wide share with no denominator. | Taboola FY2025 10-K: 'Generally, our charges are based on a CPC, CPM or CPA basis'; no split by model. Checked by an AI agent. | supported with qualification | Headline 4 now reads "Conversion optimisation does not change the billing basis", scoped to the per-click campaigns the help centres document, with target-CPA bidding distinguished from CPA billing. Ch.6 standfirst scoped the same way. E-F25 wording updated. | Accepted. | Abstract; Ch.6 standfirst; E-F25 | | R2-15 | Abstract finding 6: 'The page-view series that feed recommendation units are falling.' | Round-2 review (P1). Chapter 11 distinguishes sessions, ad requests and impressions. | Ch.11 indicators: People Inc. sessions, Ozone ad requests, Dianomi impressions. | contradicted | Abstract now names the three metrics and labels widget contraction an inference; indicator-table interpretations use the proper metric names; scenario two renamed. | Accepted. | Abstract; Ch.11 table and scenario two | | R2-16 | Abstract interests: 'None of these companies is scored in this paper.' | Round-2 review (P1). Teads, named just before, is scored and profiled. | Abstract interests paragraph. | contradicted | Replaced by: 'None of the author's current or former employers listed above is scored. Teads is scored and profiled; its commercial relationship with Samba TV is disclosed here.' Any personal Teads disclosure remains the author's decision. | Accepted. | Abstract | | R2-17 | Ch.10 'Publishers make this call on money, not quality'; Ch.11 'A long-tail widget host cannot [reprice]' and the Outside revenue share 'is why most publishers did not follow'. | Round-2 review (P1). Selected cases do not establish population motives or pricing power. | Ch.10 publisher paragraph; Ch.11 traffic and Outside paragraphs. | supported with qualification | Reworded as observed trade-offs, with the mechanism labelled an inference and no claim about what most publishers did. | Accepted. | Ch.10; Ch.11 | | R2-18 | Ch.11 scenario boxes: 'Ruled out if...' rules. | Round-2 review (P1). The rules are monitoring indicators, not decisive tests of the scenarios. | Ch.11 scenarios and indicator table. | supported with qualification | Each rule is now a 'signal against', with why it is not decisive, and a direct test with its disclosure status (native vs non-native Realize mix; widget impressions across a stable publisher set; budget response after a study). The intro says these are signals, not tests. | Accepted. | Ch.11 | | R2-19 | Ch.6 learning table header 'Minimum conversion volume'. | Round-2 review (P2). The column mixes hard gates, recommendations and learning guidance. | E-S07, E-S13, E-S15, E-S18, E-S28 to E-S30 as already cited. | supported with qualification | Header now 'Conversion volume: gate or guidance'; each row labelled hard gate, recommended, guidance or learning guidance, with its window. | Accepted. | Ch.6 table | | R2-20 | Research-question numbers: 'question four' (incrementality), 'question eight' (agentic, twice), 'question two' (growth, three times), 'question nine' (dependence). | Round-2 review (P2). The numbers follow an earlier brief, not the list in chapter 12. | Ch.12 list: incrementality is 6, AI and agentic buying is 10. | contradicted | All numbered references replaced by topic names. | Accepted; the chapter 2 and 8 references had the same fault. | Ch.2; Ch.7; Ch.8; Ch.11 | | R2-21 | Ch.1 six-axis example: the two units 'share the first two axes and nothing else'; 'Each placement inside it takes one.' | Round-2 review (P2). They differ in format and surface; targeting can take several values for one placement. | Ch.1 example and fig-category-map footer. | contradicted | Example now says both are native in the broad sense but differ on every axis named; placements usually take one value per axis, with targeting able to take several. Figure footer aligned. | Accepted. | Ch.1; fig-category-map | | R2-22 | Ch.2: IAB/PwC 'is the closest thing to an audited view'. | Round-2 review (P2). PwC does not audit the estimates. | Ch.2 and the IAB/PwC methodology note already cited. | supported with qualification | Now: 'a structured industry revenue survey, filled out with estimates. It is not an audit.' | Accepted. | Ch.2 | | R2-23 | Ch.7 and fig-rct-vs-observational: '663 tests', '663 Facebook experiments', '663 RCTs'. | Round-2 review (P2). The study counts 563 experiments, 663 treatment-control pairs and 1,673 pair-outcome observations. | arXiv 2201.07055 v2, section 3.2: '563 experiments'; '663 treatment-control pairs'; '1,673 conversion events... we will refer to each experiment-conversion event as an RCT'. Checked by an AI agent. [164] | supported with qualification | Unit made consistent in the text, the figure title, the study table and F-F04. | Accepted. | Ch.7; fig-rct-vs-observational; study-evidence; F-F04 | | R2-24 | Ch.8: 'The largest native firm's 10-K no longer leads with native.' | Round-2 review (P2). Reviewer: Taboola's FY2025 annual report is a Form 20-F. | EDGAR: the cited filing's cover reads 'FORM 10-K' (filed 2026-02-25); Taboola filed a 20-F only for FY2021 and has filed 10-Ks and 10-Qs since. Checked by an AI agent. | supported as stated | No change. | Rejected: the paper is correct; the reviewer appears to rely on Taboola's 2021 status. | Ch.8 | | R2-25 | Print edition: body text measured about 7.1pt, tables about 5.4pt, NQ figure labels about 4.1-4.5pt, references about 6.4pt. | Round-2 review (P0). An over-wide element made Chrome shrink every page. | Reproduced: in print layout at the A4 text width the matrix was 1,001px against 665px, with three tables also too wide (a 0.66 ratio). | structural | Print layout rebuilt: matrix with rotated headers and fixed column widths; wide tables on landscape pages; tables hyphenate instead of breaking mid-word; the two long ledgers print as pointers to their CSVs; figure type scaled 1.2x with a label-collision check. build/pdf.mjs now refuses to print if anything is wider than its page, and scripts/pdf-type.py measures the printed type: body 10.5pt, references 8.5pt, tables 7.5pt. | Accepted. | paper.template.html; pdf.mjs; svg.mjs; make-figures.mjs; figure-text-check.mjs; pdf-type.py | | R2-26 | PDF p.43: continuation of the capability matrix has blank column headers. | Round-2 review (P1). Continuation pages must repeat readable headers. | Reproduced in the rebuilt PDF: the repeated header row painted without its labels. | structural | Print-only header labels added in place of the sort buttons; continuation pages checked in the rebuilt PDF. | Accepted. | instruments.mjs; paper.template.html | | R2-27 | Printing from a filtered page drops hidden rows and references. | Round-2 review (P1). Print rules did not override the filter state. | Code path confirmed: .is-hidden used display:none with no print override. | structural | Print overrides added for list items and table rows; checked by filtering and printing in a browser. | Accepted. | paper.template.html | | R2-28 | Downloads link data/derived-checks.json and data/stats.json, which were not in the published data folder. | Round-2 review (P1). Broken links in the package. | Confirmed. | structural | Both files (and nq-missing-evidence.json) are now copied to dist/data. check.mjs now fails on any relative href or src in the HTML that does not resolve, and on a #key missing from its JSON file; it found one more (a figure pointing at a non-existent key), fixed. | Accepted. | make-csv.mjs; build-data.mjs; check.mjs | | R2-29 | ZIP member names use backslashes. | Round-2 review (P1). Extraction on Linux or macOS produces flat files with literal backslashes. | Confirmed: Windows PowerShell's Compress-Archive writes backslash paths. | structural | package.mjs now writes the ZIP itself with forward-slash paths (no PowerShell), and re-reads it to verify names and links. | Accepted. | package.mjs | | R2-30 | The PDF script looks only for Windows browser paths and packaging calls powershell.exe; the guide does not say Windows is required. | Round-2 review (P1). Reproduction on other systems fails. | Confirmed. | structural | Browser discovery now covers Windows, macOS and Linux (or CHROME_PATH); packaging is pure Node. The browser version used is recorded in data/pdf-build.json. README updated. | Accepted. | chrome.mjs; pdf.mjs; package.mjs; README | | R2-31 | Running the packager from the shipped source/ tree fails because source/AUTHOR-HANDOFF.md is missing. | Round-2 review (P1). The source tree cannot reproduce its own package. | Confirmed. | structural | The public handoff is now included in source/, and the packager skips files it cannot find with a warning instead of failing. A build from an extracted copy was run. | Accepted. | package.mjs | | R2-32 | Root README build commands start at the extraction root, which has no package.json. | Round-2 review (P2). npm install fails. | Confirmed. | structural | README now starts with 'cd source' and lists tool versions. | Accepted. | README | | R2-33 | 14 of the 16 checker warnings were path-validation faults. | Round-2 review (P2). The checker treated 'file#key' as a filename and looked for CSVs in the input folder. | Confirmed. | structural | File and key are now validated separately, in data/ and dist/data/; a missing key is a failure. Unused register sources are reported as information, not warnings. One long SWOT line shortened. | Accepted. | check.mjs | | R2-34 | Ch.12: 'The build fails if any of those [availability, grade, date, rationale, source] is missing.' | Round-2 review (P1). The checker did not enforce all of it and the builder defaulted availability to live and grade to C. | Confirmed in build-matrix.mjs; all 176 scored cells already carried the fields, so no output changed. | structural | The builder now refuses a scored cell with missing or invalid metadata, and check.mjs validates availability, grade and date values. | Accepted; the statement is now true. | build-matrix.mjs; check.mjs | | R2-35 | Readability 67.3 against a target of 85. | Round-2 review (P2). The target is unmet. | Reproduced with scripts/readability.py. | supported with qualification | Recorded as unmet in the handoff. The round-2 rewrites kept to plain wording; no fragmenting of technical reasoning to raise the score. | Accepted as a limitation. | AUTHOR-HANDOFF | | R2-36 | The PDF has no bookmarks or page numbers; p.61 held only a table footer. | Round-2 review (P2). Navigation and pagination. | Confirmed. | structural | The PDF now carries chapter bookmarks and printed page numbers, table footers are kept with their tables, and the type check reports near-empty pages (none). | Accepted. | pdf.mjs; paper.template.html | | R2-37 | Profile headings in print use full legal names with registry numbers and addresses (MGID, Readpeak). | Round-2 review (P2). Administrative metadata gets headline weight. | Confirmed on the Readpeak page. | structural | Headings use the trading name; the legal entity and registry detail sit in a compact line below. | Accepted. | instruments.mjs; paper.template.html | | R2-38 | Handoff and ledger closure language: 78 applied / 13 partly applied, while F41 was not fully implemented and 326 claims remain unresolved. | Round-2 review (P1). Closure and verification language must match what was checked. | F41 confirmed incomplete (see R2-05). | supported with qualification | Every ledger entry now has separate implementation and verification fields. F41 recorded as partly applied at the final review and completed in round 2, so the final-review count is 77 applied and 14 partly applied before round 2. The handoff states how many claims were not re-checked (325 after round 2) and that no external audit took place. | Accepted. | corrections-ledger; AUTHOR-HANDOFF | | R2C-01 | fig-nq units: '100 = every scored dimension at 5'; Ch.9 and Ch.12: the shared construction 'keeps the indices comparable'. | Round-2 close-out (P1). Under the implemented formula a vendor scored 5 on all twelve dimensions gets NQ 89 at the current field mean, not 100; a shared construction does not make NQ and CQ values interchangeable. | Recomputed: 100 x ((12 x 5 + 4 x field) / 16) / 5 = 88.6 at the round-2 field mean. | contradicted | Caption now states the scale and computes the all-fives endpoint from the current field mean; Ch.9 gives the same endpoint from stats.json. Ch.9 and Ch.12 now say the indices share a construction and a way of reading them, but their scores and bands are not interchangeable. | Accepted. | make-figures.mjs; stats.mjs; Ch.9; Ch.12 | | R2C-02 | PDF data-download links. | Round-2 close-out (P1). All 22 links in the PDF pointed to file:/// paths in the author's build folder. | Confirmed: 22 /URI (file:...) annotations in the PDF. | structural | pdf.mjs now removes relative links before printing, keeping the text, and refuses to write a PDF with any file: link. The downloads block in print says the files are in the data/ folder beside the PDF. The PDF and Markdown editions were added to the downloads list. | Accepted. | pdf.mjs; instruments.mjs; build-data.mjs | | R2C-03 | Abstract and handoff 'No public study with published design and results shows what open-web native adds to sales'; abstract 'No one has measured the bias...'; Ch.5 'No one has measured ad literacy at population scale' and 'No peer-reviewed replication exists'; Taboola and Teads SWOT lines. | Round-2 close-out (P1). Unbounded absence claims survived R2-11. | Located in the text and the SWOT files. | supported with qualification | Each now states what this review found in the sources screened by the cutoff. | Accepted. | Abstract; Ch.5; handoff; evidence/swot/taboola.json; evidence/swot/teads.json | | R2C-04 | Ch.8: 'the paid tiers stayed free of ads.' | Round-2 close-out (P1). Go is a paid tier and was in the test. | OpenAI's 9 February announcement (R2-S02) lists Free and Go as eligible and Plus, Pro, Business, Enterprise and Education as excluded. [241] | contradicted | Now: the test covered Free and Go; Plus, Pro, Business, Enterprise and Education were excluded from the initial test. | Accepted. | Ch.8 | | R2C-05 | Primis and EX.CO format support scored 1 as borderline in-content video players. | Round-2 close-out (P1). Player location or styling is not a paid native unit; score only what the record establishes, even if the vendor leaves the matrix. | AI check of both vendors' sites, docs, releases and trade press: Primis uses 'native' only for player styling and a placement type; EX.CO's paid units are instream and outstream video, and its only qualifying unit was Playbuzz branded content in 2015-2017. [688, 732] | contradicted | Both cells set to n/e. Both vendors fall below four evidenced dimensions and leave the scored set; their records are kept in data/matrix-excluded.json and both stay in the vendor universe with the reason. NQ, field mean, bands and every count recomputed. | Accepted. | matrix-extra.json; matrix-excluded.json; universe-overlay.json; NQ; all counts | | R2C-06 | PDF p.92 held only the universe table note; the page check counted footer text and did not fail. | Round-2 close-out (P2). Orphaned note and a checker that could not see it. | Root cause found: render.mjs appended '' after every '', closing each instrument box early, so every table footer sat outside its box on the web page and in print. | structural | Wrapper fixed so only plain Markdown tables get it; table notes also print above their table. pdf-type.py now counts body text only (footer excluded) and fails on any near-empty page. | Accepted. | render.mjs; instruments.mjs; paper.template.html; pdf-type.py | | R2C-07 | Handoff: reaching 85 'would mean breaking technical reasoning into misleading short sentences'. | Round-2 close-out (P2). That trade-off was asserted, not shown. | Handoff decision 4. | contradicted | Replaced by: the target remains unmet; reaching it needs a dedicated plain-language edit followed by a meaning check. R2-35 stays open. | Accepted. | AUTHOR-HANDOFF | | R2C-08 | scoring-rules.json bands_rule: 'the native field mean of 56'. | Round-2 close-out (P2). Stale number; the field NQ is 55. | Confirmed in scoring-rules.json and capability-matrix.json. | contradicted | The sentence no longer carries a number; the thresholds are stated from the current field mean elsewhere. | Accepted. | scoring-rules.json | | R2C-09 | Work log: '11 format scores changed'. | Round-2 close-out (P2). The two matrices differ in 13 cells by value or status. | Counted from the matrices. | contradicted | Handoff and R2-03 now say 13 cells changed value or status (12 scores, one to n/e), with three more changing grade only. | Accepted. | AUTHOR-HANDOFF; R2-03 | ### The claim ledger This lists each claim marked in the text, with its evidence class, review outcome, confidence and sources. In the web edition, each claim opens to show the supporting passage, scope, derivation and limits. Headline claims come first. Claims the research recorded but the text does not use stay in the full ledger in the data files. **Claim ledger** | ID | Claim (final wording) | Ch. | Evidence class | Review | Confidence | Sources | |---|---|---|---|---|---|---| | A-F09 (headline) | EMARKETER's December 2025 forecast puts US native display ad spending at $147.98B for 2026 (+13.1%), implying about $130.8B for 2025. An AI-assisted EMARKETER FAQ describes native formats as including in-feed, sponsored content, recommendation widgets, in-app rewarded video and sponsored outstream video. Whether the forecast itself counts rewarded or outstream video is not stated on any page opened. | abstract, 2 | forecast | supported with qualification | medium | [9] | | A-F12 (headline) | Social feeds took 74% to 84% of eMarketer-defined US native display in every vintage whose split could be read (the 2017 to 2019 vintages); a later split exists only as an unopened lead. | abstract, 2 | synthesis | supported with qualification | medium | [20, 21, 19] | | A-F20 (headline) | Taboola and Teads Holding Co. together booked $3.21B gross and $1.24B ex-TAC in FY2025, about 1.6% (gross) and 0.6% (ex-TAC) of Meta's global ad revenue. On a US basis their combined US revenue of about $1.26B is about 0.43% of IAB/PwC's $294.6B US total and about 1.7% of Meta's $74.78B US revenue. | abstract, 2 | synthesis | supported with qualification | medium | [28, 29, 23, 7] | | C-F04 (headline) | Taboola's cost of publisher guarantees (payments due under guarantee arrangements in excess of what revenue-share terms would have required) was approximately 15% of TAC in 2025 and 18% in 2024, up from 10% in 2022, 9% in 2021 and 13% in 2020; in H1 2026 it was about 15% (17% in H1 2025) and 13% in Q2 2026 (16% in Q2 2025). | abstract, 4, 10 | direct record | supported as stated | high | [1, 77, 47] | | D-F01 (headline) | In the field's anchor US experiment (fieldwork before Dec 2015), only 17 of 242 adult MTurk readers (7%) recognised a sponsored article on a newspaper-style page as advertising across all 12 disclosure conditions, and 11 of 60 students (18.3%) did so in a follow-up eye-tracking lab study that used the clearest wording ('sponsored by Dell'). | abstract, 5 | independent measurement | supported as stated | high | [99] | | D-F05 (headline) | Across 16 real native ads (each respondent saw 8) from nine to ten publisher and social platforms (the paper says eight, lists nine, and its table adds Fast Company), 37% of 896 valid SSI opt-in panel respondents (January 2015, US context) identified native ads as paid vs 81% for regular ads. Per-ad recognition ran from 21% (Forbes/SAP; Onion/Burger King) to 72% (Facebook/Jaspers). | abstract, 5 | independent measurement | supported with qualification | high | [102] | | E-F25 (headline) | For Taboola Realize, Outbrain and MGID, the per-click conversion-optimised campaigns their help centres document are billed per click and the platform's conversion model decides which clicks to buy; the advertiser still carries the conversion risk, paying per click whether or not a conversion follows. A target CPA is a bidding setting, not CPA billing; both large platforms' filings also describe CPM and CPA pricing without disclosing revenue shares by model. | abstract, 6 | synthesis | supported with qualification | medium | [136, 138, 139, 142, 143, 134, 1] | | F-F04 (headline) | Across 663 treatment-control pairs from 563 large Facebook experiments, with over 5,000 user-level features, double machine learning produced median lifts of 83%, 58% and 24% for upper-, mid- and lower-funnel outcomes against RCT medians; the study's own 'RCT' unit is the pair-outcome, 1,673 in all. | abstract, 7 | independent measurement | supported with qualification | high | [164] | | F-F21 (headline) | Taboola's 2023 Kantar meta-analysis reports average lifts such as 15% in favourability against a 4.2% all-digital norm, using exposed-versus-unexposed groups; most studies included video, and against Kantar's video-only norm the favourability lift was nearly identical, at about 15%. | abstract, 7 | commissioned measurement | supported with qualification | low | [166] | | G-F22 (headline) | On both large native self-serve platforms, placement reports appear only once a campaign has data and controls are exclusion-first; neither documents a pre-campaign site list, though Outbrain DSP documents publisher include lists at setup. | abstract, 10 | vendor assertion | supported with qualification | high | [292, 293, 294, 295, 296] | | J-F08 (headline) | Taboola, Teads and MGID do not appear as AdCP or AAMP implementers on any page opened, and neither Taboola's nor Teads' Q2 2026 results mention AdCP, AAMP, MCP or agent-to-agent buying. | abstract, 11 | inference | unresolved | medium | [375, 382, 30, 271, 386, 387] | | L-F16 (headline) | People Inc.'s core sessions fell 5% in fiscal 2025, then 13% in Q4 2025, 17% in Q1 2026 and 22% in Q2 2026, attributed by the company primarily to Google AI Overviews. | abstract, 11 | direct record | supported with qualification | high | [354, 355, 356, 215] | | B-F23 | Programmatic native transacts under IAB Tech Lab's OpenRTB Native Ads specification, introduced with OpenRTB 2.3 in January 2015 and last finalised as version 1.2 in March 2017 (adding third-party/dynamic creative, eventtrackers and a privacy link), carried as imp.native in the bid request and a JSON string in the bid adm; the latest OpenRTB 2.x release is 2.6-202309 and no IAB adoption statistics for native were found. | 1, 3 | direct record | supported with qualification | high | [4, 67] | | H-F25 | Taboola's FY2025 10-K reframes the company from 'a leader in native advertising' to a performance-advertising platform (Realize, launched February 2025) that 'addresses the limitations of native advertising alone', supports formats re-using display and social creative, and claims a $55 billion opportunity based on internal and external industry data. | 1, 8 | direct record | supported as stated | high | [1] | | K-F25 | IAB's Native Advertising Playbook 2.0 (May 2019) states that all three native ad types (in-feed/in-content, content recommendation, branded/native content) 'must include a disclosure to signal to the consumer that these are paid ads', that disclosure language must convey the content 'has been paid for' and be 'large and visible enough' to notice, and it defers to the FTC's two December 2015 documents; it is industry guidance with no legal force and has not been superseded as far as this stream found. | 1, 10 | direct record | unresolved | high | [3] | | A-F01 | The IAB Native Advertising Playbook 2.0 (May 2019) reduced native to three core ad types: In-Feed/In-Content Native Advertising, Content Recommendation Ads and Branded/Native Content, down from six types in the December 2013 playbook. | 2 | direct record | supported as stated | high | [3, 6] | | A-F02 | Playbook 2.0 explicitly removed Paid Search from the native taxonomy even though 'Search ads technically meet the IAB definition of native', and folded Promoted Listings (commerce-site in-feed ads) into the In-Feed type. | 2 | direct record | unresolved | high | [3] | | A-F03 | Under the IAB taxonomy, In-Feed native spans three feed surfaces - content feeds (e.g. CNN, Yahoo), product feeds (e.g. Amazon, Etsy, eBay) and social feeds (e.g. Facebook, Instagram, Twitter) - and 'On social feeds, there are no In-Content Ads'. | 2 | direct record | unresolved | high | [3] | | A-F04 | Playbook 2.0 dropped 'Buying & Selling' and 'Measurement' as criteria for whether an ad is native, leaving Design, Location, Ad Behavior and Disclosure; the IAB states that with the growth of native programmatic, buying route no longer defines nativeness. | 2 | direct record | unresolved | high | [3] | | A-F05 | US internet advertising revenue reached $294.6 billion in 2025 (+13.9% YoY) per IAB/PwC, and native is NOT reported as its own line: it is buried inside Display ($81.6B), which 'includes banner, rich media, sponsorship and native revenues'. | 2 | direct record | supported as stated | high | [7, 8] | | A-F06 | In IAB/PwC's FY2025 report, 'Social media' ($117.7B, +32.6%) is a cross-cutting channel rather than a format: the five formats (search $114.2B, display $81.6B, video $78.0B, audio $8.4B, other $12.5B) already sum to $294.7B, i.e. the total, so social revenue is embedded inside display and video. | 2 | inference | supported as stated | high | [7] | | A-F07 | Display (+9.8%) was the slowest of the four named formats (search, display, video, audio). The residual 'Other' category (classifieds, directories, lead generation) grew more slowly, at +6.9%. The report's own 'smallest YoY growth across the formats' wording ignores 'Other'. | 2 | direct record | supported with qualification | high | [7] | | A-F08 | IAB/PwC figures are US-based, GAAP 'earned' revenues collected by a PwC web survey of selling companies plus public data and 'a conservative revenue estimate' for non-participants; PwC states it has not audited or verified the data. | 2 | direct record | unresolved | high | [7] | | A-F10 | EMARKETER states that native's share of total US display has declined 'as connected TV and retail media expand', i.e. native is growing more slowly than total display under its definition. | 2 | forecast | unresolved | medium | [9] | | A-F11 | Public eMarketer values exist for 2019-2023 (Jan 2023 vintage: $47.29B, $58.69B, $80.64B, $87.03B, $97.46B), so the unfilled gap is 2024-2025 only. The 2019 base was itself restated from $43.90B to $47.29B. Using the later vintage's 2019 value, the implied 2019-2026 CAGR is about 17.7%, not 19.0%. | 2 | synthesis | supported with qualification | medium | [9, 21, 18, 20] | | A-F13 | In 2019 eMarketer expected native to be 95.6% of US social display spend but only 30.8% of nonsocial display spend (up from 19.1% in 2017), so 'native display' as measured is largely a relabelling of social-feed advertising. | 2 | forecast | supported with qualification | high | [18] | | A-F14 | eMarketer's digital video category explicitly includes 'outstream video ads such as native, in-feed (including video ads in Facebook's News Feed and Twitter's Promoted Tweets), in-article, in-banner, and interstitial video ads', so its 'native display' and 'digital video' totals overlap and cannot be added. | 2 | direct record | unresolved | high | [10] | | A-F15 | Meta's FY2025 advertising revenue was $196.175 billion (+22% from $160.633 billion), with ad impressions up 12% and average price per ad up 9%; volume contributed roughly 57% and price 43% of the growth. | 2 | direct record | supported as stated | high | [23] | | A-F16 | Google Network ($29.792B, -1.9%) is Google's third-party-property line. It combines AdSense (web), AdMob (apps) and Ad Manager, and the 2025 decline was driven by AdSense and partly offset by AdMob growth. The web-publisher piece therefore fell by more than 1.9%, though the filing does not quantify it. | 2 | direct record | supported with qualification | high | [24, 25] | | A-F17 | Amazon 'Advertising services' net sales were $21.317 billion in Q4 2025 (+23% YoY) and $68.6 billion for FY2025 when the four reported quarters are summed; Amazon defines the line as sales 'through programs such as sponsored ads, display, and video advertising'. | 2 | direct record | supported as stated | high | [27] | | A-F21 | MAGNA's June 2025 Global Ad Forecast (media-owner net ad revenue) put 2025 at $979 billion (+4.9%, cut 1.2 points from December 2024's +6.1%/$990B), 2026 at +6.3% passing $1 trillion, and US Digital Pure Players at $297 billion (+10%) in 2025; native is not broken out in any MAGNA public release opened. | 2 | forecast | unresolved | medium | [11, 12, 13] | | A-F22 | None of the three agency-group forecasters opened for 2026 publishes a native category: WPP Media (June 2026: global +8.9% to $1.3 trillion excluding US political) and dentsu (December 2025: 2026 +5.1%, digital 68.7% of total, US +5.0%) report search/social/retail media/video only, and Zenith's forecasts are gated. | 2 | forecast | unresolved | medium | [14, 15] | | A-F23 | IAB Europe's AdEx Benchmark 2025 (30 markets) reports EUR131 billion of European digital ad spend (+10.5%), with video EUR34.0bn (+19.6%, 'more than half of all display investment'), social EUR35.5bn (+19.2%), display +12.5%, search +8.8% and retail media EUR13.3bn (+16.7%), and has no native category. | 2 | direct record | unresolved | high | [16] | | A-F24 | The IAB Tech Lab OpenRTB Native Ads standards page (v1.2 'finalized') still describes the API as covering 'the 6 types of native ads defined by the IAB Native Advertising Playbook', i.e. the 2013 six-type taxonomy that Playbook 2.0 replaced in 2019. | 2 | synthesis | unresolved | low | [17, 3] | | B-F01 | Taboola holds a 30-year exclusive commercial agreement, announced 2022-11-28 and closed 2023-01-17, to exclusively power native advertising across all of Yahoo's digital properties, projected by the parties at approximately $1 billion of annual revenue and more than 800 billion impressions, with Yahoo receiving 24.99% of Taboola's shares (about 60% ordinary, 40% non-voting) and one board seat. | 3 | direct record | supported as stated | high | [33, 34] | | B-F02 | The Yahoo agreement was amended at least three times between June 2025 and March 2026 (Amendments No. 5, 6 and 7, effective 2025-06-01, 2025-08-11 and 2026-03-01), adding a Yahoo direct-insertion-order integration ('Yahoo App'), Salesforce portal access through 2027-11-30, and a rewritten property-level, monthly performance-assessment methodology with redacted 'Performance Decreases' calculations. | 3 | direct record | supported as stated | medium | [35] | | B-F03 | Other than Yahoo and Microsoft, no Taboola digital-property partner accounted for 5% or more of Taboola's 2025 revenues generated from advertisers on digital properties (the 10-K gives no percentages for Yahoo or Microsoft); Taboola's ten largest advertisers were under 10% of total network revenues, with none above 3%. | 3 | direct record | supported with qualification | high | [1] | | B-F05 | Taboola's FY2025 10-K states that historically the majority of its digital-property agreements have typically required exclusivity or other preferred-usage incentives for the term of the agreement, and that as of 2025-12-31 its average contract term was over two years at inception. | 3 | direct record | supported with qualification | high | [1] | | B-F08 | Taboola launched Realize on 2025-02-26 as a performance platform opening display and other non-native placements across a network it describes as ~600 million daily active users, and by 2025-10-15 TIME, Weather Channel Digital, Gannett \| USA TODAY Network, Nexstar and Slate had opened display inventory to Realize advertisers. | 3 | vendor assertion | supported as stated | high | [48, 49] | | B-F11 | Outbrain Inc. announced the acquisition of Teads from Altice on 2024-08-01 for about $1 billion ($725M upfront cash, $25M deferred, 35M shares, $105M convertible preferred) and closed on 2025-02-03 on revised terms of $625M cash plus 43.75M shares (~$900M, no deferred or preferred), leaving Altice with about 46.6% of the shares; the company renamed to Teads Holding Co. effective 2025-06-06 and began trading as TEAD on 2025-06-10. | 3 | direct record | supported as stated | high | [44, 45, 46] | | B-F13 | Teads Holding Co. describes three media-owner contract structures - revenue share, programmatic bidding, and guaranteed minimums ('We may commit to a guaranteed minimum rate... to access... inventory') - across approximately 10,000 media owners, with its top 20 media partners averaging seven years' tenure, and about 1,700 employees as of March 2026. | 3 | direct record | unresolved | high | [46] | | B-F15 | Life360 announced the acquisition of Nativo on 2025-11-10 for approximately $120 million in cash and stock and completed it on 2026-01-05 (65% cash, 35% stock); Nativo was described as having 'direct integrations to hundreds of publishers' and Life360 as having 91.6M global MAU (2025-09-30) and 50M+ US MAU. | 3 | direct record | supported as stated | high | [57, 58] | | B-F16 | Sharethrough merged with Equativ (Bridgepoint-backed) in a deal announced 2024-06-12 - not 2023 as the lead stated - creating a company of 720+ employees in 18 countries with claimed net recurring revenue above $200M, and the Sharethrough brand was retired into Equativ on 2025-06-09. | 3 | vendor assertion | supported as stated | high | [52, 53] | | B-F20 | On 2026-09-18 Dianomi's board announced it intends to unanimously recommend a Taboola Europe Limited cash acquisition, by court-sanctioned scheme of arrangement, at 64p per share (~GBP 19M) plus up to 24p of contingent consideration (up to ~GBP 27M in total). The contingent part is payable only if a subset of Dianomi publishers adopt Taboola standard terms, and is scaled to their net revenue. Irrevocables cover ~75.3%; the deal is conditional on the CMA not pursuing a Phase 2 review and on shareholder approvals. | 3 | direct record | supported with qualification | high | [60, 61, 31] | | B-F25 | Major buying and serving platforms document support for component-based open-web native (DV360, The Trade Desk since March 2016, Google Ad Manager, AdSense, Microsoft Advertising), and Amazon Publisher Services announced APS Native ads for Amazon advertiser demand on 2025-05-21 (availability, markets and third-party DSP access not stated). | 3 | synthesis | supported with qualification | high | [69, 70, 71, 72, 73, 38, 74] | | B-F27 | Between June 2024 and September 2026 the independent open-web native tier consolidated into fewer owners: Sharethrough into Equativ (announced 2024-06-12), Zemanta into Outbrain DSP (2024-08-06), Teads into Outbrain/Teads Holding Co. (closed 2025-02-03), Nativo into Life360 (closed 2026-01-05) and Dianomi into Taboola (offer 2026-09-18, pending), leaving TripleLift (Vista), MGID, Revcontent, Readpeak and Kargo as the notable remaining independents. | 3 | synthesis | unresolved | high | [52, 56, 45, 58, 31, 62, 65, 407, 408, 741] | | B-F28 | Supply is more concentrated than demand at the largest native platform: two partners (Yahoo, Microsoft) each account for at least 5% of Taboola's 2025 revenues while its top ten advertisers together are below 10%, with none above 3%. | 3 | inference | unresolved | medium | [1] | | C-F09 | About 2,200 Scaled Advertisers accounted for 84% of Taboola's Q4 2025 revenue (not full-year 2025 revenue). | 3 | direct record | supported with qualification | high | [1] | | C-F10 | Taboola's stated count of digital property partners fell from 'approximately 14,000' (Q4 2025, 10-K) to 'approximately 12,000' (Q2 2026 10-Q). | 3 | direct record | unresolved | low | [1, 47] | | B-F29 | The two listed native platforms together reported $3.2 billion of FY2025 gross revenue (Taboola $1.9B, Teads Holding Co. $1.3B) and $1.24 billion of ex-TAC gross profit, but the figures are not a native-market total because Teads includes video, CTV and display and Taboola includes display sold via Realize and Yahoo revenue recognised on a net basis. | 3, 4 | synthesis | unresolved | high | [348, 75, 41] | | C-F01 | Taboola's FY2025 gross revenue was $1,912.0M (+8.3% YoY), traffic acquisition cost $1,214.9M, GAAP gross profit $569.5M, ex-TAC gross profit $713.5M (+6.9%), Adjusted EBITDA $215.5M (+7.2%), GAAP net income $42.3M (vs a $3.8M loss in 2024) and free cash flow $163.4M. | 4 | direct record | supported as stated | high | [32, 1] | | C-F03 | Under the definition used before Q2 2026, Taboola's ex-TAC gross profit grew about 4.7% in Q2 2026. The 11.8% headline adds back a $12.2M one-time, non-cash write-off of publisher prepayments that were booked in TAC. The write-off also depresses GAAP gross-profit growth (2.9%); excluding it, gross profit grew about 11.8%. The 4.7% is therefore 'same-definition' growth, not necessarily 'underlying' growth. Which figure is more representative depends on whether the impaired prepayments are treated as a recurring cost of supply. | 4 | synthesis | supported with qualification | high | [47, 76] | | C-F05 | Taboola discloses two publisher compensation models: a revenue share (most common) and a 'Minimum guarantee model' under which it pays the greater of a revenue-share percentage or a committed guaranteed amount per thousand page views, settled monthly; guarantee contracts generally run 2-5 years, large digital property contracts 'in general, contain minimum guarantee requirements', and the average contract term at inception exceeds two years. | 4 | direct record | unresolved | high | [1, 47] | | C-F08 | The Yahoo deal was paid for in equity (24.99% of shares, ~60% ordinary/40% non-voting, one board seat), booked as a Commercial agreement asset of $270.2M at end-2025 ($262.1M at June 30 2026) that is amortised within TAC ($2.8M in 2024, $16.4M in 2025, $8.1M in H1 2026) and added back in ex-TAC; Taboola meanwhile repurchased 76.9M shares in 2025 at an average $3.30 (about $255.4M cash) and a further 16.2M shares in H1 2026 at $3.99. | 4 | direct record | unresolved | high | [80, 1, 47, 32] | | C-F11 | Taboola recognises revenue gross (as principal) for most arrangements, with payments to publishers booked as TAC in cost of revenues, and net only where it acts as agent; its auditor treats the principal-versus-agent judgment as requiring 'a high degree of auditor judgment', and TAC also includes 'cost for advertising impressions purchased from real-time advertising exchanges and other third parties' as well as up-front payments, incentives and bonuses to publishers. | 4 | direct record | unresolved | high | [1] | | C-F14 | Teads Holding Co.'s FY2025 revenue was $1,300.5M (+46% reported, driven by consolidating legacy Teads from 3 Feb 2025), TAC $770.8M, other cost of revenue $100.6M, gross profit $429.1M (33.0% margin), ex-TAC gross profit $529.7M, Adjusted EBITDA $93.4M (17.6% of ex-TAC), operating cash flow $7.6M, free cash flow -$15.2M (adjusted FCF $6.0M) and a net loss of $517.1M that included a $352.1M non-cash goodwill impairment, $15.5M of intangible/software write-offs, $28.9M acquisition costs and $15.3M restructuring charges. | 4 | direct record | supported as stated | high | [75, 46] | | C-F15 | On a pro forma basis (as if the Teads acquisition had occurred on 1 Jan 2024) Teads Holding Co.'s revenue fell 11.7% from $1,507.2M in 2024 to $1,330.9M in 2025, with a pro forma net loss of $545.5M; legacy Teads contributed $517.2M of revenue and a $350.8M net loss from 3 Feb to 31 Dec 2025, and legacy Outbrain TAC fell $97.2M 'consistent with the decline in Outbrain revenues'. | 4 | direct record | unresolved | high | [46] | | C-F16 | Teads' ~$900M acquisition closed on 3 February 2025. Its $625M cash portion was first funded by a bridge facility, which was refinanced on 11 February 2025 with $637.5M of 10.000% senior secured notes due 2030, issued at 98.087% by the subsidiary OT Midco. | 4 | direct record | supported with qualification | high | [46, 78, 82] | | C-F17 | Teads Holding Co.'s 2026 results deteriorated sharply: Q1 revenue $266.0M (-7%), ex-TAC $107.9M (+5%), Adjusted EBITDA $0.8M (-93%), adjusted FCF -$41.1M; Q2 revenue $284.6M (-17%), gross profit $95.6M (-21%), ex-TAC $123.4M (-14%), Adjusted EBITDA $7.0M (-74%), net loss $42.5M; on 6 Aug 2026 it suspended guidance including the ~$100M FY2026 Adjusted EBITDA target, citing 'the volatility of the Direct Response and SME business' facing 'open-web headwinds'. | 4 | direct record | supported as stated | high | [83, 82, 75] | | C-F18 | Teads Holding Co.'s reported H1 2026 Adjusted EBITDA of $7.7M was about a quarter of the roughly $31.4M semi-annual coupon on its 10% notes. CTV, which the company highlights as its main growth driver, was 13% of Q2 2026 revenue (7% a year earlier) after 67% YoY growth. | 4 | synthesis | supported with qualification | medium | [83, 82, 46] | | C-F19 | Teads Holding Co. (and Outbrain before it) describes three supply-payment structures - revenue share, programmatic bidding and guaranteed minimums - and warns that guarantees 'require us to pay our media partner for the ad impressions we receive, regardless of whether the consumer engages with the ad or we are paid by the advertiser', that profitability 'has been and may continue to be adversely impacted' by them, and that TAC 'may not correlate with fluctuations in revenue'; no media partner accounted for 10% of TAC and no marketer for 10% of revenue in 2023-2025. | 4 | direct record | unresolved | high | [46, 78] | | C-F20 | The Outbrain lineage shrank for three years before the merger: revenue fell from $992.1M (2022) to $935.8M (2023) to $889.9M (2024) while TAC fell from 76.3% to 75.7% to 73.5% of revenue, ex-TAC gross profit was $234.8M / $227.4M / $236.1M, Adjusted EBITDA rose from $28.5M (2023) to $37.3M (2024), free cash flow swung from -$6.5M to $51.3M, and 2024 revenue fell about $107M from 'net revenue retention of 88% on existing media partners' (lower impressions) offset by about $60M from new partners. | 4 | direct record | unresolved | high | [78] | | C-F22 | Dianomi (AIM-listed, financial/business-vertical native ads) reported FY2025 revenue of GBP27.4M (-2.1%; GBP28.1M at constant currency), publisher revenue-share payments (cost of sales) of GBP19.97M or 72.9% of revenue, gross profit GBP7.4M (27.1% margin, +100bps), adjusted EBITDA -GBP0.3M, loss before tax GBP0.8M, cash GBP5.8M with no debt; 78% of revenue was generated in the US, impressions fell 14.1% to 39.1bn while revenue per click rose 7.4% to 58p. | 4 | direct record | unresolved | high | [86] | | C-F23 | Dianomi returned to growth in H1 2026 (revenue GBP13.4M, +2%, +4.5% constant currency; gross margin 28.9% vs about 25.5%; EBITDA loss GBP0.1M vs GBP0.6M; cash GBP6.0M) with impressions up 10% to 22.6bn from expanded CNN and Associated Press placements and July-August revenue up 14% YoY, while attributing the 14% impression decline of 2025 partly to 'AI-generated and so-called zero-click summaries'. | 4 | direct record | unresolved | medium | [87] | | C-F26 | Taboola is diversifying its supply monetisation beyond recommendation units: on 19 Aug 2026 it announced that Realize will power 'global programmatic display advertising' on NBCNews.com and TODAY.com, described as 'the first of its kind for Taboola', and its Q2 2026 release credited momentum to 'Realize, the addition of Fox News and other strategic wins'; its CEO separately claimed Taboola 'paid over $1.5 billion to publishers last year', which exceeds FY2025 reported TAC of $1,214.9M. | 4 | vendor assertion | unresolved | medium | [50, 76, 81, 1] | | G-F03 | The joint ANA/4A's/WFA/ISBA definition (26 Sep 2023) makes 'high percentage of paid traffic sourcing' a core MFA characteristic and, in the ANA report's elaboration, explicitly names 'content recommendations platforms' alongside social networks as the clickbait-ad channels that source MFA visits. | 4 | direct record | supported as stated | high | [88, 89] | | G-F16 | In the Forbes case, bid requests routed via Media.net's Prebid server to Magnite, PubMatic, Xandr and TripleLift truncated the page URL and rewrote the subdomain from www3.forbes.com to www.forbes.com, and IAS/Moat/DoubleVerify publisher tools deployed on www.forbes.com were not observed on www3, so buyers' verification did not flag it; agencies including Goodway Group paused Forbes buys. | 4 | independent measurement | supported with qualification | medium | [92, 93] | | C-F02 | Taboola's TAC equalled 63.5% of FY2025 gross revenue (59.3% in 2022). TAC is mostly, but not only, payments to publishers: it also includes programmatic inventory bought on exchanges, up-front and incentive payments, and $16.4M of non-cash Yahoo-asset amortisation. It should not be called a pure 'publisher payout ratio'. The trend since 2024 is also affected by a change in how some Yahoo-related amounts are presented (revenue versus a TAC offset). | 4, 10 | synthesis | supported with qualification | high | [1, 77] | | G-F01 | In the ANA's 2023 transparency study (21 marketers, $123M, 35.5B impressions, Sep 2022-Jan 2023), made-for-advertising websites accounted for 21% of impressions and 15% of spend, with every advertiser exposed (range 0.13% to 42% of spend). | 4, 10 | independent measurement | supported as stated | high | [89] | | G-F04 | MFA sites pass standard buyer quality screens: the ANA reports they show high measurability, good viewability, low IVT, brand-safe environments and CPMs 25% below non-MFA sites, while Jounce Media estimates their ads are at least 50% less likely than the internet average to be attributed with a sale. | 4, 10 | synthesis | unresolved | medium | [89] | | G-F07 | Among ANA benchmark participants, MFA share of spend fell from 15% (2023 study) to 6.2% in 2024 according to the Dec 2024 release, although the ANA's own Q1 2025 report restates the 2024 value as 1.1%. It then sat at 0.4-0.6% through 2025 in the sequential cost waterfall (0.4% in Q1, 0.5% in Q3). The median fell from 10% (2023) to 1.1% (Q4 2024), rose to 2.3% (Q1 2025), then fell to 0.8% (Q2) and 0.39% (Q3 2025). The ANA says waterfall, average and median values are not comparable. | 4, 10 | independent measurement | supported with qualification | medium | [94, 95, 96, 97] | | G-F08 | Dispersion persisted even as medians fell: in Q3 2025 the top quartile of ANA benchmark marketers still had 3.3% to 27.4% of web spend on MFA, after up to 28.7% in Q2 2025 and a bottom-quartile high of 25% in Q1 2025. | 4, 10 | independent measurement | supported as stated | high | [97, 96, 95] | | G-F11 | MFA spend among ANA benchmark participants rose from 0.6% in Q4 2025 to 1.1% in Q1 2026 (2.1% for lower-performing vs 0.9% for higher-performing advertisers), with 'AI slop' cited as an emerging sub-type that SSPs say may not meet all three classic MFA criteria. | 4, 10 | independent measurement | contradicted | medium | [98, 742] | | G-F15 | Adalytics (Mar 2024) reported, from Similarweb estimates, that www3.forbes.com sourced more than 70% of its readership from paid display ads on Taboola, Outbrain and other paid traffic-acquisition sources. In one observed session, a 52-slide slideshow served 201+ ads, against about 3-10 ads on a normal www.forbes.com article. The subdomain had served ads since at least 3 May 2017 and became inaccessible on 2 April 2024, after the WSJ contacted Forbes. | 4, 10 | independent measurement | supported with qualification | medium | [92, 93] | | D-F02 | Disclosure position mattered in the opposite direction to regulatory guidance: a mid-article disclosure raised the odds of recognition 5.1x relative to the top-of-page position (Study 1), and in Study 2 90% of readers fixated on a mid-article disclosure versus 40% for a top disclosure (odds ratio 12), with all 11 recognisers having fixated on the label. | 5 | independent measurement | supported as stated | medium | [99] | | D-F03 | Disclosure wording drove recognition: 'advertisement' (12%) and 'sponsored content' (13%) produced far higher recognition than 'brand-voice' (2%) or 'presented by [sponsor]' (3%), odds ratio 1:0.121 for the vaguer wordings. | 5 | independent measurement | supported as stated | medium | [99] | | D-F04 | Readers who recognised the sponsored article as advertising rated it less credible (M 4.11 vs 5.03 on 7-point), liked the sponsor less (4.67 vs 5.67), judged story quality lower (4.24 vs 5.31) and were less willing to share (3.29 vs 4.65) in Study 1; in Study 2 only the credibility penalty replicated (4.53 vs 5.35). | 5 | independent measurement | unresolved | medium | [99] | | D-F06 | Shown in isolation, labels containing 'paid' were read as advertising by 83-89% of respondents ('Paid Ad' 89%), 'sponsored' labels by 76-79%, while industry neologisms scored 57-64% ('Brand Voice' 64%, 'Presented By' 60%, 'Partner' 57%) and 'Written By' only 23%; in context, however, 'Brand Voice' content on Forbes was recognised by just 21-38%. | 5 | independent measurement | unresolved | high | [102] | | D-F07 | Adding a prominent 'Paid Ad' bar to two real native ads (inserted above the text on Forbes/Fidelity; replacing the small 'Sponsor Content' label on Vanity Fair/Hennessy) raised recognition from 40% to 56% (t=7.31); 33% still judged the content unpaid and 11% did not know. | 5 | independent measurement | supported with qualification | high | [102] | | D-F08 | The FTC's own exploratory lab study (48 participants, 2014-2015) found that applying its recommended disclosure fixes raised the probability of recognising an ad by 21 percentage points overall (about 47% to 68%; 95% CI 15-27), by 23 points for native-ad conditions (95% CI 14-32), with similar gains on desktop (21) and mobile (22), while 'a significant percentage of participants still did not recognize' the ads. | 5 | independent measurement | supported with qualification | medium | [103] | | D-F09 | FTC eye-tracking showed labels in the top-right corner of a page or content-recommendation widget were rarely fixated: on the Time mobile page 'only a handful' looked at the 'Sponsored Content' label in the widget's upper-right corner while 'most' attended to the 'Around the Web' label at upper-left, and moving/centring the label and switching to 'Paid Content' left 'only a few' failing to recognise the ADT ad; improved disclosures also cut time spent looking at ads by 21% overall (26% for native, not statistically significant). | 5 | independent measurement | unresolved | low | [103] | | D-F10 | The FTC's 2015 business guidance sets a performance standard for native disclosure - 'do consumers recognize the native ad as an ad?' - recommending 'Ad', 'Advertisement', 'Paid Advertisement' or 'Sponsored Advertising Content', discouraging 'Promoted'/'Promoted Stories' as 'at best ambiguous', requiring placement 'in front of or above the headline', and rejecting a single disclosure for a mixed grouping of paid and unpaid items in a recommendation widget. | 5 | direct record | unresolved | high | [2] | | D-F11 | A 2020 meta-analysis of 61 papers / 57 datasets / 473 effect sizes / 278,791 respondents found disclosing sponsored content raises recognition of the content as advertising (r=.255, k=50) and understanding of persuasive intent (r=.257, k=63), lowers brand attitude (r=-.108, k=89), credibility (r=-.132, k=82) and source evaluation (r=-.054, p<.10), and has no significant effect on behavioural intention (r=-.023, k=66, n=137,601) or attitude toward the ad. | 5 | independent measurement | supported as stated | high | [108] | | D-F12 | Meta-analytic moderators: the word 'advertising' in a disclosure increased both recognition and credibility; visual disclosures outperformed audio; disclosures after the content decreased brand evaluation and credibility while before/during placement was less negative; effects were generalisable across online/offline media and student/non-student samples; adults showed more recognition and more negative brand evaluation than minors; and brand-evaluation penalties became less negative over the study period. | 5 | independent measurement | unresolved | medium | [108] | | D-F13 | Mechanism: in the meta-analytic path model, recognition-as-advertising drives brand memory (total effect .363) but not brand attitude, whereas understanding of persuasive intent drives the brand-attitude penalty (total effect -.326); a 2023 experiment (N=133) adds an opposing path in which disclosure raises perceived transparency, lowers persuasion knowledge and increases brand attitude and purchase intent. | 5 | synthesis | unresolved | medium | [108, 112, 109] | | D-F14 | In a YouGov-representative sample of 800 US adults, only 9% of the 738 shown a sponsored article recognised it as advertising; high-prominence labels raised the odds 1.97x, high and medium language explicitness 3.66x and 3.01x, and a sponsor logo 1.64x (p<.10); more education increased and older age decreased recognition. | 5 | independent measurement | supported with qualification | high | [100, 101] | | D-F15 | Recognising the sponsored article lowered perceived publisher credibility (M 3.93 vs 4.46 for non-recognisers and 4.39 for a display-ad control, p<=.001 and p<.05) and attitude toward the publisher (4.09 vs 4.49, p<.05), with no difference between legacy (NYT/WSJ) and digital-only publishers. | 5 | independent measurement | supported as stated | medium | [100] | | D-F16 | Alternatives to text labels: a companion banner for the same brand raised recognition of article-style native ads to the same degree as a traditional disclosure across two experiments, and negative reactions to recognition were muted when sponsorship transparency was perceived as high, for familiar and unfamiliar brands alike. | 5 | independent measurement | unresolved | low | [113] | | D-F17 | The only located randomised online-plus-field study comparing in-feed native with display on a news page (Aribarg & Schwartz, JMR 2020, abstract only) found native earned a higher CTR at the same position but less visual attention, brand recognition and website trust. A larger randomised field experiment on search-native ads (Sahni & Nair 2020, >200,000 users) found no evidence of deception. | 5 | independent measurement | supported with qualification | medium | [115] | | D-F18 | In a 2025 German eye-tracking study of a mock Instagram feed (N=152; 8 sponsored, 21 organic posts), organic posts received 25% more fixations than sponsored posts (eta-squared .404) and about 250 ms longer dwell; where users fixated early on a disclosure or call-to-action, 80.6% of significant tests showed reduced subsequent dwell, and participants reported classifying ads mainly by product visibility (178 mentions), logos (81) and CTA buttons (77) rather than by the 'sponsored' label. | 5 | independent measurement | contradicted | medium | [116] | | D-F20 | In a January 2020 crawl of 6,498 mainstream news and 1,055 misinformation sites, 44.6% of 2,419 manually coded ads on 300 sampled sites were 'problematic' (content farms, supplements, insurance/mortgage/product advertorials, investment pitches, sponsored search, misleading polls); 87% of loaded native ads were problematic versus 20% of display ads; Taboola served 61.1% of all problematic ads and 85.7% of Taboola's coded ads were problematic (659 of 769), while 16.9% of Google-served ads were problematic (225 of 1,332), Zergnet was 100% content farms and 57.6% of RevContent ads were supplements. | 5 | independent measurement | supported with qualification | medium | [119] | | D-F21 | In a US Prolific survey (n=1,025, Aug-Sep 2020) rating 500 real web ads, 45% of ads were given a negative label by a majority of raters, 20.6% were majority-labelled 'clickbait' and 11.2% 'deceptive'; the four lowest-rated clusters (mean 2.21-2.8 on a 1-7 scale) were majority-labelled clickbait (61-68%) and 43-72% of ads in those clusters were native/content-recommendation ads. | 5 | independent measurement | unresolved | medium | [120] | | D-F22 | Political clickbait rides recommendation units: across 1,402,245 ads crawled on 745 US news/media sites (26 Sep 2020-19 Jan 2021), 3.9% carried political content, 52.0% of political ads promoted 'political news and media' articles, and 79.4% of those article ads were served by the content-recommendation network Zergnet (19,690 ads, 1,388 unique), followed by Taboola (10.0%), Revcontent (5.7%) and Content.ad (1.8%); a 2024 replication on the same 745 sites (15,110 ads, 4 Oct-7 Nov 2024) again found a 'prevalence of clickbait political news ads' (315 political ads, about 2.1%). | 5 | independent measurement | supported with qualification | medium | [121, 122] | | D-F23 | Deceptive advertorial landing pages are a long-standing enforcement pattern: in April 2011 the FTC sued 10 operations running 'fake news' sites with fabricated reporters and network logos to sell acai weight-loss products, driven by 'attention-grabbing ads on search engines and high volume websites'; the defendants had 'collectively paid more than $10 million to advertise their fake news sites' and consumers paid $70-$100 per purchase. | 5 | direct record | unresolved | high | [123] | | D-F24 | Vendor-commissioned attention/trust evidence for native units is undocumented at the method level: an Outbrain-commissioned Lumen study (900+ consumers, UK/France/Germany, 2019) claimed native ads on 'premium' news sites were '+31%' more likely to be trusted, '+16%' clicked and '+18%' to lead to future purchase than social-media ads, and '62% easier to understand than display ads', with no baselines, definitions or method disclosed. | 5 | vendor assertion | supported as stated | low | [117] | | D-F25 | A 2015 vendor survey (Contently; about 509 US adults balanced to Census) found that for four of six real native ads a majority read the piece as an article rather than an ad, 48% said they had felt deceived on realising content was sponsored, and 62% said a news site loses credibility when it publishes native ads; Hyman et al. summarise the same survey as 20-71% recognition by example. | 5 | commissioned measurement | unresolved | low | [118, 102] | | G-F25 | US regulatory authority over deceptive advertorial landing pages rests on the FTC's December 2015 Enforcement Policy Statement and Native Advertising Guide, which cite earlier cases where consumers reached 'fake news websites by clicking on ads presented as attention-getting news headlines, which frequently appeared on legitimate news websites', treat recommendation-widget links formatted as headlines as requiring clear disclosure, call 'Promoted' labels ambiguous, and extend liability to ad agencies and affiliate networks. | 5 | direct record | unresolved | high | [125, 2] | | D-F26 | As of 2026 the literature has two systematic reviews but no native-specific meta-analysis: Gupta, Halder & Narang (IJCS, March 2026) review 113 articles and organise them as individual, advertisement and disclosure characteristics driving recognition, which drives cognitive, attitudinal and behavioural outcomes; Huebner, Henseler & Thalmann (JMC, April 2026) review social-media native perception; and for AI answer interfaces the only located work (Webis, 2024) is computational - sentence transformers detect LLM-generated native ads with precision and recall above 0.9 while 'the investigated LLMs struggle', with no human recognition study reported. | 5, 8 | synthesis | unresolved | medium | [110, 111, 251] | | E-F01 | Taboola's FY2025 10-K states that its technology ranks each candidate ad by combining the predicted probability that the user will interact, the predicted probability of conversion after click/view, and the advertiser's bid into a 'relative value', and that its algorithms run auctions across CPC and CPM pricing models simultaneously. | 6 | direct record | supported as stated | high | [1] | | E-F02 | Taboola's FY2025 10-K states 500,000 recommendation-related requests per second (a figure repeated unchanged in every 10-K since FY2022, while the other scale metrics rose), over 600 million daily active users, up to 1.2 trillion recommendations served monthly (a peak, not an average) across seven front-end data centres, and four back-end data centres processing over 170TB of data per day to train its AI engine. | 6 | direct record | supported with qualification | high | [1, 126] | | E-F03 | In 2022 Taboola described serving more than a billion requests a day with hundreds of CTR-prediction models deployed daily on distinct traffic segments; each teacher model trained from scratch for ~2.5 hours on ~100 million impressions from the prior 14 days, while incremental 'student' models fine-tuned on ~12 million fresh samples in ~12 minutes (12.5x speedup) and were redeployed every 4 hours (6 cycles/day). | 6 | vendor assertion | unresolved | medium | [127] | | E-F04 | Taboola's knowledge-distillation training pipeline produced RPM lifts of 0.53% to 0.85% across four major traffic segments in a ~6-month A/B test, and the full warm-start-plus-distillation pipeline produced a 0.61% RPM lift in a ~2-month A/B test, with the paper noting warm-start alone caused an RPM regret. | 6 | vendor assertion | supported as stated | medium | [127] | | E-F05 | Taboola's 2018 Deep Density Networks paper states the recommendation engine must respond within strict time constraints (<50 ms), defines the objective as RPM = CTR x CPC x 1000 under fixed-CPC advertiser payment, splits the algorithm into exploitation and exploration modules to handle tens of thousands of new candidate recommendations per day, and reports a 2.9% online RPM lift for DDN over the tuned regression baseline plus 6.5% more new targets and 2.1% more advertisers discovered by exploration at a 0.05% RPM cost. | 6 | vendor assertion | unresolved | medium | [128] | | E-F06 | Zemanta (Outbrain's DSP) reported receiving over a million bid requests per second in 2021 and 'millions' in 2022, a maximum allowed auction response time of 100 ms, DeepFM-family CTR models computing over 300 million predictions per second (2021) and over 600 million per second (2022), served on CPUs inside a monolithic bidder with in-process autobatching that halved TensorFlow CPU usage at a cost of about 5 ms average latency and <0.01% timeouts. | 6 | vendor assertion | supported as stated | medium | [129, 130] | | E-F07 | In a multi-day online A/B test (hundreds of millions of impressions per arm), uncertainty-directed exploration raised revenue 5.0% and CTR 9.6% versus control, and random exploration raised revenue 3.2% but cut CTR 2.3%. In a subsequent offline evaluation of the final models on later data, AUC rose 0.24% and log loss fell 0.26% for the uncertainty arm, while random exploration raised AUC 0.19% and worsened log loss 0.8%. Revenue lifts are reported only 'where business constraints were respected'. | 6 | vendor assertion | supported with qualification | medium | [130] | | E-F08 | Adding a learned soft frequency-capping feature to OFFSET produced a 7.3% revenue lift in an online bucket test (CIKM 2019). The abstract does not say that the rule-based hard cap was removed. | 6 | vendor assertion | supported with qualification | medium | [131, 132] | | E-F09 | Post-auction creative and asset selection is a documented, separate stage: Yahoo's Carousel Asset Optimization ranked carousel cards by CTR with successive elimination and produced +8.6% CTR and +4.3% CPM/revenue versus control (KDD 2019), and its conversion-based dynamic creative optimization used an auxiliary OFFSET CVR model to weight title/image combinations, producing a 53.5% CVR lift versus a control that served combinations uniformly at random (IEEE Big Data 2022). | 6 | vendor assertion | supported with qualification | medium | [133, 134] | | E-F10 | Clicks are a noisy objective in native units: Verizon Media reported that accidental short-dwell clicks mislabel training data, that simply removing them causes under-prediction, and that soft re-weighting of such clicks yielded a 1.18% revenue lift in an online A/B test (CIKM 2021). | 6 | vendor assertion | unresolved | medium | [135] | | E-F13 | Realize recommends, but does not say it defaults to, Maximize Conversions. Realize's own help centre does state pricing rules: Enhanced CPC reduces winning CPC bids to the second-highest bid (a second-price rule), Creative Expedition charges only the second price even when bids are raised up to 7x, Fixed Bid charges the entered bid, and inventory sold programmatically through Realize clears first-price in PMPs and second-price on the open exchange. | 6 | vendor assertion | supported with qualification | high | [136] | | E-F14 | Realize's Maximize Conversions has a 2-5 day learning phase, requires at least 50 conversions in 7 consecutive days before a target CPA can be applied, recommends a daily cap of at least 10-15x expected CPA (or $50 if CPA is under $5), and without Taboola Pixel or server-to-server tracking optimises only toward page views and impressions. | 6 | vendor assertion | supported as stated | high | [145] | | E-F15 | Outbrain's Conversion Bid Strategy (automatic opt-in when Sales, Leads or App Installs is the campaign goal, per an Outbrain blog; the help pages describe selecting a mode under the 'Conversions' or 'App Installs' objective) offers four modes. Three of them (Max Conversions, Target CPA, Target ROAS) work by adjusting the campaign CPC, and Semi Manual routes traffic without adjusting CPC. Target CPA aims for 7-10 conversions a day, expects daily spend of at least 7x target CPA, has a learning phase of up to a week and asks for a 48-hour wait. | 6 | vendor assertion | supported with qualification | high | [137, 138, 139] | | E-F16 | Outbrain's Engagement Bid Strategy optimises to four on-site metrics (Max Clicks, Max Pages Per Session, Max Session Duration, Min Bounce Rate) using Google Analytics 4 data filtered by Outbrain's click ID plus contextual signals, requires no pixel, and may raise the CPC up to 3x the entered bid. | 6 | vendor assertion | unresolved | high | [140, 141] | | E-F18 | MGID's CPA Tune (announced 2025-06-25) adds Target CPA and MaxConversions strategies that keep CPC pricing while the algorithm 'decides which clicks to buy based on predicted conversion probability'; learning takes 7-14 days or 10-30 conversions, an exploration phase bids higher until the first conversion or 3x tCPA spend, daily budgets should be 5-7x tCPA (about 200x for Search Feed campaigns), and MaxConversions may exceed tCPA by 2x (product) or 1.5x (search feed). | 6 | vendor assertion | supported as stated | medium | [142, 143, 144] | | E-F19 | MGID's Performance Prediction scores a creative before launch as Poor/Medium/Good/Excellent from its title, image and campaign targeting settings, with the forecast considered valid for up to 30 days; no accuracy figures are published. | 6 | vendor assertion | unresolved | high | [155] | | E-F20 | Teads launched Teads Conversions in October 2023 after a beta of 150+ campaigns in 35 markets in which it says client CPA goals were met in 75% of campaigns; by November 2025 it marketed a 'Predictive AI' creative pre-test built on eye-tracking and EEG data, and its 2026 Ad Manager page lists 'Predictive AI CPC Prediction', 'Predictive AI Post-Click Engagement' and '4 Billion Signals/minute'. | 6 | vendor assertion | unresolved | medium | [156, 157, 158] | | E-F21 | Taboola's generative creative stack is documented as GenAI Ad Maker (image-to-image, text-to-image, background replacement and title/description generation using 'Chat GPT and Stable Diffusion' plus a proprietary framework trained on Realize best practices and network trends; help page updated 2025-05-22) and, from 2026-04-23, Realize+ in beta with a Decision Engine/Budget Allocator that reallocates budget in real time and an Element Generator that creates and revises ads and targeting, alongside an opening to Claude Skills. | 6 | vendor assertion | unresolved | high | [151, 152, 1] | | E-F22 | Taboola's 2026-01-28 press release reports a raw CTR of 0.76% for AI-generated ads versus 0.65% for human-made sibling ads, but says the two 'performed comparably' under the study's tightest statistical controls. It asserts, without figures, that AI visuals did not reduce downstream conversion performance. The underlying paper was not located. | 6 | vendor assertion | contradicted | low | [159] | | E-F23 | Taboola reports that its widget must return recommendations 'in a matter of hundreds of milliseconds' with strict database-query limits (2025), and that its TRECS rendering engine cut its scripts' main-thread blocking time by 485 ms (-70%) and improved publisher INP at p75 by 6-36% across four publishers (e.g., 75 ms to 48 ms) without negative impact on ad CTR or RPM (2024); a broken feature caught by its Sherlock skew detector, when fixed, lifted global RPM 1.31%. | 6 | vendor assertion | unresolved | medium | [161, 126] | | E-F24 | OpenRTB 2.6, the transport for programmatic native, signals the auction rule per request via the 'at' attribute (1 = First Price, 2 = Second Price Plus; default 2; exchange-specific values from 500) and the bidder time budget via 'tmax' in milliseconds, with native markup carried under the companion Dynamic Native Ads API. | 6 | direct record | unresolved | high | [150] | | E-F26 | Taboola's and Outbrain's published target-bidding rules imply roughly 5-10 conversions a day and daily budgets of 7-15x target CPA. MGID's are lower: 10-30 conversions cumulatively over 7-14 days, with 5-7x tCPA budgets (search feed about 200x). Target-CPA and target-ROAS modes are therefore hard to use for low-volume or high-CPA advertisers, but conversion-labelled modes without a target (Realize Maximize Conversions, Outbrain Max Conversions/Semi Manual, MGID) have no such entry thresholds. They run on proxies or network learnings instead. | 6 | synthesis | supported with qualification | medium | [145, 137, 139, 143, 144] | | E-F27 | Inference: the documented systems share a six-stage architecture — candidate retrieval, pCTR model, pCVR model, expected-value ranking with bid, post-auction creative/asset selection, and an exploration module fed by incremental training — and this architecture is established as live for Taboola (10-K plus papers), Zemanta (papers, 2021-22) and Yahoo (papers, 2019-22), but only as of the dates of those documents. | 6 | inference | unresolved | medium | [1, 127, 128, 129, 130, 131, 133, 134] | | F-F01 | Across 25 large US field experiments (millions of customers, USD 2.8 million of digital ad spend), the median confidence interval on advertising ROI was over 100 percentage points wide, and informative experiments 'can easily require more than 10 million person-weeks'. | 7 | independent measurement | supported as stated | high | [162] | | F-F03 | Using 15 US Facebook advertising experiments (500 million user-experiment observations, 1.6 billion impressions), observational methods 'often fail to produce the same effects as the randomized experiments, even after conditioning on extensive demographic and behavioral variables'. | 7 | independent measurement | unresolved | high | [163] | | F-F05 | The stream summary ('activity bias inflates exposed-vs-unexposed comparisons by up to 350%') and table F-T3 ('up to 350% overstatement' under exposed-vs-unexposed matching) misstate the design: the 350% is a before-after comparison within the treated group in one experiment, not an exposed-vs-unexposed comparison and not an upper bound. The finding text itself is correct. | 7 | independent measurement | supported as stated | high | [165] | | F-F06 | Google's 'ghost ads' methodology identifies control-group counterparts of exposed users inside randomized experiments; in one retargeting campaign it measured +17.2% website visits and +10.5% purchases, and lets advertisers 'measure ad lift just as precisely while spending at least an order of magnitude less' than PSA-style controls, with over 100 million predicted ghost ads logged daily. | 7 | independent measurement | unresolved | medium | [177] | | F-F07 | eBay's 2012 US experiments found brand-keyword paid search had no measurable short-term benefit and non-brand keywords negative average returns, with returns 'a fraction of conventional non-experimental estimates'; new and infrequent users responded to ads while frequent users, who generated most of the spend, did not. | 7 | independent measurement | supported as stated | high | [175] | | F-F08 | A randomized retargeting experiment with an online home-improvement retailer found retargeting caused 14.6% more users to return within four weeks, that 33% of the first week's advertising effect occurred on the first day, and that week-2 ads were more effective when the user had received week-1 ads. | 7 | independent measurement | supported as stated | high | [173] | | F-F09 | In an online travel firm's field experiment, dynamic retargeted ads showing previously viewed products were on average less effective than generic brand ads, and stopped underperforming only when browsing behaviour (e.g., visiting review sites) indicated narrowed preferences. | 7 | independent measurement | unresolved | medium | [174] | | F-F10 | For 288 US CPG brands, TV advertising elasticities were substantially smaller than the published literature, a sizable share were insignificant or negative, marginal ROI was negative for more than 80% of brands and overall ROI of the observed schedule positive for only one third, which the authors attribute partly to publication bias in prior studies. | 7 | independent measurement | supported with qualification | high | [176] | | F-F12 | Google Ads' default click-through conversion window is 30 days (configurable 1-90 days by campaign type), the default view-through window is 1 day (configurable 1-30 days), and the default engaged-view window is 3 days; view-through conversions count only where at least 50% of a Display ad was on screen for at least 1 second, exclude users who interacted with any other ad, are reported outside the main 'Conversions' column, and cannot be reported for browsers that block cross-site cookies. | 7 | direct record | supported as stated | high | [178, 179] | | F-F13 | Meta deprecated the 7-day view (7d_view) and 28-day view (28d_view) attribution windows in the Ads Insights API effective 12 January 2026 (announced on Meta's Developer Blog on 16 October 2025), leaving 1-day click, 7-day click, 28-day click, 1-day engaged view and 1-day view windows. | 7 | synthesis | supported with qualification | medium | [180] | | F-F15 | Google's Meridian documentation states 'there is no single formula to translate an experiment result into a prior', warns that 'results from a different time period might not be directly applicable', that short experiments may miss long-term effects, and that 'the MMM counterfactual is zero spend, whereas some experiments might define ROI against a different baseline, such as reduced spend'; its CalibrationBuilder adds uncertainty for older and shorter experiments. | 7 | direct record | supported as stated | high | [187, 188] | | F-F16 | Robyn's docs cite an Analytic Edge whitepaper for a '25% average difference to the ground truth' in uncalibrated models, and a simulation by Robyn 2022 Hackathon winners (two simulated channels) for accuracy that 'improves strongly with up-to 10 studies per channel'. | 7 | vendor assertion | contradicted | medium | [189] | | F-F17 | IAB and IAB Europe (Nov 2025) define incrementality as 'the causal impact of marketing by identifying the additional business outcomes directly driven by a campaign or tactic, compared to what would have occurred in the absence of marketing activity', state that 'incrementality differs from attribution and ROAS: those methods show what happened, not whether marketing caused the result', and rate experiments 'Strong', model-based counterfactuals 'Strong to Moderate', MMM 'Moderate to Weak' and hybrid proxies (new-to-brand percentage, baseline-vs-exposed analysis, platform-reported incrementality, simple MTA) 'Weak' on causal strength. | 7 | direct record | supported as stated | high | [171] | | F-F18 | The IAB/MRC Retail Media Measurement Guidelines (Jan 2024) define incrementality as 'the potential causal impact of marketing', recommend RCTs 'when feasible', require that attribution/lookback windows be 'empirically supported' and 'disclosed before campaign execution', note typical digital windows are '3, 7, 14, 28, or 30 days', require day-level lag data (view/click to conversion) be made available, and define new-to-brand as a shopper with no purchase from the brand 'within a defined time frame that needs to be disclosed by the retailer'. | 7 | direct record | unresolved | high | [184] | | F-F19 | IAB's Modernizing MMM guide (Dec 2025) instructs marketers to 'use experiment results to calibrate models' but cautions that experiments must be 'matched carefully to MMM outcomes (same KPI, region, and time window)' and 'replicated or validated over time before being treated as reliable anchors', and describes market demand for 'weekly data refreshes and monthly model retrains'. | 7 | direct record | unresolved | high | [191] | | F-F20 | The 55% and 51% survey figures come from Kantar research commissioned by Meta (per the page); sample size and geography are not disclosed. | 7 | vendor assertion | supported with qualification | low | [190] | | F-F22 | Outbrain's 2018 'True Engagement' brand-lift case (Nielsen survey commissioned by Outbrain; 1,573 users surveyed June-July 2017; PHD Germany automotive client) reports +23% brand lift among all respondents, but its comparison groups were 'those who clicked on the page and those who did not', i.e., clickers versus non-clickers rather than a randomized holdout. | 7 | commissioned measurement | unresolved | low | [167] | | F-F23 | The release uses 'incremental' language in its headline and quotes but gives no incrementality test, holdout or measurement partner behind the 2.4x claim. | 7 | vendor assertion | supported with qualification | high | [168, 169, 170] | | F-F25 | Haus's Cyber Week 2025 analysis of hundreds of geo experiments reports that 41% of the incremental value of media appeared in the post-treatment window, total efficiency improved by more than 75% when latent BFCM impact was included (USD 1 iROAS becoming USD 1.75), efficiency doubled in 44% of experiments, evergreen post-treatment windows add only 26% more incremental value, and OTT/CTV showed a 344% median efficiency improvement with the post-treatment window. | 7 | vendor assertion | unresolved | low | [182] | | F-F26 | ANA's July 2024 retail media survey (fielded November 2023) found 55% of marketers cite lack of standardization across platforms as the greatest challenge, 48% attribution and 40% data timeliness, 71% report sales conversion as the biggest goal, and ANA is working with MRC on measurement 'must haves'; the widely repeated claim that '71% of advertisers rank incrementality as the most important retail media KPI' is not what this press release says. | 7 | direct record | unresolved | medium | [193] | | F-F27 | The IAB / BWG Global 'State of Data 2026' report (released 2 February 2026; 400+ US brand and agency decision-makers per IAB's summary) finds that three out of four marketers say their attribution, incrementality and MMM approaches 'aren't delivering the speed, accuracy or trust they need', and about half of buy-side marketers are already scaling AI within measurement. | 7 | synthesis | unresolved | medium | [192] | | H-F01 | Taboola's FY2025 10-K describes its e-commerce offering (via Connexity) as access for publishers to over 500 million structured product listings, monetised on a CPC or CPA basis, and positions it as a data asset that pairs readership with purchasing data. | 8 | direct record | supported as stated | high | [1] | | H-F02 | Taboola reported FY2025 revenue of $1,912.0M (+8.3%) and TAC of $1,214.9M; its reported ex-TAC Gross Profit (non-GAAP) was $713.5M (2024: $667.5M, +6.9%). Revenue minus TAC is $697.1M; the $16.4M gap is non-cash amortization of the Commercial agreement asset that Taboola adds back. The 10-K reports a single segment and no e-commerce revenue line. | 8 | direct record | supported with qualification | high | [201, 1] | | H-F03 | Taboola announced the Connexity acquisition on 2021-07-23 for about $800M (about $260M cash, $300M debt-financed cash and $240M in shares); Connexity reported 2020 revenue of $176M, ex-TAC gross profit of $78M and Adjusted EBITDA of $38M, with 1,600+ direct merchants and 6,000 publishers, and had acquired Skimlinks in May 2020. | 8 | direct record | supported as stated | medium | [195] | | H-F04 | Taboola Turnkey Commerce, launched 2023-02-14 with TIME (a managed offering in which Taboola supplies the editorial team, technology and affiliate-link monetisation for a publisher commerce section), was extended to The Associated Press's 'AP Buyline' on 2024-03-12; commercial terms are undisclosed and Taboola's cited scale ('more than 1 million monthly transactions', 'nearly 600 million daily active users') is unverified. | 8 | vendor assertion | unresolved | medium | [196, 197] | | H-F05 | Skimlinks (a Taboola company) states it affiliates commerce links across 48,500 merchants and serves 23 of the top 25 publishers worldwide; its support documentation states a typical 75/25 revenue share, 75% of the merchant's commission to the publisher. | 8 | vendor assertion | unresolved | low | [198] | | H-F07 | Teads' FY2025 10-K names retail media (alongside CTV, in-app and LLMs) as a planned high-growth focus environment and a growth-plan risk, but discloses no retail-media or commerce revenue, product or partner (Pentaleap is not mentioned); the Q1 2026 10-Q does not mention retail media or commerce at all and discloses revenue only by geography. | 8 | inference | contradicted | medium | [46, 202] | | H-F08 | On 2025-07-24 Pentaleap and Teads announced an RTB integration letting advertisers buy retailers' onsite Sponsored Product Ads through Teads Ad Manager; the release names no launch retailer, and by 2026-09-27 no follow-up (live retailer, volumes or results) was found on Pentaleap's news listing. | 8 | vendor assertion | supported as stated | medium | [203, 204] | | H-F09 | Teads/Outbrain accepts conversions four ways: browser pixel, a first-party server-to-server postback (ob_click_id plus event name, with optional order value, currency, order ID and timestamp), CSV offline import emailed to a service account keyed to the dicbo click ID (only conversions within 72 hours of the click are used for optimisation; later ones can still be reported), and since 2025-11-20 a free Shopify pixel app. Conversion windows are advertiser-set; the pixel page's '24 hours' matching statement should not be read as a hard attribution ceiling. | 8 | direct record | supported with qualification | high | [220, 221, 222] | | H-F10 | Taboola's server-to-server conversion path is a postback to trc.taboola.com keyed to the 'tblci' click ID captured from the landing-page URL (sent as 'click-id' with an event 'name'), with a bulk submission alternative; Taboola states S2S tracks attributed conversions only, which limits audience marking, retargeting and optimisation relative to the pixel. | 8 | direct record | supported with qualification | high | [226, 227, 228] | | H-F11 | Compared with Meta's Conversions API reference design (server events with event_id deduplication against the pixel, hashed customer identifiers such as em and ph, unhashed fbc/fbp click and browser IDs, and offline/CRM events usable for measurement and audiences), the native vendors' documented ingestion is narrower: a click-ID-plus-event-name postback (Taboola) or CSV import (Outbrain) tied to a single ad click. | 8 | synthesis | unresolved | medium | [229, 230, 226, 220] | | H-F12 | What advertisers can export from native platforms is aggregate reporting: Taboola's Backstage API offers campaign summary reports broken down by day, campaign, site (publisher), country, platform and item, and lets advertisers upload first-party audiences by email or device ID in 26 (email) or 24 (device) countries, while Outbrain's Amplify API exposes Marketer, Budget, Campaign, PromotedLink and 'PerformanceBy*' entities and is 'available for a select number of partners by request'; no documentation read describes exporting audiences, click-level logs or model learnings. | 8 | synthesis | unresolved | medium | [231, 232, 233, 234, 235] | | H-F13 | Amazon Marketing Cloud returns only aggregated, anonymous query outputs, states that an advertiser's own uploaded signals cannot be exported or accessed by Amazon, is offered at no cost to eligible advertisers (Amazon DSP users by request, sponsored-ads advertisers self-service), and sells paid features in beta. | 8 | direct record | unresolved | high | [214] | | H-F14 | Amazon Publisher Cloud, built on AWS Clean Rooms, lets publishers plan inventory packages, enrich first-party signals to unlock Amazon DSP demand and receive reporting, with Dotdash Meredith quoted as mapping 'intent-to-buy signals' across more than 1.5 million articles; this is the one documented route by which retail purchase signals reach open-web publisher inventory without data leaving Amazon. | 8 | vendor assertion | unresolved | medium | [208, 209] | | H-F15 | Amazon itself announced in August 2023 that Sponsored Products would run automatically, with no advertiser action and on existing CPC bids, on Pinterest, BuzzFeed, Hearst Newspapers, Raptive and Ziff Davis sites. In mid-2025, Amazon-ad vendors (Feedvisor, JumpFly) reported a new off-Amazon campaign setting with 'Maximize reach' as the default and 'Minimize spend' as the alternative; the default-on claim remains vendor-reported. | 8 | vendor assertion | supported with qualification | medium | [211, 212, 213] | | H-F16 | Criteo's Retail Media segment generated $263.9M revenue and $259.7M Contribution ex-TAC in 2025 (each +2% year on year), about 22% of group Contribution ex-TAC of $1,174.6M, with Q4 2025 down 17% (revenue $76.3M vs $91.9M) which Criteo attributes to scope changes with two clients; Criteo defines onsite as ads on retailer websites and offsite as 'across the open internet', serves about 235 retailers and 17,000 clients, became 'Google's first onsite Retail Media partner' in 2025, and does not disclose an onsite/offsite revenue split in the material read. | 8 | direct record | supported as stated | high | [205, 206] | | H-F17 | Criteo Commerce Grid, a commerce-focused SSP, advertises native among its supported formats (display, video, native, in-app, CTV), integration with 60+ third-party DSPs, and lets retailers activate first-party data across offsite inventory 'without the risk of data leaks'; Criteo claims $2.2B annual ad spend accessed and 17,000+ clients. | 8 | vendor assertion | unresolved | medium | [207, 206] | | H-F18 | People Inc. (renamed from IAC on 2026-06-04) reported Q2 2026 performance marketing revenue of $68.8M, up 13% from $61.1M, driven by 11% affiliate commerce growth on higher transaction volumes, equal to about 24% of digital revenue of $289.9M (+6%), even as Core Sessions fell 22% 'due primarily to the impact of the growing prominence of Google AI Overviews'. | 8 | direct record | supported as stated | high | [215, 216] | | H-F19 | BuzzFeed's commerce and other revenue was $56.5M in FY2025 (-8.3%), of which affiliate commerce was $55.5M (-6.9%), about 30% of total revenue of $185.3M; Q4 2025 affiliate commerce fell 22.7% to $16.1M, attributed to 'changes in supplemental bonus structures from our partners', and the 10-K states that approximately 28% of FY2025 revenue was derived from Amazon, primarily from affiliate commerce transactions. | 8 | direct record | supported as stated | high | [217, 218] | | H-F20 | Future plc's B2C eCommerce affiliates revenue fell 9% (4% organic) to GBP 76.7M in the year to 2025-09-30 (H1 +10%, H2 -22%). Future attributes the fall to audience and consumer confidence, plus FX and closures; full-year unique page views were down 13% and transactions down 6%. | 8 | direct record | supported with qualification | high | [219] | | H-F21 | Under the FTC Endorsement Guides revised effective 2023-07-26, a disclosure must be 'difficult to miss (i.e., easily noticeable) and easily understandable' and 'unavoidable' in interactive media; the Commission struck the word 'small' from the affiliate-link blogger example so that any affiliate compensation requires disclosure, and FTC staff guidance says 'affiliate link' alone is insufficient while 'paid link' next to the link would work, with the disclosure placed as close to the recommendation as possible. | 8 | direct record | unresolved | high | [236, 237] | | H-F22 | The FTC's Trade Regulation Rule on the Use of Consumer Reviews and Testimonials (16 CFR Part 465), announced 2024-08-14 by a 5-0 vote and effective 2024-10-21, prohibits fake or false reviews and testimonials, buying positive or negative reviews, undisclosed insider reviews, company-controlled review websites, review suppression and fake social-media indicators, and authorises courts to impose civil penalties for knowing violations; incentivised reviews remain lawful if not conditioned on sentiment, with incentive disclosure governed by the FTC Act and the Endorsement Guides rather than the Rule. | 8 | direct record | unresolved | high | [238, 239] | | H-F23 | Every commerce revenue line examined (People Inc. performance marketing, BuzzFeed affiliate commerce, Future eCommerce affiliates, Taboola/Connexity CPC-CPA) is booked as attributed commissions or attributed conversions, and none of the native vendor conversion documentation read describes holdout or incrementality measurement; Taboola's S2S doc explicitly limits S2S to 'attributed conversions only' and Outbrain optimises only on conversions within 72 hours of a click. | 8 | inference | unresolved | medium | [218, 215, 219, 1, 228, 220] | | H-F24 | Publisher content-to-commerce economics are governed by affiliate-programme terms and traffic rather than by native ad technology: BuzzFeed's 2025 decline came from partner bonus-structure changes with about 28% of revenue dependent on Amazon, Future's H2 decline from audience loss, and People Inc.'s growth from transaction volume despite a 22% session decline. | 8 | synthesis | unresolved | medium | [217, 218, 219, 215] | | J-F15 | Google's ads on AI surfaces progressed from 'expanding ads in AI Overviews to desktop' and 'bringing ads to AI Mode' at Google Marketing Live 2025 (2025-05-21) to 'new ad formats built with Gemini for AI Search' that are 'instantly tailored to a person's unique query' plus an expanded 'Direct Offers pilot' at GML 2026 (2026-05-20); neither post gives adoption, share-of-queries or revenue figures. | 8 | vendor assertion | unresolved | high | [257, 258] | | J-F20 | OpenAI announced ChatGPT ads on 16 January 2026 and began testing them on 9 February 2026 with logged-in US adults on the Free and Go tiers. A reported OpenAI statement of 31 August to 1 September 2026 put ChatGPT Ads at a $1 billion annualised revenue run rate, reached in under 200 days; the revenue figure is the seller's own, via reporting. | 8 | vendor assertion | supported with qualification | medium | [242, 243] | | J-F21 | Perplexity's ads, launched in the US on 2024-11-12 as labelled 'sponsored follow-up questions' beside AI-generated answers, were phased out during 2025. Perplexity stopped taking new advertisers, and the FT reported on 2026-02-17 that it has no plans to return to ads. | 8 | synthesis | supported with qualification | medium | [248, 249] | | J-F22 | The Agentic Commerce Protocol (ACP) is co-maintained by OpenAI and Stripe, licensed Apache 2.0, labelled 'Beta', with spec releases dated 2025-09-29, 2025-12-12, 2026-01-16, 2026-01-30 and 2026-04-17; its ChatGPT surface, Instant Checkout, launched 2025-09-29 for US Free, Plus and Pro users with Etsy live and Shopify merchants to follow, merchants paying 'a small fee on completed purchases', and OpenAI stating results 'are not sponsored, and ranked on relevance alone' — i.e., a commerce rail adjacent to, not part of, advertising. | 8 | synthesis | supported as stated | high | [259, 260, 261, 262] | | G-F02 | The ANA's MFA measurement matched DSP domain logs against a DeepSee.io list of 4,500 known MFA sites; about 2,200 of those carried study spend and the top 500 MFA sites accounted for 98% of MFA spend, so exposure is concentrated in a few hundred domains. | 10 | independent measurement | supported as stated | high | [89] | | G-F13 | Jounce Media's market-wide estimates put MFA at roughly 5% of web auctions in early 2020, 15% of the ad-supported web in 2021, 20% in 2022 and nearly 30% of web auctions by mid-2023. | 10 | independent measurement | unresolved | medium | [89, 291] | | G-F18 | Taboola's MFA remedy is Taboola Select: placements on about 15% of its most premium publishers (2 Oct 2024 release). Taboola says it excludes MFA from Select using Jounce Media's detection data plus its own ML models. The commercial terms are undisclosed, and Taboola publishes no MFA-exposure figure for the rest of its network. | 10 | vendor assertion | supported with qualification | medium | [299, 300] | | G-F19 | Taboola/Realize landing-page policy (updated 27 Apr 2025) requires 'Advertorial' across the top of promotional pages that could be confused with editorial, 'Sponsored By [Company]' disclosure, accessible company contact details, no false-urgency timers, user-initiated video audio and disclosure of third-party ads on the landing page; the pages opened contain no numeric ad-density or refresh cap. | 10 | vendor assertion | unresolved | high | [302, 303] | | G-F20 | Teads' (Outbrain Inc., renamed Teads Holding Co.) advertiser guidelines impose numeric landing-page limits - no more than 5 ad placements and 1 video player per viewport and no ad auto-refresh above 3 times per minute - plus 'Advertisement'/'Advertorial' labelling, Contact Us and Privacy Policy links, and a ban on 'sites that seek to intentionally deceive the reader'. | 10 | vendor assertion | supported as stated | high | [301] | | G-F21 | On the supply side, Teads' publisher guidelines bar sites 'primarily designed to generate ad revenue with little to no original or meaningful content' and require traffic that is 'legitimate, human-driven', prohibiting incentivised clicks and bot-generating acquisition; the equivalent Taboola publisher policy was not opened. | 10 | vendor assertion | unresolved | medium | [304] | | G-F26 | The publisher-side dispute record is mostly commercial rather than quality-led: by January 2018 Outbrain and Taboola were replacing $1-2M annual guarantees with revenue-share or CPC deals (Conde Nast chose revenue share for placement flexibility ahead of paywalls), while Vice's January 2021 report on disinformation in Taboola widgets led Taboola to stop working with The Federalist. | 10 | synthesis | unresolved | medium | [79, 311] | | G-F27 | Jounce Media's 2024 'cheap reach' category - inventory with under 15% viewability such as below-the-fold recommendation widgets, comment sections and slideshows - names Outbrain and Taboola among operators, and Jounce dropped weather.com (52% cheap-reach bid requests) and dailymail.co.uk (22%) from its bellwether list; buyers called it 'a necessary evil'. | 10 | independent measurement | unresolved | medium | [312] | | I-F01 | Outbrain closed its acquisition of Teads on 3 February 2025 for approximately $900 million ($625 million cash plus 43.75 million Outbrain shares) and the combined business began operating under the Teads brand while still trading as Nasdaq: OB. | 10 | direct record | supported as stated | high | [45] | | I-F02 | The legal entity was renamed Teads Holding Co. and its shares began trading under the Nasdaq ticker TEAD on 10 June 2025, four months after the deal closed; the 10-K dates the name change effective 6 June 2025. | 10 | direct record | unresolved | high | [273, 46] | | I-F03 | Teads' FY2025 10-K describes managed and self-service buying (Item 1 names CPC and CPM), holding-company MSAs for Teads Ad Manager, and about 50 typically non-binding JBPs averaging $4M of annual spend. Its revenue-recognition note adds CPV and cost-per-incremental-action (CPA) pricing, with variable consideration tied to measured outcomes. 93% of 2025 revenue came directly from advertisers and the rest indirectly via DSPs. | 10 | direct record | supported with qualification | high | [46] | | I-F04 | In Q2 2026 Teads reported revenue of $284.6 million (down 17% year over year), Ex-TAC gross profit of $123.4 million (down 14%), Adjusted EBITDA of $7.0 million (down 74%) and a net loss of $42.5 million, and suspended full-year guidance citing volatility in its Direct Response and SME business while renewing enterprise JBPs with Stellantis, LVMH, Warner Brothers and Dyson. | 10 | direct record | supported as stated | high | [271] | | I-F05 | On 11 June 2026 Teads announced EngageOS, a publisher 'feed operating system' that places editorial recommendations and ads in one auction to optimise total session yield, with Magnite's Demand Server integrated so publishers can bring Prebid Server-compatible SSP demand into native placements; no adoption or lift figures were disclosed. | 10 | vendor assertion | unresolved | high | [265, 271] | | I-F06 | Omnicom completed its acquisition of Interpublic on 26 November 2025 (0.344 Omnicom shares per IPG share; legacy Omnicom ~60.6% / legacy IPG ~39.4% ownership; pro forma revenue above $25 billion), with John Wren as Chairman & CEO and Philippe Krakowsky and Daryl Simm as Co-Presidents and COOs. | 10 | direct record | unresolved | high | [274] | | I-F07 | At the cutoff Omnicom Media's agency roster is OMD, PHD, UM, Initiative and Hearts United, the last formed by combining Hearts & Science and Mediahub and officially launched on 28 August 2026 with approximately $9.1 billion of 2025 billings across 40 markets; the playbook's list of Hearts & Science and Mediahub as separate brands is stale. | 10 | direct record | supported as stated | high | [281, 282, 272] | | I-F08 | Omnicom Media self-describes $70 billion-plus buying power, 40,000-plus specialists in 70-plus markets, Omni as its unified data/AI platform and Acxiom Real ID as an identity layer (2.6 billion IDs), and claims $3.3 billion of new billings in H1 2026; Annalect is not named on the site. | 10 | vendor assertion | supported with qualification | medium | [282] | | I-F09 | GroupM was rebranded WPP Media on 28 May 2025, described by WPP as managing about $60 billion in annual media investment; Mindshare, Wavemaker and EssenceMediacom continue as agency brands, and by 2026 WPP Media's site lists those three plus Choreograph and The Goat Agency, with Open Media Studio inside WPP Open. | 10 | synthesis | unresolved | medium | [275, 276] | | I-F10 | Life360 completed its acquisition of Nativo on 5 January 2026 for approximately $120 million (65% cash, 35% stock), removing Nativo from the set of independent native platforms. | 10 | direct record | supported as stated | high | [277] | | I-F11 | Criteo completed its redomiciliation to Luxembourg on 29 Jul 2026 and plans a US domicile via cross-border merger, expected Jan 2027 and subject to shareholder approval. Its releases do not mention a sale. However, Bloomberg reported on 6 Jul 2026 (relayed by MediaPost) a takeover bid from Vista Equity Partners and Quinti Capital at a >50% premium, and Criteo declined to comment. The playbook's 'acquisition speculation, no buyer confirmed' therefore matches the public record at the cutoff. | 10 | direct record | supported with qualification | high | [284, 272] | | I-F12 | Reddit launched Max campaigns in beta for select advertisers on 5 Jan 2026 and opened them to all advertisers in early September 2026 (reported 3-4 Sep 2026), adding Ads API access, a Smartly integration and asset groups. Its 17% lower CPA and 27% more conversions come from 17 self-reported tests run Jun-Aug 2025. | 10 | vendor assertion | supported with qualification | medium | [279] | | I-F13 | Taboola announced Realize+ on 23 April 2026 as an agentic layer on its Realize performance platform after 'several months in beta', with phase 2 beta expanding in Q2 2026 and a first Claude Skill for campaign setup and optimisation; a Taboola-sponsored survey of 200 senior marketers at US and UK enterprises with 1,000+ employees found 80% would increase open-web investment if walled-garden-style automated AI solutions existed and 86% would allocate up to a quarter of performance budgets. | 10 | commissioned measurement | supported with qualification | medium | [153, 310] | | I-F15 | Taboola's FY2025 10-K reports about 2,200 Scaled Advertisers (more than $100K of trailing-four-quarter gross spend) in Q4 2025, working with it directly or through agencies. It states that in each of the last three years a significant majority of revenue came from Scaled Advertisers working directly rather than via an agency. It also reports ~14,000 digital property partners, top-ten advertisers under 10% of revenue with none above 3%, and CPC, CPM or CPA pricing. | 10 | direct record | supported with qualification | high | [1] | | I-F16 | The complete FY2025 10-Ks of Teads and Taboola contain no advertiser-facing outcome guarantee, make-good or published minimum spend. Their guarantee language covers minimum payments to publishers or media partners, and Teads has this as well as Taboola. Both filings do document performance-priced (CPA) campaigns, in which the advertiser pays only per acquisition or incremental action. Outcome-based commercial terms therefore exist as pricing models, not as guarantees. | 10 | inference | supported with qualification | medium | [46, 1, 266, 158, 267] | | I-F17 | The Trade Desk began rolling out Kokai Zuma to Kokai clients globally on 27 Aug 2026, with some capabilities still in open or closed beta. It reports that 'initial results' of the improved modelling showed an average 32% improvement in CPA performance, with no baseline or sample disclosed. | 10 | vendor assertion | supported with qualification | medium | [280] | | I-F19 | Vevo's 'Attention Guaranteed' (announced 27 January 2026 with Adelaide) guarantees a minimum Adelaide AU attention score for a campaign across CTV, mobile and desktop; it is an attention-threshold guarantee, not a CPA or sales-outcome guarantee, and no make-good or refund mechanism is disclosed. | 10 | vendor assertion | unresolved | medium | [307, 308] | | I-F20 | In the ANA Programmatic Transparency Benchmark for Q3 2025 (39 participating marketers, 21 contributing log-level data, 17.5 billion impressions), the median share of web ad spend on Made-for-Advertising inventory was 0.40% (down from 0.75% in Q2 2025) and the all-environment median was 0.39% (down from 0.8%), while the 2023 study had found MFA at 21% of impressions and 15% of spend for 21 advertisers. | 10 | independent measurement | supported with qualification | high | [97, 290, 743] | | I-F21 | 81.6% is the cohort-total (aggregate) PMP share of spend, and the median advertiser's PMP share is 88.8%. The 47.1% working-media figure is an average-based cost-waterfall value that excludes agency, ad-serving and managed-service fees, not a median. | 10 | independent measurement | supported with qualification | high | [97, 290] | | I-F23 | ANA contracting guidance (2016 principles; contract template v2.0, July 2018) prescribes disclosure of when an agency acts as principal rather than agent, affiliate definitions reaching the highest holding-company entity, caps on mark-ups for principal inventory (K2 Intelligence found 30-90% mark-ups), advertiser access to transaction data, treatment of 'value pots' as rebates and auditor NDAs as contract exhibits; the 4A's issued separate Transparency Guiding Principles in January 2016 that the ANA publicly called insufficient on audit rights. | 10 | direct record | unresolved | high | [305, 306] | | I-F26 | Publicis Groupe's UK media page lists Starcom, Zenith, Spark Foundry and Performics with PMX as the trading arm ('the second largest media buyer in the world'), and the CitrusAd brand now presents itself as having evolved into 'Epsilon Retail Media', a unified on-site and off-site platform; the playbook's 'CitrusAd (powered by Epsilon)' and 'Epsilon / CitrusAd + PMX' labels are directionally right but use the superseded brand. | 10 | vendor assertion | unresolved | medium | [286, 287, 272] | | I-F27 | Documented enterprise buying routes for open-web native at the cutoff are (a) managed service via vendor sales/strategic-account teams, (b) vendor self-serve buying interfaces (Teads Ad Manager under holding-company MSAs; Taboola Realize with an optional Claude Skill front end), (c) DSP/programmatic integrations, and (d) publisher-side native supply exposed to SSP demand (EngageOS/Magnite); published minimums exist only as third-party blog ranges, and MGID, Taboola and Teads product pages state no floor. | 10 | synthesis | unresolved | medium | [46, 1, 153, 265, 266, 158, 267, 268] | | I-F28 | The playbook's commercial arithmetic (120 qualified opportunities at a 20% close rate to 23 deals at $144-230K ACV yielding $3.3M-plus; $3.25M-$4.7M booked revenue per FTE; $2B-plus media at Hearts & Science; Verve Group 114% on-target in Q1 2024, LATAM 274%) is the author's illustrative operating model and personal track-record assertion, labelled as such on the page, and is not supported by any industry benchmark located in this stream. | 10 | illustrative scenario | unresolved | high | [272] | | K-F01 | Under the FTC's Enforcement Policy Statement of 22 December 2015 (approved 4-0), an advertisement's format is deceptive under Section 5 of the FTC Act if it materially misleads reasonable consumers about the ad's commercial nature or source, judged on the ad's overall net impression; the standard applies regardless of medium. | 10 | direct record | supported as stated | high | [125, 313] | | K-F03 | FTC guidance says everyone who participates directly or indirectly in creating or presenting native ads (which covers publishers, though the guide names only ad agencies and affiliate-network operators as examples) must avoid misleading consumers; the FTC has acted on this against a publisher at least once, in Inside Publications (complaint Nov 2018, final order Feb 2019). | 10 | direct record | supported with qualification | high | [2, 125] | | K-F05 | Lord & Taylor (complaint 15 March 2016; final consent order 23 May 2016) remains the FTC's landmark native-advertising case: a paid article in Nylon and roughly 50 paid Instagram influencer posts without disclosure, resolved by a 20-year order requiring clear and conspicuous disclosure and monitoring. | 10 | direct record | supported as stated | high | [315] | | K-F08 | TruHeight (Vanilla Chip LLC; complaint and proposed consent 13 April 2026, final order 15 July 2026) charged violations of both Section 5 and the Reviews and Testimonials Rule for website reviews written by employees and Amazon 5-star reviews incentivised with reimbursements or 10 percent discounts, showing the Rule being pleaded in 2026 enforcement. | 10 | direct record | supported as stated | high | [316, 317] | | K-F09 | The FTC Legal Library lists roughly ten review-related matters dated 2023-2026 (dates are 'last updated' dates). The Rytr order barring AI review-generation services (final December 2024) was reopened and set aside by a 2-0 Commission vote on 22 December 2025, citing the Administration's AI Action Plan. | 10 | synthesis | contradicted | medium | [318] | | K-F12 | Under the CCPA as amended, 'sharing' covers disclosing personal information for cross-context behavioural advertising. A business that sells or shares must honour valid opt-out preference signals and generally must also post the opt-out link; it may omit the link only if it processes signals in a frictionless manner and makes the required disclosures (11 CCR 7025(e)-(g)). The link is not an alternative to honouring the signal. | 10 | direct record | supported with qualification | high | [320, 321] | | K-F13 | California regulation section 7025 requires a business that sells or shares personal information to treat any compliant opt-out preference signal (for example an HTTP header or JavaScript object) as a valid opt-out of sale/sharing 'for that browser or device and any consumer profile associated with that browser or device, including pseudonymous profiles'; the updated regulation package was approved by OAL on 22 September 2025 and took effect 1 January 2026. | 10 | direct record | unresolved | high | [319, 322] | | K-F14 | AB 566, the California Opt Me Out Act (Chapter 465, approved 8 October 2025), makes it unlawful from 1 January 2027 for a business to 'develop or maintain a browser that does not include functionality configurable by a consumer that enables the browser to send an opt-out preference signal', which will put a universal opt-out control in mainstream browsers used for open-web native inventory. | 10 | direct record | supported as stated | high | [324] | | K-F15 | As of the IAPP chart dated 24 Nov 2025, 19 states had signed comprehensive privacy laws (17 with a profiling/targeted-ads opt-out). By the 27 Sep 2026 cutoff at least 21 had (adding Oklahoma, signed Mar 2026 and effective 1 Jan 2027, and Alabama, signed Apr 2026 and effective 1 May 2027); further 2026 enactments were reported but not verified. The statement 'none later than 1 Jan 2026' is no longer true. | 10 | synthesis | supported with qualification | medium | [325, 326] | | K-F16 | EU law blacklists undisclosed advertorials outright: UCPD Annex I item 11 ('Using editorial content in the media to promote a product where a trader has paid for the promotion without making that clear in the content or by images or sounds clearly identifiable by the consumer') and item 11a (undisclosed paid search results or paid ranking) are unfair in all circumstances, and Article 7(2) treats failure to identify commercial intent as a misleading omission. | 10 | direct record | unresolved | high | [329] | | K-F17 | Under the Digital Services Act, since 17 February 2024 all online platforms must ensure ads are 'clearly labelled' with information about 'who is placing the ad and why you are seeing it', may not target ads using profiling on sensitive data such as sexual orientation, religion or ethnicity, may not target ads at minors based on profiling, and very large platforms must maintain a public ad repository. | 10 | direct record | unresolved | high | [330, 331] | | K-F18 | Article 50 applied from 2 August 2026, except that the Article 50(2) machine-readable marking duty for generative systems already on the market before that date was deferred by the AI Omnibus to 2 December 2026. | 10 | direct record | supported with qualification | high | [336, 337, 338] | | K-F19 | The AI Omnibus postponed high-risk obligations to 2 Dec 2027 (Annex III) and 2 Aug 2028 (Annex I), added a ban on AI nudification/CSAM generation applying from 2 Dec 2026, and gave generative systems placed on the market before 2 Aug 2026 until 2 Dec 2026 to meet the Article 50(2) machine-readable marking duty; the other Article 50 duties applied from 2 Aug 2026. | 10 | direct record | contradicted | high | [337, 339, 340] | | K-F21 | UK CAP Code rule 2.1 requires that 'Marketing communications must be obviously identifiable as such' and rule 2.4 that 'Marketers and publishers must make clear that advertorials are marketing communications; for example, by heading them "advertisement feature"', with both rule 2.3 and 2.4 annotated as reflecting the statutory blacklist in DMCC Act Schedule 20. | 10 | direct record | unresolved | high | [341] | | K-F22 | The UK's DMCC Act 2024 Schedule 20 (in force 6 April 2025) makes undisclosed paid editorial promotion (paragraph 12) and fake or concealed-incentive consumer reviews (paragraph 13) unfair commercial practices in all circumstances, giving the CMA direct enforcement powers over conduct the ASA previously policed only through the CAP Code. | 10 | direct record | unresolved | high | [342] | | K-F23 | On 18 February 2026 the ASA upheld a complaint against Health Bridge Ltd t/a Zava and Mumsnet Ltd: an advertorial headed 'Medicated weight loss, an expert guide' with only a small 'Created by ZAVA' credit below the headline and no 'advertisement feature' label was not obviously identifiable as marketing (CAP Code 2.1, 2.3, 2.4) and also breached medicines rules 12.12 and 12.18. | 10 | direct record | supported as stated | high | [343] | | K-F24 | On 22 April 2026 the ASA upheld a complaint against a paid Google search ad for consumertestreports.org ('5 Best Cordless Vacuums 2026') because it presented an import-export trader's commercial site as 'an independent review organisation', breaching CAP Code 2.3 and 3.1, illustrating that UK recognition rules reach affiliate-style review sites, not only advertorials. | 10 | direct record | unresolved | high | [344, 345] | | K-F26 | Across the three jurisdictions the substantive rule for native placements is convergent (paid content must be identifiable as paid), while the enforcement architecture diverges: US case-by-case Section 5 actions plus civil penalties only for review/testimonial conduct under 16 CFR 465; EU per se blacklist (UCPD Annex I 11/11a) plus platform-level DSA duties on labelling, advertiser identity and targeting parameters; UK self-regulatory ASA rulings plus, since 6 April 2025, statutory CMA enforcement under DMCC Schedule 20. | 10 | inference | unresolved | medium | [125, 736, 329, 330, 341, 342] | | K-F27 | No rule found in this stream expressly governs paid placements inside AI assistant answers; the closest US authority is the 2015 Policy Statement's treatment of paid search results (disclosure of paid nature required), the closest EU authority is UCPD item 11a (undisclosed paid ranking) together with DSA labelling duties where the assistant qualifies as an online platform, and AI Act Article 50 addresses AI-generated content, not sponsorship. | 10 | inference | unresolved | medium | [125, 329, 330, 338] | | J-F01 | AdCP's GitHub marks v3.1.24 (2026-09-23) as the 'Latest' full release and the 3.2.0 line as pre-releases (rc.6 on 2026-09-24, rc.7 on 2026-09-27); the docs use 3.1.24 and the home page still announces 3.2.0-rc.3. Eight tags were cut 2026-09-09 to 2026-09-24 (nine by the cutoff day). | 11 | direct record | supported with qualification | high | [375, 376, 377] | | J-F03 | AgenticAdvertising.org asserts that 'the first-ever agent-to-agent media buy was executed' on 2025-10-16 with 'real inventory from LG Ads' and 'human oversight'; no independent observation of that transaction was found. | 11 | vendor assertion | unresolved | medium | [385, 375] | | J-F04 | AdCP owner surfaces do name native: the 3.1.24 docs refer to native formats, the catalog carries native_standard and native_content entries, and an open roadmap issue (2026-05-17) proposes a canonical 'native' format deferred to the 3.2 track pending an audit against OpenRTB Native 1.2. Native is therefore representable but not yet a canonical first-class format in AdCP. | 11 | inference | contradicted | medium | [377, 384, 390] | | J-F05 | The only open-source AdCP sell-side implementation opened, Prebid Sales Agent, describes itself as 'Status: alpha. A pre-1.0 implementation of a pre-release protocol (AdCP 3.1.1)', targets Google Ad Manager, Broadstreet, Kevel, Triton Digital and a mock ad server, and had 38 GitHub stars at access; AdCP's own docs call it a community example that is not officially maintained. | 11 | direct record | supported as stated | high | [382, 383, 377, 384] | | J-F06 | IAB Tech Lab named its agentic umbrella 'AAMP (Agentic Advertising Management Protocols)' on 2026-02-26 (the standards index still shows that as the 'Most Recent Release' date); AAMP 3.0 was announced on 2026-09-22, and its new OpenProposal specification is in public comment until 2026-10-22. ARTF is at v1.0 Public Comment; Agentic Audiences has no release listed. | 11 | direct record | supported with qualification | high | [378, 379, 380] | | J-F07 | The two protocol stacks do not reference each other on any page opened: the IAB Tech Lab AAMP page and hub repo do not mention AdCP or AgenticAdvertising.org, and the AdCP site, docs and README do not mention IAB Tech Lab or AAMP, although both cite MCP and A2A as transports. | 11 | inference | unresolved | medium | [379, 380, 375, 377, 384] | | J-F09 | Scope3 announced its rebrand to Apostra on 2026-09-23. It says the platform is built on AdCP and that campaigns are live in six named markets, and its own post gives no customer names or spend. Trade press (Adweek) reports a spend-scale figure ('around $1.2' followed by a truncated unit) that this verifier could not read in full. | 11 | vendor assertion | supported with qualification | high | [391, 390] | | J-F10 | Taboola introduced Realize in its Q4 2024 results release (2025-02-26) as 'our new independent performance platform that goes beyond search and social' reaching 'approximately 600 million daily active users', and by 2026-09 realize.com markets 'Realize+' as 'an agentic system that turns advertiser goals into outcomes' alongside 'Realize ID' and 'Performance AI', with formats listed as native, display, vertical, performance video, carousel and app promotion across '11,000+' publishers. | 11 | vendor assertion | unresolved | high | [351, 386, 393] | | J-F11 | Taboola's Q2 2026 results release (2026-08-05) mentions Realize only qualitatively and contains no statement about Abby, GenAI Ad Maker, Realize ID, Realize+ or AI-assistant distribution partners, while reporting Q2 2026 revenue of $476.8M (+2.4% YoY), ex-TAC gross profit of $192.4M (+11.8%) and Adjusted EBITDA of $55.5M (+22.8%), with FY2026 guidance of $1,930-1,956M revenue and $772-783M ex-TAC gross profit. | 11 | direct record | supported as stated | high | [30] | | J-F12 | Taboola announced Abby on 2024-10-15 (English press release; the German localisation is dated 2024-10-17). Abby is a conversational generative-AI assistant for campaign setup and creative edits. Taboola's only performance claim is that early-test campaigns were 'estimated to go live 75% faster' than manual setup, with no sample or method. | 11 | vendor assertion | supported with qualification | medium | [392, 393] | | J-F13 | MGID's MIA (launched 2026-07-16) is a diagnostic assistant that reads live account data and explains spend problems, creative rejections, account health and CPA Tune performance, but 'the decision to act remains yours'; MGID's advertiser page names CTR Guard, CPA Tune and smart bidding, and no generative image or headline tool is named on either page. | 11 | vendor assertion | unresolved | medium | [394, 267] | | J-F14 | Teads' Q2 2026 results (2026-08-06) describe 'predictive AI technology', the launch of 'Teads EngageOS, an AI-powered operating system for publishers' and a 'Teads CTV Ensemble', with CTV revenue up 67% YoY to 13% of Q2 revenue (7% in Q2 2025), while total revenue fell 17% YoY to $284.6M, ex-TAC gross profit fell 14% to $123.4M, Adjusted EBITDA fell 74% to $7.0M and FY2026 guidance was suspended citing Direct Response/SME volatility; no generative-creative or agentic product is named. | 11 | direct record | supported as stated | high | [271, 387] | | J-F16 | Google's advertiser-side agents at GML 2026 are 'Ask Advisor', a unified Gemini-powered agent across Google Ads, Google Analytics, Merchant Center and Google Marketing Platform, and 'Ads Advisor' with three agentic safety and policy features; by 2026-08-27 Asset Studio's Multimodal Video Creation was generally available, and on 2026-09-16 Google announced a US-only beta of a Business Agent for YouTube ads. | 11 | vendor assertion | unresolved | high | [258, 398, 399] | | J-F17 | Google attributes an average 30% increase in conversions or conversion value to 'hundreds of improvements' made to Demand Gen in H2 2025, citing internal global experiment results. It does not disclose the experiment design, baseline, sample or variance. | 11 | vendor assertion | supported with qualification | low | [398] | | J-F18 | Meta's 2026-09-03 post positions Advantage+ sales campaigns as 'Set It Up Once, Watch It Perform' with audience, placement and budget optimisation layered in, Advantage+ creative generating AI variations, and a WhatsApp 'Business Agent'; supporting figures are advertiser case studies (e.g., Zanskar Health '64% lower cost per purchase' in an A/B test; HubX '28% more incremental subscriptions' via a conversion lift study; Underneat '13% incremental lift in purchases'), not platform-wide measurements. | 11 | vendor assertion | unresolved | medium | [400] | | J-F19 | Ads Agent's listed availability is 3 North American countries, Brazil, 16 European, 8 Middle Eastern and 6 Asia-Pacific markets. The finding's EU (9) and ME (6) counts are wrong. | 11 | vendor assertion | supported with qualification | high | [395, 396, 397] | | J-F23 | The Trade Desk's Q4 2025 results release (2026-02-25) references Kokai only qualitatively and discloses no share of clients or spend on Kokai, no Koa/AI, agentic or Deal Desk metrics, while reporting Q4 2025 revenue of $847M (+14% YoY), FY2025 revenue of $2,896M (+18%) and 2025 gross spend of $13.4B. | 11 | direct record | unresolved | high | [401] | | J-F24 | Every automated or agentic system whose documentation was opened retains an explicit human approval gate: Amazon Ads Agent ('Campaigns only launch after you review and approve'), AdCP ('check_governance' before launch, escalation to humans when 'something exceeds the agent's authority', audit logs via 'get_plan_audit_logs'), MGID MIA ('the decision to act remains yours') and Taboola Abby (question-led guided setup); no page opened documents unattended campaign launch or budget changes. | 11 | synthesis | unresolved | medium | [395, 377, 394, 392, 375] | | J-F25 | Machine-readable reporting and feedback are specified in AdCP (get_media_buy_delivery, log_event, provide_performance_feedback) and offered in Amazon's Ads Agent via natural-language SQL over Amazon Marketing Cloud, whereas no native vendor page or filing opened (Taboola, Teads, MGID) documents an agent-facing reporting or goal-specification API. | 11 | inference | unresolved | low | [377, 395, 386, 387, 267] | | J-F26 | IAB Tech Lab's OpenRTB Dynamic Native Ads API remains at v1.2 (July 2017) on the standards index while the same body's agentic specifications iterate to AAMP 3.0 and AdCP cuts eight tags in sixteen days, so the native format's programmatic contract is nine years old at the moment agent-to-agent buying specs are being written. | 11 | synthesis | unresolved | high | [378, 379, 376] | | L-F02 | Teads' Q2 2026 revenue fell 17% YoY to $284.6 million, ex-TAC gross profit fell 14% to $123.4 million, Adjusted EBITDA fell 74% to $7.0 million and net loss widened to $42.5 million; H1 2026 revenue was down 13% with adjusted free cash flow of negative $37.9 million. | 11 | direct record | supported as stated | high | [271] | | L-F03 | On 2026-08-06 Teads suspended all guidance, including its previously reaffirmed FY2026 Adjusted EBITDA guidance of approximately $100 million, citing 'the volatility of the Direct Response and SME business' and stating that this business 'faced open-web headwinds'. | 11 | direct record | supported as stated | high | [271, 346, 29] | | L-F04 | Teads executed a restructuring in late 2025 that reduced headcount by approximately 10% (anticipated annualised savings $35-40 million), booked $15.3 million of restructuring and $28.9 million of acquisition/integration costs in FY2025, and replaced its CCO, CMO and Managing Director North America; CEO David Kostman remained in post through the Q1 2026 release and no CEO-departure 8-K was found. | 11 | direct record | unresolved | medium | [29, 346] | | L-F05 | Teads' FY2025 10-K states in the past tense that generative-AI answers in search engines and browsers have 'reduced [users'] need to click through to original publisher websites' and that the resulting reduction in media-partner traffic 'has affected, and could in the future have a significant effect on, our revenue and results of operations'. | 11 | direct record | supported as stated | high | [46] | | L-F08 | Taboola's FY2026 revenue guidance was cut between May and August 2026 (from $2,006-2,062M to $1,930-1,956M, i.e. $76M lower at the bottom and $106M lower at the top) while ex-TAC gross profit guidance was raised ($760-781M to $772-783M) and Adjusted EBITDA guidance was raised at the low end ($222-240M to $228-240M). | 11 | direct record | supported with qualification | high | [348, 347, 30] | | L-F09 | Taboola's Q2 2026 revenue of $476.8 million (+2.4% YoY) came in below its own Q2 guidance of $492-505 million, while ex-TAC gross profit of $192.4 million (+11.8% YoY) landed inside the $189-194 million guidance range. | 11 | direct record | supported as stated | high | [30, 347] | | L-F10 | The 2024 facts are correct. Delete or reword 'repeating the 2024 pattern in which Yahoo revenue shifted from gross to net presentation' (top finding 2) and 'echoing 2024 when Yahoo revenue moved from gross to net presentation' (summary): the 2026 revenue cut shares only the surface pattern (revenue cut while ex-TAC is held or raised). Management attributed it to a Google product-policy change and publisher exits, not to a presentation change. | 11 | direct record | supported with qualification | high | [349, 350, 41, 351] | | L-F15 | Taboola's filings never use the phrase 'AI Overviews', its 10-Qs filed July-September 2026 do not contain the phrase 'generative AI', and its 2026 quarterly releases make no statement about AI search reducing publisher traffic, whereas its FY2023-FY2025 10-Ks do contain the phrase 'generative AI'. | 11 | inference | unresolved | medium | [353, 30, 347, 348] | | L-F17 | Despite a 22% Core Sessions decline, People Inc.'s Q2 2026 Digital revenue rose 6% to $289.9 million: advertising was flat at $174.2 million (open programmatic up on 'higher rates, partially offset by lower impression volumes'), performance marketing rose 13% to $68.8 million and licensing rose 23% to $47.0 million helped by a Meta content partnership signed in Q4 2025. | 11 | direct record | unresolved | high | [215] | | L-F18 | Ad-request declines are Q2 (April-June) 2026. Programmatic spend changes (-44% US, -14.3% UK, -30.6% combined) are FIRST HALF 2026, not Q2. Outstream -76% and apps +23% are JUNE 2026 only, and the apps figures appear to be US. The data-table row 'Ozone programmatic spend YoY ... Q2 2026' and top finding 3 ('Q2 2026 ... with US programmatic spend -44%') need the same period fix. | 11 | commissioned measurement | supported with qualification | medium | [357] | | L-F19 | Dianomi plc (AIM-listed financial-content native ads) reported FY2025 revenue of £27.4 million (-2.1% vs £28.0 million), an Adjusted EBITDA loss of £0.3 million and £5.8 million cash with no debt, citing 'lower publisher traffic levels linked to the growing use of AI-powered content discovery tools'; H1 2026 revenue rose 2% to £13.4 million (+4.5% at constant currency) and July-August 2026 revenue was up 14% YoY. | 11 | direct record | unresolved | medium | [358, 359] | | L-F21 | Equativ acquired Sharethrough in June 2024 and retired the Sharethrough brand on 2025-06-09, announcing 'the full unification of Sharethrough into its global brand' with a combined team of over 750 people in 20 countries and native listed as one of five channels. | 11 | vendor assertion | supported with qualification | high | [360] | | L-F22 | Outbrain's acquisition of Ligatus GmbH was effective 2019-04-01, per the acquired-subsidiary schedule in a bank loan amendment filed with Outbrain's 2021 S-1, and 'Ligatus' still appears in Teads Holding Co.'s FY2025 Exhibit 21.1 subsidiary list. | 11 | direct record | unresolved | medium | [362, 363] | | L-F24 | The only documented publisher removal of a recommendation widget opened in this stream is Outside magazine, which in 2017 replaced Outbrain (about 10% of its digital ad revenue) with in-house widgets, reporting 1.3% CTR overall vs about 1% for Outbrain and 5.8% vs 2.5% among readers reaching the end of an article; the same article says Outbrain 'accounts for 30% of revenue for some publishers'. | 11 | vendor assertion | unresolved | high | [374] | | L-F26 | Microsoft retired the 'Microsoft Start' brand in November 2024 and reverted its web portal and app to the MSN name; this was a branding change, not a product or supply discontinuation, and Taboola's Microsoft partnership continued at 5% or more of revenue through FY2025. | 11 | synthesis | unresolved | low | [361, 1] | | C-F07 | Taboola's disclosed Yahoo related-party revenue (advertiser spend billed by Yahoo) was $201.6M in 2025 (10.5% of revenue) versus $233.6M (13.0%) in 2024, while TAC paid to Yahoo for supply rose to $349.0M from $275.5M; in H1 2026 Yahoo-billed revenue rose to $148.9M (15.8%) from $94.8M and TAC to Yahoo to $201.1M from $159.6M. | abstract, 3, 4 | direct record | supported as stated | high | [1, 47, 80] | ### What to distrust in this paper Six things. Each is named here so it is not discovered later. The first is a list of what the paper did not cover, so that silence is not read as absence. It did not measure sponsored editorial sold directly by publisher studios, creator and influencer content, or in-mail and newsletter native as surfaces. It did not assemble the competition-law record, such as the 2019 to 2020 review of the proposed Taboola and Outbrain merger or the Google ad-tech ruling beyond the two suits it cites. It did not cover rules for child-directed pages, industry self-regulation, or sector rules for finance, investment advice and prescription drugs, where much native demand sits. It did not check each vendor's TAG certification or MRC accreditation status, and it did not map product retirements outside the native specialists, such as Yahoo's own native marketplace or Microsoft's buy-side platform. Asia-Pacific markets and their native-ad guidelines were out of scope. The open-web native market has no measured size, and this paper does not invent one. Each spend figure it quotes is somebody's fence around the market. Chapter 2 says whose. The financial comparisons rest on non-GAAP measures that each company defines for itself. Ex-TAC gross profit is roughly revenue minus traffic acquisition cost. That cost is what a platform pays publishers and other suppliers for placements. The measure cannot be compared across Taboola and Teads Holding Co. unless it is adjusted. The paper says so each time it uses it. A reader who wants one number will find the GAAP lines in the figure data. The consumer evidence is old. The main ad-recognition studies were fielded between 2014 and 2018. Newer experiments from 2022 and 2024 used smaller or student samples. Two further leads from 2025 and 2026 were found too late to screen. No representative US replication was found. The paper reports the studies as the best that could be found. It does not claim they describe 2026 readers. The vendor profiles are built from public records. A company can be under-rated for being private rather than for being weak. The SWOT in each profile applies the same rule. Strengths and weaknesses rest on the record. Opportunities and threats are the author's reading of dated facts, and are labelled as inference. MGID's ownership, Revcontent's and TripleLift's revenue, and Readpeak's scale could not be established. Their cells say n/e rather than guessing. The author has interests to declare, set out in the abstract. The research was designed to test the companies and standards he is connected to as hard as the rest of the field. Whether it succeeded is for the reader to judge from the ledger. That is why the ledger is published. ## Appendix: instruments, data and definitions *The study table, the data behind each figure, the steps to rebuild the paper and the terms it uses, in one place. The numbered source list follows as its own section.* ### Study evidence table This lists each consumer and causal study opened for this paper. For each one it gives the design, sample, setting, outcome measure, headline result, sponsor and how widely it applies. Studies with unlike outcomes are not pooled, on purpose. You can filter by whether a study is independent or native-specific. You can sort by any column. **Study evidence table** | Study | Year | Design | Sample | Setting | Format / channel | Outcome measure | Headline result | Uncertainty | Sponsor | Generalisability | Src | |---|---|---|---|---|---|---|---|---|---|---|---| | Wojdynski & Evans, Study 1 | 2015 | 3x4 between-subjects online experiment | 242 adults (MTurk) | US, newspaper-style page | Article-style sponsored content | Open-ended ad recognition; credibility; brand attitude; share intent | 7% recognition (17/242); middle vs top position OR 5.10, p<.05; explicit vs vague wording OR 1:0.121, p<.01; recognisers rate credibility 4.11 vs 5.03 | stated in result | None stated (academic) | Convenience sample; single fictitious brand; tiny recogniser counts | [99] | | Wojdynski & Evans, Study 2 | 2015 | 3-condition lab eye-tracking experiment | 60 undergraduates | US, USA Today-style page | Article-style sponsored content (Dell) | Fixation on disclosure; recognition; credibility | 18.3% recognition (11/60); 90%/60%/40% fixated middle/bottom/top; middle vs top attention OR 12, p<.01; all 11 recognisers fixated; credibility 4.53 vs 5.35, p<.01 | stated in result | None stated | Students; 75% female; one page | [99] | | Hyman, Franklyn, Yee & Rahmati | 2017 | Online survey with randomised label modification | 896 valid of 1,228 (SSI panel) | US, screenshots of nine platforms | 16 real native ads, 5 regular ads, 8 editorial | Forced-choice paid/unpaid/don't know | 37% recognised native vs 81% regular (t=42.36, p<.001); per-ad 21-72%; 'Paid Ad' bar raised 40%->56% (t=7.31) | stated in result | None stated (law faculty) | Non-probability panel; static images | [102] | | FTC staff, Blurred Lines | 2017 | Usability-lab between-subjects with eye tracking and think-aloud | 48 (24 per condition) | US lab; 4 native and 4 search scenarios, desktop and mobile | Native articles and content-recommendation widgets (Gear Patrol, Time mobile, Yahoo, email click-in) | Coder-judged recognition; fixation time | +21 pp overall (47%->68%, 95% CI 15-27); native +23 pp (CI 14-32); looking time -21% (native -26%, n.s.) | stated in result | US regulator | Exploratory; multiple changes per ad; small n | [103] | | Amazeen & Wojdynski (Journalism) | 2018 | 2x2x2x3 online experiment | 800 US adults (YouGov representative; 738 in native conditions) | US; NYT/WSJ vs digital-only layouts | Article-style sponsored content (Bank of America) | Open-ended recognition; publisher attitude and credibility | 9% recognition; prominence OR 1.97, high explicitness OR 3.66, logo OR 1.64 (p<.10); publisher credibility 3.93 vs 4.46 (p<=.001) | stated in result | None stated | Single stimulus; small recogniser subgroup | [100] | | Amazeen & Wojdynski (Mass Comm & Society) | 2018 | Same YouGov experiment, antecedent analysis | 738 | US | Article-style sponsored content | Recognition; critical responses | Fewer than 1 in 10 recognised; recognition higher among younger, more educated, information-seeking readers | not reported | None stated in abstract | Abstract only | [101] | | Eisend, van Reijmersdal, Boerman & Tarrahi | 2020 | Meta-analysis (HLM of correlations) | 61 papers, 57 datasets, 473 effect sizes, 278,791 respondents | Mixed countries; all sponsored-content formats | TV placement, advergames, blogs, influencer, native | Recognition; persuasion knowledge; brand attitude; credibility; behavioural intention | r=.255 recognition, .257 persuasive intent, -.108 brand attitude, -.132 credibility, -.023 behavioural intention (n.s.) | stated in result | German Research Foundation EI 508/10-1 | No native-only estimate; heterogeneous | [108] | | van Reijmersdal, Brussee, Evans & Wojdynski | 2023 | Single experiment | 133 | Not stated (blog post) | Sponsored blog post | Persuasion knowledge; perceived transparency; brand attitude; purchase intent | Disclosure lowers persuasion via persuasion knowledge and raises it via transparency; source effects n.s. | stated in result | None stated | Abstract only; small n | [112] | | Campbell & Evans | 2018 | Two experiments | Not in abstract | Not stated | Article-style native with companion banner | Recognition; evaluations | Companion banner equals traditional disclosure for recognition; transparency mutes negative reactions | not reported | None stated | Abstract only | [113] | | Jung & Heo | 2019 | Two experiments | Not in abstract | Not stated (social media style) | Natively formatted social ad | Ad evaluation; persuasion knowledge | Evaluation driven by knowledge of social-media ad tactics and recognition, not disclosure explicitness | not reported | None stated | Abstract only | [109] | | Aribarg & Schwartz | 2019 | Randomised online and field experiments with clickstream, eye movement, survey | Not in abstract | US news site | In-feed native vs display, disclosure styles | CTR; visual attention; brand recognition; website trust | Native higher CTR at same position; display more attention, brand recognition and website trustworthiness (magnitudes not accessible) | not reported | None stated | Abstract only; publisher pages blocked | [115] | | Huebner, Thalmann & Henseler | 2025 | Within-subjects eye tracking plus cued retrospective think-aloud | 152 (mean age 25) | Germany lab; mock Instagram feed | In-feed social native (8 sponsored/21 organic) | Fixations; dwell time; recognition cues | Organic posts +25% fixations (eta2 .404), ~250 ms longer dwell; early disclosure/CTA fixation reduced dwell in 80.6% of significant cases | stated in result | Not stated | Young sample; 110-second cap; retrospective bias | [116] | | Boerman, van Reijmersdal & Neijens | 2012 | 3-condition experiment | 116 | Netherlands, TV | Brand placement disclosure 0/3/6 s | Persuasion knowledge; brand memory; brand attitude | 6 s disclosure activates attitudinal persuasion knowledge and lowers brand attitude indirectly; any disclosure raises brand memory | not reported | None stated | Abstract only; TV not digital | [744] | | Zeng, Kohno & Roesner (ConPro) | 2020 | Crawler measurement plus manual coding | 6,498 news + 1,055 misinformation sites crawled; 5,414 ads sampled, 2,419 coded | US crawler, Jan 15-19 2020 | Native recommendation units vs display | Share of ads coded problematic by platform and site type | 44.6% problematic; native 87% vs display 20%; Taboola 85.7% of its ads, 61.1% of all problematic; Google 16.9% | not reported | Academic | Snapshot; one-third unrendered; broad codebook | [119] | | Zeng, Kohno & Roesner (CHI) | 2021 | Two surveys (qualitative n=60; rating n=1,025) | 1,025 (Prolific) | US, Jun-Sep 2020 | 500 real web ads incl. native/chumbox | Opinion labels; 1-7 liking | 45% of ads majority-negative; 20.6% clickbait; 11.2% deceptive; lowest clusters 43-72% native | not reported | Academic | Young, ad-blocker-heavy sample | [120] | | Zeng, Wei, Gregersen, Kohno & Roesner (IMC) | 2021 | Longitudinal crawler measurement | 745 sites; 1,402,245 ads | US six metros, 26 Sep 2020-19 Jan 2021 | All ads; political subset | Share political; platform attribution | 3.9% political; 52.0% of political ads = news/media articles; Zergnet 79.4%, Taboola 10.0%, Revcontent 5.7%, Content.ad 1.8% of those | not reported | Academic | Crawler impressions | [121] | | Yoshikawa & Roesner (arXiv) | 2025 | Crawler measurement | 745 sites; 15,110 ads; 315 political | US three metros, 4 Oct-7 Nov 2024 | All ads; political subset | Share political; content types | ~2.1% political; clickbait political news ads prevalent; voter-eligibility ads 46% of political ads | not reported | Academic | Preprint; single coder; no platform attribution | [122] | | Outbrain / Lumen (via MediaPost) | 2019 | Unspecified consumer study | 900+ | UK, France, Germany | Native recommendations on premium news vs social/display ads | Self-reported trust, click and purchase likelihood | +31% trusted, +16% clicked, +18% future purchase vs social; 62% easier to understand than display (no baselines) | not reported | Commissioned by Outbrain | Method undisclosed; report not located | [117] | | Contently | 2015 | Online survey | ~509 (Research Now, Census-balanced) | US | Six real sponsored articles | Ad vs article; felt deceived; publisher credibility | Majority read 4 of 6 as articles; 48% felt deceived; 62% say news site loses credibility | not reported | Vendor (Contently) | No significance tests; vendor interest | [118] | | Gupta, Halder & Narang | 2026 | Hybrid bibliometric + systematic review | 113 articles | Global literature | Native advertising broadly | Framework of antecedents -> recognition -> outcomes | Narrative; no pooled effect | not reported | None stated | Abstract only | [110] | | Lewis & Rao, QJE | 2015 | 25 randomized field experiments | millions of customers; USD 2.8M spend | Display and search (Yahoo!) for US retailers/brokerages | Display and search (Yahoo!) for US retailers/brokerages | lift vs randomized control | Median ROI CI > 100 percentage points wide; informative tests may need >10M person-weeks | stated in result | Yahoo! researchers; academic | High for the power problem: native effect sizes are unlikely to be larger | [162] | | Gordon, Zettelmeyer, Bhargava & Chapsky, Marketing Science | 2019 | 15 RCTs vs observational models (matching, regression, propensity) | 500M user-experiment observations; 1.6B impressions | Facebook feed ads | Facebook feed ads | lift vs RCT lift | Observational methods 'often fail' to recover RCT effects even with extensive covariates (magnitudes in full paper, not read) | not reported | Academic + Facebook co-authors | High for the method critique; medium for format | [163] | | Gordon, Moakler & Zettelmeyer, Marketing Science | 2023 | 663 treatment-control pairs from 563 experiments vs DML and stratified propensity matching | 563 experiments, 663 pairs, 1,673 pair-outcome observations; >5,000 user features | Facebook feed ads | Facebook feed ads | lift vs RCT lift | DML median lift 83/58/24% vs RCT 29/18/5% (upper/mid/lower funnel); SPSM 173/176/64% | stated in result | Academic + Meta co-author | High for the method critique | [164] | | Lewis, Rao & Reiley, WWW | 2011 | 3 controlled experiments vs exposed/unexposed and before-after | MTurk N=1,600 (806 treated) | Yahoo! display; video on Mechanical Turk | Yahoo! display; video on Mechanical Turk | lift vs randomized control | Before-after would overstate effect by 350% in Expt 2; true effect small and insignificant ('activity bias') | not reported | Yahoo! Research | Very high: exposed-vs-unexposed is the design used in native brand-lift studies | [165] | | Johnson, Lewis & Nubbemeyer, JMR | 2017 | Ghost ads / predicted ghost ads within RCT | >100M predicted ghost ads/day; one campaign reported | Google display retargeting | Google display retargeting | lift vs ghost-ad control | Site visits +17.2%, purchases +10.5%; same precision at >=10x lower cost than PSA controls | not reported | Google co-authors; academic | Medium: method requires platform implementation not shown by native vendors | [177] | | Blake, Nosko & Tadelis, Econometrica | 2015 | Brand keyword suspension; DMA-level geo experiments | eBay US; multiple DMAs | eBay paid search (brand and non-brand) | eBay paid search (brand and non-brand) | lift vs geo control / pre-period | Brand keywords: no measurable short-term effect; non-brand: negative average ROI; new/infrequent users respond, frequent users do not | not reported | eBay Research Labs; academic | Medium: heterogeneity lesson transfers; brand-search result does not | [175] | | Sahni, Narayanan & Kalyanam, JMR | 2019 | User-level randomized experiment | not in abstract | Banner retargeting, online home-improvement retailer | Banner retargeting, online home-improvement retailer | lift vs randomized control | +14.6% users returning within 4 weeks; 33% of week-1 effect on day 1; week-1 exposure raises week-2 effect | not reported | Academic | Medium-high: retargeting via native widgets is a core native use case | [173] | | Lambrecht & Tucker, JMR | 2013 | Field experiment | not in abstract | Dynamic vs generic retargeting banners, online travel firm | Dynamic vs generic retargeting banners, online travel firm | lift vs generic creative | Dynamic retargeted ads on average less effective than generic; parity only when preferences narrowed | not reported | Academic | Medium | [174] | | Shapiro, Hitsch & Tuchman, Econometrica | 2021 | Observational panel (Nielsen), generalizable design | 288 brands, 2010-2014 | US TV, 288 CPG brands | US TV, 288 CPG brands | lift vs model-based | Negative marginal ROI for >80% of brands; overall ROI positive for one third; elasticities far below literature | not reported | Academic | Low for format; high for publication-bias lesson | [176] | | Haus 'Meta Report' | 2025 | 640 geo experiments with synthetic control | 640 tests; avg USD 14M annual Meta spend per brand | Meta (DTC brands) | Meta (DTC brands) | lift vs geo control | Avg 19% KPI lift; Meta 7-day-click under-reports incrementality by 15% (DTC-only); 32% of impact in non-DTC channels; Advantage+ over-reported by 12 pp vs manual | not reported | Haus (measurement vendor) | Low for format; medium for lag/halo mechanics | [745] | | Haus 'Cyber Week Incrementality Report' | 2025 | Hundreds of geo experiments, Q4 2025 | not disclosed | CTV, YouTube, paid social (DTC brands) | CTV, YouTube, paid social (DTC brands) | lift vs geo control incl. post-treatment window | 41% of incremental value in post-treatment window; efficiency +75% with latent BFCM impact; evergreen PTW adds 26% | not reported | Haus (measurement vendor) | Low: native not broken out | [182] | | Taboola-Kantar brand-lift meta-analysis | 2023 | Exposed vs unexposed survey (pixel/cookie exposure) | study count and n not disclosed | Taboola recommendation units (Europe, LatAm, US) | Taboola recommendation units (Europe, LatAm, US) | lift vs unexposed users; Kantar vertical norms | Favorability 15% vs 4.2% norm; consideration 6.7% vs 4.6%; message association 87% vs 10.1%; awareness 10% vs 10% (unit pp/% unstated) | not reported | Taboola with Kantar | Direct format match, but observational design (activity-bias exposed) | [166] | | Outbrain-Nielsen True Engagement brand lift (automotive, PHD Germany) | 2018 | Clickers vs non-clickers survey | 1,573 users, June-July 2017 | Outbrain recommendation units (Germany) | Outbrain recommendation units (Germany) | lift vs non-clickers | +23% brand lift among all respondents (selection on engagement; no CI) | stated in result | Outbrain commissioned Nielsen | Direct format match, but non-causal design | [167] | | Taboola Realize ID press release | 2026 | none disclosed | none disclosed | Taboola/Realize performance | Taboola/Realize performance | lift vs none disclosed | 'up to 2.4x increase in conversion efficiency' | not reported | Taboola | Direct format match; unusable as evidence of lift | [168] | ### Data downloads - `native-advertising-2026.pdf` — the print edition: A4, with page numbers and chapter bookmarks - `native-advertising-2026.md` — the reading edition in Markdown, with numbered references - `data/source-register.csv` — every source with publication, event and access dates, type, geography, funding, access limits and reliability notes - `data/claim-ledger.csv` — every material claim with sources, passage locator, scope, evidence class, review outcome, confidence and final wording - `data/capability-matrix-wide.csv` — vendors × dimensions, scores only (0–5, n/e, n/a) - `data/capability-matrix-long.csv` — one row per cell with availability, evidence grade, date, rationale and sources - `data/nq.csv` — Native Quotient per vendor: group means, coverage, band, the four reweightings and every dimension score - `data/nq-method.csv` — NQ groups, weights, shrinkage, field mean, band thresholds, and the reweighting, shrinkage and missing-evidence tests - `data/nq-missing-evidence.csv` — every scored cell turned into n/e one at a time, with the NQ before and after - `data/swot.csv` — every SWOT line for the profiled vendors with its basis (record, vendor claim, inference) and sources - `data/deep-profiles.csv` — the deep-profile summaries: what is sold, buyers, earnings, dependencies, fit, limits - `data/deep-profile-records.csv` — every profiled capability, performance claim, financial figure and timeline event with sources - `data/vendor-universe.csv` — the reasoned vendor sample with scope group, owner, formats, routes and status - `data/study-evidence.csv` — consumer and causal studies with design, sample, sponsor and limits - `data/corrections-ledger.csv` — every change made after verification and challenge passes - `data/agentic-map.csv` — AI and agentic capabilities with status labels and observability - `data/rules.csv` — disclosure and governance instruments by jurisdiction - `data/estimates.csv` — market-size estimates with inclusions, exclusions and combinability - `data/decision-tool.csv` — the measurement decision tool rows - `data/figure-data.json` — the canonical dataset behind every data figure - `data/derived-checks.json` — every derived number with its formula and inputs - `data/stats.json` — every count the prose and figures quote about the research itself, computed from the ledgers - `data/scoring-rules.json` — dimension definitions, anchors and cell rules fixed before scoring - `data/corpus-map.csv` — the author corpus audit ### Reproduction guide, in brief The package that comes with this paper holds all that is needed to rebuild it. The master text is `paper.md`, with one file per chapter. Each citation is an ID that resolves against `data/source-register.json`. Each claim marker resolves against `data/claim-ledger.json`. An unknown ID fails the build, so a citation cannot be orphaned. The script `scripts/make-figures.mjs` builds the data figures from `data/figure-data.json`. That file is recomputed from the evidence files by `scripts/derive.mjs`. Each formula and input is recorded in `data/derived-checks.json`. The instruments are built from the same data files by `build/instruments.mjs`. The HTML and Markdown editions come from `build/render.mjs`, and the CSV set from `build/make-csv.mjs`. The PDF comes from `build/pdf.mjs`, through headless Chrome. The script `scripts/check.mjs` checks that references resolve and that claims tie back to their sources. It also checks figure sidecars, matrix cell rules and headline counts. It can also probe each source URL. The script `scripts/readability.py` measures Flesch Reading Ease on the Markdown edition. Its method is documented. It needs Node 22, Python 3.10 or later, and six small markdown packages listed in `package.json`. The full guide is `README.md`. ### The author's corpus, audited The research began from the author's own Performance & Native playbook. It also drew on his related essays, glossary entries, standards pages and position ledger. Each item was classed as retain, update, test, qualify or reject. The table below is the corpus map. The full records are in `data/corpus-map.csv`. Nothing from the author's corpus is used as evidence for a market fact. Each claim that started there was traced to its primary source or dropped. | Item | Type | Published | Disposition | Why | |---|---|---|---|---| | Performance & Native Playbook | playbook | 2026, market facts validated June 2026 | test | hypothesis source; every market fact re-checked in workstream I, operating figures labelled as the author's model | | Glossary entries on native, recommendation platforms, commerce media | glossary | 2026 | update | starting definition; extended to six axes and separated from targeting method | | The Open Web Isn't Dead. It's Uninsured. | essay | 2026-08-22 | retain | its primary sources (Taboola, Magnite, People Inc. filings; Ozone data) re-opened directly | | The Great Repackaging | essay | 2026-07-16 | qualify | vendor-published survey; cited as such | | The Evidence Premium | essay | 2026-09-11 | qualify | author's framework only; its straw poll is not evidence about native buyers | | Measurement, on the Browser's Terms; position ledger on attribution | essay and ledger | 2026-07 | retain | credit-is-not-cause distinction used in chapter 7 | | Position ledger on measurement (iROAS, Scalarization Trap) | ledger | 2025–2026 | test | tension to test against native vendors' CPA optimisation claims | | AdCP vs AAMP and related agentic essays | essay series | 2026-07 to 2026-09 | retain | protocol versions and registry counts re-verified at the cutoff; author's working-group role disclosed | | Standards deep dives on measurement science, retail media, incrementality, conversion APIs | reference pages | validated June to September 2026 | retain | used as a source map; each primary document opened directly | | The Intelligence Economy; The Customer's Agent Arrived First; Measurement for Agentic Commerce | essays | 2026-07 | qualify | only the primary sources they cite are reused | | The Packaging Problem and Commercial Architecture playbook | essay and playbook | 2026-06-13 | qualify | labelled author perspective in chapter 10 only | | Data is the New Currency to power Attribution | essay | 2015-09-18 | reject | records a change in the author's thinking; not used as evidence | | Proof and Story: contextual advertising in 2026 | research paper | 2026-08 | qualify | presentation benchmark and contextual-versus-native distinction only; its conclusions and counts are not reused | ### Definitions Native advertising. Paid ad units whose form and placement match the content or feed around them. The term describes format and placement surface only. Contextual targeting. Choosing which reader sees an ad from signals on the page or in the session. It describes the basis for the choice, not a promise that no personal data is used. It is a targeting method, not a format. Recommendation unit. A native format, usually at the end of an article. It shows paid and sometimes unpaid links made to look like further reading. It is also called a content-recommendation widget. In-feed unit. A native format placed between items of a feed. The feed can be on a publisher page, in a social feed or in a product feed. Sponsored editorial. Content made or paid for by an advertiser and run in the host's editorial style. It is also called branded content or an advertorial. Traffic acquisition cost (TAC). What a platform pays publishers and other suppliers for the placements it sells. Taboola and Teads Holding Co. report it inside cost of revenue. Ex-TAC gross profit. A non-GAAP measure that each company defines for itself. It is roughly revenue minus traffic acquisition cost, with add-backs that vary by company. It cannot be compared across firms unless it is adjusted. Minimum guarantee. A contract where the platform pays the publisher the greater of a revenue share or an agreed amount per thousand page views. The platform pays this no matter what advertisers pay. Made-for-advertising (MFA). Sites built to host ads rather than to be read. They are marked by a high share of paid traffic sourcing, high ad density and templated content. Classifiers do not all define it the same way. Incrementality. The effect that ads cause, set against what would have happened without them. It is shown by a holdout: a random group or a region that is kept from the ads. Attribution gives credit. It does not show cause. Attribution window. The time after a click or view in which a conversion, such as a sale, is credited to the ad. It is a reporting rule, not a measure of lag. Learning phase. The time a bidding system tuned for conversions needs, at a stated conversion volume, before its target can be trusted. Each vendor publishes it as rules. Exposed-versus-unexposed design. A comparison of outcomes for those who saw an ad and those who did not. No one is assigned at random. It is subject to activity bias and has been shown to overstate effects. OpenRTB Native Ads. The IAB Tech Lab spec that carries a native ad's parts through the programmatic bid stream. It was last finalised in March 2017, as version 1.2. AdCP and AAMP. Two protocol stacks for agentic advertising. AdCP is the Ad Context Protocol, governed by AgenticAdvertising.org. AAMP is IAB Tech Lab's Agentic Advertising Management Protocols. AdCP's docs refer to native formats, and a standard native format is deferred to its 3.2 release. No large native platform was found implementing either at the cutoff. Evidence class. The paper's label for what kind of thing a claim rests on. The classes are direct record, independent measurement, commissioned measurement and vendor assertion. The others are synthesis, inference, forecast and illustrative scenario. Review outcome. What the independent check found for a claim. It is one of: supported as stated, supported with qualification, contradicted, unresolved or withdrawn. Claims not picked for the verification sample are logged as unresolved, with a note saying so. n/e and n/a. In the capability matrix, n/e means the record gave no basis to score. The label n/a means the dimension does not apply to the vendor's model. Neither is a zero. A zero is kept for documented absence. ## References The 808 sources cited in this paper, numbered in order of first citation. The full register of 977 sources opened during the research, with access dates, access limits and reliability notes, is in data/source-register.csv. [1] Taboola.com Ltd. Form 10-K for fiscal year ended December 31, 2025. Taboola.com Ltd. (SEC EDGAR), 2026-02. https://www.sec.gov/Archives/edgar/data/1840502/000184050226000004/tbla-20251231.htm Accessed 2026-09-27. Access: partial extract. [2] Native Advertising: A Guide for Businesses. Federal Trade Commission, 2015-12. https://www.ftc.gov/business-guidance/resources/native-advertising-guide-businesses Accessed 2026-09-27. [3] IAB Native Advertising Playbook 2.0. IAB (working group led by Susan Borst; 69 member companies), 2019-05. https://www.iab.com/wp-content/uploads/2019/05/IAB-Native-Advertising-Playbook-2_0_Final.pdf Accessed 2026-09-27. [4] OpenRTB Dynamic Native Ads API Specification Version 1.2 (Final). IAB Tech Lab (GitHub), 2017-03. https://github.com/InteractiveAdvertisingBureau/Native-Ads/blob/main/OpenRTB-Native-Ads-Specification-Final-1.2.md Accessed 2026-09-27. [5] (Author's work) Proof and Story: The State of Contextual Advertising in 2026. Evgeny Popov, No Fluff Advisory, 2026-08. https://nofluffadvisory.com/research/contextual-2026/ Accessed 2026-09-27. [6] With Market Maturity Comes an Updated IAB Native Advertising Playbook and Predictions for the Future. IAB, 2019-05-09. https://www.iab.com/news/2019-native-insights/ Accessed 2026-09-27. [7] IAB/PwC Internet Advertising Revenue Report: Full Year 2025. IAB / PwC, 2026-04. https://iabhongkong.com/sites/default/files/2026-04/IAB_PwC_Internet_Ad_Revenue_Report_Full_Year_2025_April_2026.pdf Accessed 2026-09-27. [8] Digital Ad Revenue Climbs to Nearly $300B as IAB Celebrates 30 Year Anniversary (press release). IAB, 2026-04-16. https://www.iab.com/news/digital-ad-revenue-climbs-to-nearly-300b-as-iab-celebrates-30-year-anniversary/ Accessed 2026-09-27. [9] FAQ on native advertising: Formats, AI opportunities, and the best metrics for 2026. EMARKETER, 2026-01-15. https://www.emarketer.com/content/faq-on-native-advertising--formats--ai-opportunities--best-metrics-2026 Accessed 2026-09-27. Access: paywalled excerpt. [10] US Ad Spending 2023: Amid Deceleration, CTV and Retail Media Are Silver Linings. eMarketer / Insider Intelligence (Ethan Cramer-Flood), 2023-05. https://www.emarketer.com/uploads/pdf/US_Ad_Spending_2023.pdf Accessed 2026-09-27. [11] Magna latest to downgrade global ad spending forecast, expects $979B. Marketing Dive (reporting MAGNA June 2025 Global Ad Forecast), 2025-06-17. https://www.marketingdive.com/news/magna-latest-to-downgrade-global-ad-spending-forecast-expects-979b/750871/ Accessed 2026-09-27. [12] MAGNA Holds US Radio's Ad Forecast As Is; Digital Shifts Up. Radio Ink (reporting MAGNA June 2025 US forecast), 2025-06-15. https://radioink.com/2025/06/15/magna-holds-us-radios-ad-forecast-as-is-digital-shifts-up/ Accessed 2026-09-27. [13] MAGNA Advertising Forecasts (4A's listing of the December 2024 release). MAGNA (IPG Mediabrands) via 4A's, 2024-12-08. https://www.aaaa.org/research-report/magna-ad-spend-forecasts/ Accessed 2026-09-27. Access: partial extract. [14] This Year Next Year 2026 Midyear Forecast: Global ad revenue to grow 8.9% to $1.3 trillion. WPP Media, 2026-06-16. https://www.wppmedia.com/news/report-this-year-next-year-midyear-2026 Accessed 2026-09-27. [15] Global Ad Spend Set to Surpass $1 Trillion for the First Time in 2026 as the Algorithmic Era Redefines Growth. dentsu, 2025-12-03. https://www.dentsu.com/news-releases/global-ad-spend-set-to-surpass-one-trillion-for-the-first-time-in-2026-as-the-algorithmic-era-redefines-growth Accessed 2026-09-27. [16] IAB Europe's AdEx Benchmark 2025 Report Records 10.5% Growth to Reach EUR131bn. IAB Europe, 2026-07-07. https://iabeurope.eu/iab-europes-adex-benchmark-2025-report/ Accessed 2026-09-27. [17] OpenRTB Native Ads (standards page). IAB Tech Lab, undated (v1.2 finalized). https://iabtechlab.com/standards/openrtb-native/ Accessed 2026-09-27. Access: partial extract. [18] Driven by Social, Native Accounts for Nearly Two-Thirds of Display Ad Spend. eMarketer (Ross Benes), 2019-03-13. https://contentstorage-na1.emarketer.com/54158586950cc0c01478d24fd0439ed3/Driven_by_Social_Native_Accounts_for_Nearly_Two_Thirds_of_Display_Ad_Spend_eMarketer.pdf Accessed 2026-09-27. [19] EMarketer: Native ads to grab 63% of display spending in 2019. Marketing Dive (reporting eMarketer), 2019-04-18. https://www.marketingdive.com/news/emarketer-native-ads-to-grab-63-of-display-spending-in-2019/553040/ Accessed 2026-09-27. [20] Native Insider: Native Digital Display Ad Spending Projected To Reach $22B. MediaPost (reporting eMarketer), 2017-03-22. https://www.mediapost.com/publications/article/297532/native-digital-display-ad-spending-projected-to-re.html Accessed 2026-09-27. [21] US Native Digital Display Advertising Forecast (report landing page). eMarketer, 2018-04-10. https://www.emarketer.com/content/us-native-digital-display-advertising-forecast Accessed 2026-09-27. Access: paywalled excerpt. [22] US Native Ad Spending, 2019-2023 (billions, % change and % of total display ad spending) - chart 261233. eMarketer / Insider Intelligence, 2023-01 (chart page timestamp 2023-02-28). https://www.emarketer.com/chart/storage/chart_images/286812/280159.svg Accessed 2026-09-27. Access: not recorded (added by verifier). [23] Meta Reports Fourth Quarter and Full Year 2025 Results (Form 8-K Exhibit 99.1). Meta Platforms, Inc., 2026-01-28. https://www.sec.gov/Archives/edgar/data/1326801/000162828026003832/meta-12312025xexhibit991.htm Accessed 2026-09-27. [24] Alphabet Inc. Form 10-K FY2025 - Revenues - Revenue by Type (Details) (XBRL report R44). Alphabet Inc., 2026-02. https://www.sec.gov/Archives/edgar/data/1652044/000165204426000018/R44.htm Accessed 2026-09-27. [25] Alphabet Announces Fourth Quarter and Fiscal Year 2025 Results (earnings release PDF). Alphabet Inc., 2026-02-04. https://s206.q4cdn.com/479360582/files/doc_financials/2025/q4/2025q4-alphabet-earnings-release.pdf Accessed 2026-09-27. [26] Alphabet Inc. Form 10-K FY2025 (main document). Alphabet Inc., 2026-02. https://www.sec.gov/Archives/edgar/data/1652044/000165204426000018/goog-20251231.htm Accessed 2026-09-27. Access: partial extract. [27] Amazon.com Announces Fourth Quarter Results (earnings release PDF). Amazon.com, Inc., 2026-02-05. https://s2.q4cdn.com/299287126/files/doc_earnings/2025/q4/earnings-result/AMZN-Q4-2025-Earnings-Release.pdf Accessed 2026-09-27. [28] Taboola Reports Strong Fourth Quarter & Full Year 2025 Financial Results. Taboola.com Ltd. (GlobeNewswire release; syndicated copy on The Globe and Mail), 2026-02-25. https://www.theglobeandmail.com/investing/markets/markets-news/GlobeNewswire/410101/taboola-reports-strong-fourth-quarter-full-year-2025-financial-results/ Accessed 2026-09-27. [29] Teads Holding Co. Announces Fourth Quarter and Full Year 2025 Results (Form 8-K Exhibit 99.1). Teads Holding Co. (formerly Outbrain Inc.), 2026-03-05. https://www.sec.gov/Archives/edgar/data/1454938/000145493826000007/q42025er-exhibit991.htm Accessed 2026-09-27. [30] Taboola Q2 2026 results press release (Form 8-K, EX-99.1). Taboola.com Ltd. (SEC EDGAR), 2026-08-05. https://www.sec.gov/Archives/edgar/data/1840502/000114036126031336/ef20078191_ex99-1.htm Accessed 2026-09-27. [31] Recommended Acquisition of Dianomi plc (by Taboola Europe Limited). Dianomi plc / Taboola Europe Limited (RNS via Investegate), 2026-09-18. https://www.investegate.co.uk/announcement/rns/dianomi--dnm/recommended-acquisition-of-dianomi-plc/9778532 Accessed 2026-09-27. [32] Taboola Reports Strong Fourth Quarter & Full Year 2025 Financial Results. Taboola.com Ltd. (via GlobeNewswire), 2026-02-25. https://www.globenewswire.com/news-release/2026/02/25/3244402/0/en/taboola-reports-strong-fourth-quarter-full-year-2025-financial-results.html Accessed 2026-09-27. [33] Yahoo and Taboola Enter 30-Year Commercial Agreement ... Partnership to Generate Approximately $1 Billion in Annual Revenue. Yahoo Inc. (press release), 2022-11-28. https://www.yahooinc.com/press/yahoo-and-taboola-enter-30-year-commercial-agreement-developing-a-leading-offering-for-advertisers-publishers-and-merchants-on-the-open-web-partnership-to-generate-approximately-1-billion-in-annual-revenue Accessed 2026-09-27. [34] Yahoo and Taboola Announce Closing of Deal, 30-Year Strategic Partnership Sees Taboola Power Recommendations for Yahoo. Taboola (press release), 2023-01-17. https://www.taboola.com/press-releases/taboola-and-yahoo-close-deal/ Accessed 2026-09-27. [35] Amendments No. 5, 6 and 7 to the Digital Property and Demand Services Agreement between Yahoo Inc. and Taboola (exhibit to Taboola Form 10-Q, FY2026). Taboola.com Ltd. / Yahoo Inc. (SEC EDGAR), 2026. https://www.sec.gov/Archives/edgar/data/0001840502/000184050226000009/yamendments.htm Accessed 2026-09-27. [36] Taboola and MSN Launch Multi-Year Strategic Partnership, Bringing Native to 50 Global Markets. Taboola (company blog), 2016-01-02. https://www.taboola.com/taboola-updates/taboola-and-msn-launch-multi-year-strategic-partnership-bringing-native-to-50-global-markets/ Accessed 2026-09-27. [37] Taboola Reaches 10 Years in Serving Advertising on Microsoft Properties ... Now Serves Advertising Across Microsoft Productivity Suite, Including Outlook Mail and Office 365. Taboola (GlobeNewswire), 2025-03-18. https://www.globenewswire.com/news-release/2025/03/18/3044686/31297/en/Taboola-Reaches-10-Years-in-Serving-Advertising-on-Microsoft-Properties-Including-MSN-Edge-Browsers-and-Windows-Experiences-Taboola-Now-Serves-Advertising-Across-Microsoft-Producti.html Accessed 2026-09-27. [38] Display and Native ads | Microsoft Advertising. Microsoft Advertising, n/a (living product page). https://about.ads.microsoft.com/en/solutions/ad-products-formats/display Accessed 2026-09-27. [39] Microsoft signs MGID for native advertising. MGID (PR Newswire), 2020-08-19. https://www.prnewswire.com/news-releases/microsoft-signs-mgid-for-native-advertising-301113997.html Accessed 2026-09-27. [40] Apple taps infamous Taboola to sell ads for Apple News and Stocks. 9to5Mac (citing Axios / Sara Fischer), 2024-07-16. https://9to5mac.com/2024/07/16/apple-taps-infamous-taboola-as-new-ad-seller-for-apple-news-and-stocks/ Accessed 2026-09-27. [41] Taboola Q2 2024 results press release (Exhibit 99.1 to Form 8-K). Taboola.com Ltd. (SEC EDGAR), 2024-08-07. https://www.sec.gov/Archives/edgar/data/1840502/000114036124035979/ef20033470_ex99-1.htm Accessed 2026-09-27. [42] Taboola Q3 2024 results press release (Exhibit 99.1 to Form 8-K). Taboola.com Ltd. (SEC EDGAR), 2024-11-07. https://www.sec.gov/Archives/edgar/data/1840502/000114036124045483/ef20037783_ex99-1.htm Accessed 2026-09-27. [43] Apple dinged over low-quality, scammy-looking ads on Apple News. 9to5Mac (citing Kirk McElhearn), 2026-02-06. https://9to5mac.com/2026/02/06/apple-dinged-over-low-quality-scammy-looking-ads-on-apple-news/ Accessed 2026-09-27. [44] Category Leaders Outbrain and Teads to Merge, Creating an Independent End-to-End Advertising Platform for the Open Internet. Outbrain Inc. (GlobeNewswire), 2024-08-01. https://www.globenewswire.com/news-release/2024/08/01/2922739/0/en/Category-Leaders-Outbrain-and-Teads-to-Merge-Creating-an-Independent-End-to-End-Advertising-Platform-for-the-Open-Internet.html Accessed 2026-09-27. [45] Outbrain Completes the Acquisition of Teads (Exhibit 99.1 to Form 8-K). Outbrain Inc. (SEC EDGAR), 2025-02-03. https://www.sec.gov/Archives/edgar/data/1454938/000114036125002802/ny20037354x5_ex99-1.htm Accessed 2026-09-27. [46] Teads Holding Co. Form 10-K for fiscal year ended December 31, 2025. Teads Holding Co. (SEC EDGAR), 2026-03. https://www.sec.gov/Archives/edgar/data/1454938/000145493826000015/tead-20251231.htm Accessed 2026-09-27. Access: partial extract. [47] Taboola.com Ltd. Form 10-Q for the quarter ended June 30, 2026. Taboola.com Ltd. (SEC EDGAR), 2026-08. https://www.sec.gov/Archives/edgar/data/1840502/000184050226000012/tbla-20260630.htm Accessed 2026-09-27. [48] Taboola Expands Beyond Native Ads with Realize; New Technology Platform Allows Taboola to Serve the Entire Performance Advertising Market. Taboola (press release), 2025-02-26. https://www.taboola.com/press-releases/taboola-realize/ Accessed 2026-09-27. [49] Taboola Announces Strong Momentum for Realize, Pushing Beyond Native with Expanded Partnerships ... Taboola Expects to Pay Over $1.5 Billion to Publishers and OEMs in 2025. Taboola (GlobeNewswire), 2025-10-15. https://www.globenewswire.com/news-release/2025/10/15/3167133/31297/en/Taboola-Announces-Strong-Momentum-for-Realize-Pushing-Beyond-Native-with-Expanded-Partnerships-with-Leading-Global-Publishers-Including-TIME-Weather-Channel-Digital-Gannett-USA-TOD.html Accessed 2026-09-27. [50] Taboola Announces Expanded Partnership with NBC News ... in Programmatic Display Advertising. Taboola (corporate press release), 2026-08-19. https://www.taboola.com/press-releases/taboola-announces-expanded-partnership-with-nbc-news-one-of-the-worlds-leading-news-organizations-in-programmatic-display-advertising-advancing-taboolas-vision-to-become-the-economic-b/ Accessed 2026-09-27. [51] Taboola opens its DeeperDive advertising engine to generative AI experiences. Taboola, 2026-06-16. https://www.taboola.com/press-releases/taboola-ad-platform-for-genai/ Accessed 2026-09-27. [52] Equativ and Sharethrough Merge to Form One of the Largest Global Independent Ad Platforms and Marketplaces. Equativ (press release), 2024-06-12. https://www.equativ.com/press/equativ-and-sharethrough-merge Accessed 2026-09-27. [53] Equativ and Sharethrough Will Now Operate Under Equativ Brand. Equativ (press release), 2025-06-09. https://www.equativ.com/press/equativ-and-sharethrough-will-now-operate-under-equativ-brand-solidifying-global-position-as-leading-end-to-end-media-platform Accessed 2026-09-27. [54] Sharethrough homepage — An Equativ Group Company. Sharethrough / Equativ, undated (accessed 2026-09-27). https://www.sharethrough.com/ Accessed 2026-09-27. [55] Outbrain acquires Zemanta. Zemanta (company blog), 2017-07-25. https://www.zemanta.com/blog/outbrain-acquires-zemanta/ Accessed 2026-09-27. [56] Outbrain DSP: Simplified, Multi-Channel Advertising on the Open Web. Outbrain (company blog), 2024-08. https://www.outbrain.com/blog/outbrain-dsp-2/ Accessed 2026-09-27. [57] Life360 to Acquire Nativo, Accelerating Growth and Expanding Its Advertising Platform. Life360, Inc. (ads.life360.com newsroom), 2025-11-10. https://ads.life360.com/newsroom/life360-to-acquire-nativo Accessed 2026-09-27. [58] Life360 Completes Acquisition of Nativo and Surpasses 50 Million U.S. MAU. Life360, Inc. (GlobeNewswire), 2026-01-05. https://www.globenewswire.com/news-release/2026/01/05/3212816/0/en/life360-completes-acquisition-of-nativo-and-surpasses-50-million-us-mau.html Accessed 2026-09-27. [59] Form 8-K, Item 7.01 — completion of acquisition of Nativo, Inc.. Life360, Inc., 2026-01-05. https://www.sec.gov/Archives/edgar/data/1581760/000119312526001632/d81240d8k.htm Accessed 2026-09-27. [60] Dianomi plc Final Results (year ended 31 December 2024). Dianomi plc (RNS via Investegate), 2025-05-19. https://www.investegate.co.uk/announcement/rns/dianomi--dnm/final-results/8883818 Accessed 2026-09-27. [61] Dianomi maintains revenue resilience as AI initiatives support second-half recovery (DNM). Yahoo Finance UK (market report), 2026-05-29. https://uk.finance.yahoo.com/news/dianomi-maintains-revenue-resilience-ai-082025031.html Accessed 2026-09-27. [62] TripleLift Acquired by Vista Equity Partners. TripleLift (press release), 2021-03-29. https://triplelift.com/press/triplelift-acquired-by-vista-equity-partners/ Accessed 2026-09-27. [63] TripleLift Announces Leadership Changes. TripleLift (press release), 2024-07-12. https://triplelift.com/press/triplelift-announces-leadership-changes/ Accessed 2026-09-27. [64] Vista Equity Partners' TripleLift implements 'significant' layoffs. Digiday, 2025-07-14. https://digiday.com/marketing/vista-equity-partners-triplelift-implements-significant-layoffs/ Accessed 2026-09-27. [65] MGID Announces New CEO Ivan Doruda as Sergii Denysenko Becomes Chairman of the Board. MGID (company blog), 2025-10-07. https://www.mgid.com/blog/mgid-announces-new-ceo-ivan-doruda-as-sergii-denysenko-becomes-chairman-of-the-board Accessed 2026-09-27. [66] We've Acquired ADYOULIKE To Build A More Sustainable, Healthier Web. OpenWeb (company blog), 2022-04-11. https://www.openweb.com/blog/weve-acquired-adyoulike-to-build-a-more-sustainable-healthier-web/ Accessed 2026-09-27. [67] OpenRTB (Real-Time Bidding) standards page. IAB Tech Lab, 2023-09 (latest listed release). https://iabtechlab.com/standards/openrtb/ Accessed 2026-09-27. [68] Prebid.js Native Implementation Guide. Prebid.org, 2024. https://docs.prebid.org/prebid/native-implementation.html Accessed 2026-09-27. [69] About native ads - Display & Video 360 Help. Google, n/a (living help page). https://support.google.com/displayvideo/answer/7359723?hl=en Accessed 2026-09-27. [70] Native Advertising Platforms | Programmatic Native Marketing | The Trade Desk. The Trade Desk, n/a (living product page). https://www.thetradedesk.com/our-demand-side-platform/omnichannel-advertising/native Accessed 2026-09-27. [71] DSP Trade Desk adds native ad support, with Sharethrough as first integrated SSP under IAB standard. MarTech, 2016-03-07. https://martech.org/dsp-trade-desk-adds-native-ad-support-with-sharethrough-as-first-integrated-ssp/ Accessed 2026-09-27. [72] How native works in Google Ad Manager - Google Ad Manager Help. Google, n/a (living help page). https://support.google.com/admanager/answer/9414028 Accessed 2026-09-27. [73] About native ads - Google AdSense Help. Google, n/a (living help page). https://support.google.com/adsense/answer/7186747?hl=en Accessed 2026-09-27. [74] Amazon Publisher Services Unveils Suite of Innovations at Annual Summit 2025. Amazon Publisher Services, 2025-05-21. https://aps.amazon.com/aps/resource/amazon-publisher-services-unveils-suite-of-innovations-at-annual-summit/ Accessed 2026-09-27. [75] Teads Holding Co. Announces Fourth Quarter and Full Year 2025 Results. Teads Holding Co. (GlobeNewswire), 2026-03-05. https://www.globenewswire.com/news-release/2026/03/05/3249964/0/en/Teads-Holding-Co-Announces-Fourth-Quarter-and-Full-Year-2025-Results.html Accessed 2026-09-27. [76] Taboola Reports Strong Q2 2026 Financial Results, & Raises Full-Year ex-TAC Gross Profit and Adjusted EBITDA Guidance. Taboola.com Ltd. (via GlobeNewswire), 2026-08-05. https://www.globenewswire.com/news-release/2026/08/05/3339108/31297/en/Taboola-Reports-Strong-Q2-2026-Financial-Results-Raises-Full-Year-ex-TAC-Gross-Profit-and-Adjusted-EBITDA-Guidance.html Accessed 2026-09-27. [77] Taboola.com Ltd. Form 10-K for fiscal year ended December 31, 2022. Taboola.com Ltd. (SEC EDGAR), 2023-03. https://www.sec.gov/Archives/edgar/data/1840502/000114036123011303/brhc10049391_10k.htm Accessed 2026-09-27. [78] Outbrain Inc. Form 10-K for fiscal year ended December 31, 2024. Outbrain Inc. (SEC EDGAR), 2025-03-07. https://www.sec.gov/Archives/edgar/data/1454938/000145493825000020/outbrain-20241231.htm Accessed 2026-09-27. [79] In a shift, publishers can no longer count on content-recommendation guarantee checks. Digiday, 2018-01-30. https://digiday.com/media/shift-publishers-can-no-longer-count-content-recommendation-guarantee-checks/ Accessed 2026-09-27. [80] Yahoo and Taboola Enter 30-Year Commercial Agreement ... Partnership to Generate Approximately $1 Billion in Annual Revenue (Form 6-K Exhibit 99.1). Taboola.com Ltd. (SEC EDGAR), 2022-11-28. https://www.sec.gov/Archives/edgar/data/1840502/000114036122043198/ny20005769x2_ex99-1.htm Accessed 2026-09-27. [81] Taboola's next act: an AI answer engine for publishers. Sara Guaglione, Digiday, 2026-05-01. https://digiday.com/media/taboolas-next-act-an-ai-answer-engine-for-publishers/ Accessed 2026-09-27. [82] Teads Holding Co. Announces Second Quarter 2026 Results. Teads Holding Co. (via GlobeNewswire), 2026-08-06. https://www.globenewswire.com/news-release/2026/08/06/3340042/0/en/teads-holding-co-announces-second-quarter-2026-results.html Accessed 2026-09-27. [83] Teads Holding Co. Announces First Quarter 2026 Results. Teads Holding Co. (via GlobeNewswire), 2026-05-07. https://www.globenewswire.com/news-release/2026/05/07/3289745/0/en/Teads-Holding-Co-Announces-First-Quarter-2026-Results.html Accessed 2026-09-27. [84] Form 8-K, Item 3.01: Notice of Delisting or Failure to Satisfy a Continued Listing Rule. Teads Holding Co., 2026-08-14. https://www.sec.gov/Archives/edgar/data/1454938/000145493826000057/tead-20260811.htm Accessed 2026-09-27. [85] Form 8-K, Items 7.01/8.01: lawsuit against Google LLC and Alphabet Inc.; updated risk factor. Teads Holding Co., 2026-08-03. https://www.sec.gov/Archives/edgar/data/1454938/000145493826000048/tead-20260803.htm Accessed 2026-09-27. [86] Dianomi plc Annual Report 2025 (year ended 31 December 2025). Dianomi plc, 2026-06-01. https://www.dianomi.com/wp-content/uploads/2026/06/178114-Dianomi-RA-WEB.pdf Accessed 2026-09-27. [87] Dianomi plc Interim Results for the six months ended 30 June 2026. Dianomi plc, 2026-09-25. https://www.dianomi.com/wp-content/uploads/2026/09/Dianomi-Interim-Results-H126-FINAL.pdf Accessed 2026-09-27. [88] Leading Trade Groups Define 'Made For Advertising' Websites. ANA, 4A's, WFA, ISBA (press release), 2023-09-26. https://www.ana.net/content/show/id/pr-2023-09-made-for-advertising Accessed 2026-09-27. [89] ANA Programmatic Media Supply Chain Transparency Study: Complete Report (December 2023). Association of National Advertisers (ANA) with TAG TrustNet, Fiducia, PwC, Kroll; MFA data from DeepSee.io, 2023-12-05. https://www.adslot.com/wp-content/uploads/2023/12/ana-programmatic-media-supply-chain-transparency-study.pdf Accessed 2026-09-27. [90] Is Chumbox Economics Feeding The Industry's MFA Problem? (The Sell Sider). Ruben Schreurs, Chief Strategy Officer, Ebiquity, in AdExchanger, 2023-12-06. https://www.adexchanger.com/the-sell-sider/is-chumbox-economics-feeding-the-industrys-mfa-problem/ Accessed 2026-09-27. [91] Inside The Secret Meetings To Define MFA. AdExchanger, 2024-03-11. https://www.adexchanger.com/publishers/inside-the-secret-meetings-to-define-mfa/ Accessed 2026-09-27. [92] Does Forbes operate 'made for advertising' pages and transact mis-declared ad inventory?. Adalytics Research, 2024-03. https://adalytics.io/blog/ads-observed-on-www3-forbes-subdomain Accessed 2026-09-27. [93] 'More distrust in the marketplace': Agency execs press pause on Forbes spend after domain spoofing report. Digiday, 2024-04-12. https://digiday.com/media-buying/more-distrust-in-the-marketplace-agency-execs-press-pause-with-forbes-after-domain-spoofing-report/ Accessed 2026-09-27. [94] ANA Releases 2024 Programmatic Benchmark Study (press release). ANA with TAG TrustNet and Fiducia, 2024-12-12. https://www.ana.net/content/show/id/pr-2024-12-programmatic Accessed 2026-09-27. [95] Programmatic Transparency Benchmark: Q1 2025 Benchmark Findings (25-page PDF). ANA, TAG TrustNet, Fiducia, 2025-05. https://s3.amazonaws.com/media.mediapost.com/uploads/ANAQ12025Programmatic.pdf Accessed 2026-09-27. [96] The ANA Q2 2025 Programmatic Transparency Benchmark Finds $26.8B in Wasted Programmatic Spend (press release). ANA with TAG TrustNet and Fiducia, 2025-08-14. https://www.ana.net/content/show/id/pr-2025-08-programmatictrans Accessed 2026-09-27. [97] Programmatic Transparency Benchmark: Q3 2025 Benchmark Findings (24-page PDF). ANA, TAG TrustNet, Fiducia; MFA data by DeepSee.io, 2025-11. https://s3.amazonaws.com/media.mediapost.com/uploads/Q32025ProgrammaticTransparencyBenchmark.pdf Accessed 2026-09-27. [98] MFA Ad Spend Is Increasing. Is AI Slop To Blame?. AdExchanger (reporting on ANA Q1 2026 benchmark), 2026-07-08. https://www.adexchanger.com/ai/mfa-ad-spend-is-increasing-is-ai-slop-to-blame/ Accessed 2026-09-27. [99] Going Native: Effects of Disclosure Position and Language on the Recognition and Evaluation of Online Native Advertising. Bartosz W. Wojdynski and Nathaniel J. Evans, University of Georgia; Journal of Advertising 45(2), 2015-12-14. https://www.the7eye.org.il/wp-content/uploads/2016/01/00913367.2015.pdf Accessed 2026-09-27. [100] The effects of disclosure format on native advertising recognition and audience perceptions of legacy and online news publishers. Michelle A. Amazeen and Bartosz W. Wojdynski; Journalism (Sage), online 2018, vol. 21(8) 2020, 2018 (online first); 2020-12 (issue). https://cdn.the7eye.org.il/uploads/2018/04/Amazeen-2017.pdf Accessed 2026-09-27. [101] Reducing Native Advertising Deception: Revisiting the Antecedents and Consequences of Persuasion Knowledge in Digital News Contexts. Michelle A. Amazeen and Bartosz W. Wojdynski; Mass Communication and Society, 2018-10-30. https://api.semanticscholar.org/graph/v1/paper/DOI:10.1080/15205436.2018.1530792?fields=title,abstract,authors,year,publicationDate,venue,externalIds Accessed 2026-09-27. Access: abstract only. [102] Going Native: Can Consumers Recognize Native Advertising? Does It Matter?. David A. Hyman, David J. Franklyn, Calla Yee, Mohammad Hossein Rahmati; Yale Journal of Law and Technology 19, 77-112, 2017. https://yjolt.org/sites/default/files/hyman_19yjolt77_0.pdf Accessed 2026-09-27. [103] Blurred Lines: An Exploration of Consumers' Advertising Recognition in the Contexts of Search Engines and Native Advertising (FTC Staff Report). Federal Trade Commission staff (Bureau of Consumer Protection), 2017-12. https://www.ftc.gov/system/files/documents/reports/blurred-lines-exploration-consumers-advertising-recognition-contexts-search-engines-native/p164504_ftc_staff_report_re_digital_advertising_and_appendices.pdf Accessed 2026-09-27. [104] Consumer Responses to Covert Marketing Communications: A Case of Native Advertising Disclosure in News Contexts. Ilwoo Ju, Hyunmin Lee, Brett Sherrick; Journal of Promotion Management 28(7):1107-1128, 2022-03-27. https://doi.org/10.1080/10496491.2022.2054899 Accessed 2026-09-27. Access: not recorded (added by verifier). [105] The role of theory of mind and edutainment messages in recognizing native advertising: a test of the CARE model. Eun-Joo Choi, Matthew A. Lapierre; International Journal of Advertising 44(5):907-924, 2024-11-04. https://doi.org/10.1080/02650487.2024.2422783 Accessed 2026-09-27. Access: not recorded (added by verifier). [106] Identified disclosure to increase consumers’ detection of native advertising. Shilian and Pe’er, Journal of Economic Psychology, 2025-03. https://doi.org/10.1016/j.joep.2025.102800 Accessed 2026-09-27. Access: partial extract. [107] Hidden in Plain Sight: Understanding Consumer Responses and Ad Recognition of Sponsored Content in Search Engine Gen-AI Overviews. Maggie Liao, Yuan Sun and Nathaniel Evans (University of Georgia), American Academy of Advertising Annual Conference, 2026-03. https://grady.uga.edu/research/hidden-in-plain-sight-understanding-consumer-responses-and-ad-recognition-of-sponsored-content-in-search-engine-gen-ai-overviews/ Accessed 2026-09-27. Access: partial extract. [108] A Meta-Analysis of the Effects of Disclosing Sponsored Content. Martin Eisend, Eva A. van Reijmersdal, Sophie C. Boerman, Farid Tarrahi; Journal of Advertising 49(3), 344-366, 2020-06-23. https://pure.uva.nl/ws/files/53919254/A_Meta_Analysis_of_the_Effects_of_Disclosing_Sponsored_Content.pdf Accessed 2026-09-27. [109] Ad Disclosure vs. Ad Recognition: How Persuasion Knowledge Influences Native Advertising Evaluation. A.-Reum Jung and Jun Heo; Journal of Interactive Advertising 19(1), 2019-01-02. https://api.semanticscholar.org/graph/v1/paper/DOI:10.1080/15252019.2018.1520661?fields=title,abstract,authors,year,publicationDate,venue,externalIds Accessed 2026-09-27. Access: abstract only. [110] Native Advertisement: An Integrated Systematic Review and Research Agenda. Shubham Gupta, Deepa Halder, Sheela Narang; International Journal of Consumer Studies, 2026-03-01. https://api.semanticscholar.org/graph/v1/paper/DOI:10.1111/ijcs.70201?fields=title,abstract,authors,year,publicationDate,venue,externalIds Accessed 2026-09-27. Access: abstract only. [111] Consumers' perception of native advertising on social media - A systematic literature review. Maike Huebner, Joerg Henseler, Julia Thalmann; Journal of Marketing Communications, 2026-04-13. https://api.semanticscholar.org/graph/v1/paper/DOI:10.1080/13527266.2026.2648839?fields=title,abstract,authors,year,publicationDate,venue,externalIds Accessed 2026-09-27. Access: access error; text or metadata via another route. [112] Disclosure-Driven Recognition of Native Advertising: A Test of Two Competing Mechanisms. Eva A. van Reijmersdal, Eline Brussee, Nathaniel Evans, Bartosz W. Wojdynski; Journal of Interactive Advertising, 2023-01-05. https://api.semanticscholar.org/graph/v1/paper/DOI:10.1080/15252019.2022.2146991?fields=title,abstract,authors,year,publicationDate,venue,externalIds Accessed 2026-09-27. Access: abstract only. [113] The Role of a Companion Banner and Sponsorship Transparency in Recognizing and Evaluating Article-style Native Advertising. Colin Campbell and Nathaniel Evans; Journal of Interactive Marketing, 2018-08-01. https://api.semanticscholar.org/graph/v1/paper/DOI:10.1016/j.intmar.2018.02.002?fields=title,abstract,authors,year,publicationDate,venue,externalIds Accessed 2026-09-27. Access: abstract only. [114] Sponsorship Disclosure and Consumer Deception: Experimental Evidence from Native Advertising in Mobile Search. Navdeep S. Sahni and Harikesh S. Nair, Stanford GSB; Marketing Science 39(1):5-32, 2020-01. https://ideas.repec.org/a/inm/ormksc/v39y2020i1p5-32.html Accessed 2026-09-27. Access: not recorded (added by verifier). [115] Native Advertising in Online News: Trade-Offs Among Clicks, Brand Recognition, and Website Trustworthiness. Anocha Aribarg and Eric M. Schwartz, University of Michigan; Journal of Marketing Research, 2019-11-10. https://api.semanticscholar.org/graph/v1/paper/DOI:10.1177/0022243719879711?fields=title,abstract,authors,year,publicationDate,venue,externalIds,citationCount Accessed 2026-09-27. Access: abstract only. [116] Blending in or standing out? The disclosure dilemma of ad cues of social media native advertising. Maike Huebner, Julia Thalmann, Joerg Henseler (Ruhr West University of Applied Sciences; University of Twente; Universidade Nova de Lisboa); Frontiers in Psychology 16:1636910, 2025-08-13. https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2025.1636910/full Accessed 2026-09-27. [117] Outbrain Study Shows Added Trust In Native Ads, Extends Partnership With SPH (Publishers Daily). MediaPost reporting on an Outbrain-commissioned study conducted with Lumen Research, 2019-11-27. https://www.mediapost.com/publications/article/343916/outbrain-study-shows-added-trust-in-native-ads-ex.html Accessed 2026-09-27. [118] Article or Ad? When It Comes to Native, No One Knows. Contently (content-marketing vendor) consumer survey, 2015-09-08. https://contently.com/2015/09/08/article-or-ad-when-it-comes-to-native-no-one-knows/ Accessed 2026-09-27. [119] Bad News: Clickbait and Deceptive Ads on News and Misinformation Websites. Eric Zeng, Tadayoshi Kohno, Franziska Roesner, University of Washington; Workshop on Technology and Consumer Protection (ConPro '20), 2020-05. https://badads.cs.washington.edu/files/Zeng-ConPro2020-BadNews.pdf Accessed 2026-09-27. [120] What Makes a 'Bad' Ad? User Perceptions of Problematic Online Advertising. Eric Zeng, Tadayoshi Kohno, Franziska Roesner; CHI Conference on Human Factors in Computing Systems 2021, 2021-05. https://homes.cs.washington.edu/~yoshi/papers/Zeng-CHI2021-BadAds.pdf Accessed 2026-09-27. [121] Polls, Clickbait, and Commemorative $2 Bills: Problematic Political Advertising on News and Media Websites Around the 2020 U.S. Elections. Eric Zeng, Miranda Wei, Theo Gregersen, Tadayoshi Kohno, Franziska Roesner; ACM Internet Measurement Conference 2021, 2021-11. https://badads.cs.washington.edu/files/Zeng-IMC2021-PoliticalAds.pdf Accessed 2026-09-27. [122] Exploring Political Ads on News and Media Websites During the 2024 U.S. Elections. Emi Yoshikawa and Franziska Roesner, University of Washington; arXiv 2503.02886, 2025-02-11. https://arxiv.org/html/2503.02886 Accessed 2026-09-27. [123] FTC Seeks to Halt 10 Operators of Fake News Sites from Making Deceptive Claims About Acai Berry Weight Loss Products. Federal Trade Commission press release, 2011-04-19. https://www.ftc.gov/news-events/news/press-releases/2011/04/ftc-seeks-halt-10-operators-fake-news-sites-making-deceptive-claims-about-acai-berry-weight-loss Accessed 2026-09-27. [124] Paid to Look Like Truth: The Prevalence and Dark Patterns of Advertorials in News Outlets. Emmanouil Papadogiannakis, Panagiotis Papadopoulos, Nicolas Kourtellis, Evangelos Markatos; arXiv 2602.12810, 2026-02-13 (updated 2026-09-14). https://arxiv.org/abs/2602.12810 Accessed 2026-09-27. Access: not recorded (added by verifier). [125] Enforcement Policy Statement on Deceptively Formatted Advertisements. US Federal Trade Commission, 2015-12-22. https://www.ftc.gov/system/files/documents/public_statements/896923/151222deceptiveenforcement.pdf Accessed 2026-09-27. [126] How content recommendation provider Taboola used LoAF to improve INP up to 36% for its publisher partner websites. web.dev (Google) case study with Taboola, 2024-02-01 (last updated). https://web.dev/case-studies/taboola-inp Accessed 2026-09-27. [127] An Incremental Learning framework for Large-scale CTR Prediction (RecSys '22, DOI 10.1145/3523227.3547390). Katsileros, Mandilaras, Mallis, Pitsikalis, Theodorakis (Deeplab, Greece) and Chamiel (Taboola, Israel), 2022-09-01 (arXiv); RecSys 2022-09-18/23. https://arxiv.org/abs/2209.00458 Accessed 2026-09-27. [128] Deep density networks and uncertainty in recommender systems. Zeldes, Theodorakis, Solodnik, Rotman, Chamiel, Friedman (Taboola, Tel Aviv; DeepLab, Athens), 2017-11-07 (v1); revised 2018-05-06. https://arxiv.org/abs/1711.02487 Accessed 2026-09-27. [129] Scaling TensorFlow to 300 million predictions per second (RecSys '21, DOI 10.1145/3460231.3474605). Jan Hartman, Davorin Kopič (Zemanta, an Outbrain company, Slovenia), 2021-09-20 (arXiv); RecSys 2021-09-27/10-01. https://arxiv.org/abs/2109.09541 Accessed 2026-09-27. [130] Exploration with Model Uncertainty at Extreme Scale in Real-Time Bidding (RecSys '22, DOI 10.1145/3523227.3547383). Jan Hartman, Davorin Kopič (Zemanta, an Outbrain company), 2022-08-03 (arXiv); RecSys 2022-09-18/23. https://arxiv.org/abs/2208.01951 Accessed 2026-09-27. [131] Soft Frequency Capping for Improved Ad Click Prediction in Yahoo Gemini Native (CIKM '19). Aharon, Kaplan, Levy, Somekh, Blanc, Eshel, Shahar, Singer, Zlotnik (Yahoo Research), 2019-11 (CIKM 2019, Beijing); arXiv 2023-12-08. https://arxiv.org/abs/2312.05052 Accessed 2026-09-27. Access: abstract only. [132] OFF-Set: One-pass Factorization of Feature Sets for Online Recommendation in Persistent Cold Start Settings (RecSys 2013). Aharon, Aizenberg, Bortnikov, Lempel, Adadi, Benyamini, Levin, Roth, Serfaty (Yahoo Labs), 2013-08-08 (arXiv); RecSys 2013. https://arxiv.org/abs/1308.1792 Accessed 2026-09-27. Access: abstract only. [133] Carousel Ads Optimization in Yahoo Gemini Native (KDD 2019). Aharon, Somekh, Shahar, Singer, Trayvas, Vogel, Dobrev (Yahoo Research), 2019 (KDD 2019; Yahoo page shows 2019-01-01 default). https://www.yahooinc.com/research/publications/carousel-ads-optimization-in-yahoo-gemini-native Accessed 2026-09-27. Access: abstract only. [134] Conversion-Based Dynamic-Creative-Optimization in Native Advertising (IEEE Big Data 2022). Kaplan, Koren, Shtoff, Shadi, Somekh (Yahoo), 2022-11-13 (arXiv); IEEE Big Data 2022. https://arxiv.org/abs/2211.11524 Accessed 2026-09-27. Access: abstract only. [135] Unbiased Filtering Of Accidental Clicks in Verizon Media Native Advertising (CIKM 2021). Kaplan, Krasne, Shtoff, Somekh (Verizon Media), 2021 (CIKM 2021); arXiv 2023-12-08. https://arxiv.org/abs/2312.05017 Accessed 2026-09-27. Access: abstract only. [136] Setting Your Bid Strategy (Realize Advertiser Help Center). Taboola / Realize help center, 2026-07-23 (last updated). https://realize.com/help/en/articles/10441744-setting-your-bid-strategy Accessed 2026-09-27. [137] What is Conversion Bid Strategy and what are its best practices?. Outbrain advertiser help center, undated. https://www.outbrain.com/help/advertisers/conversion-bid-strategy-best-practices/ Accessed 2026-09-27. [138] Which Conversion Bid Strategy is the best fit for my campaign objectives?. Outbrain advertiser help center, undated. https://www.outbrain.com/help/advertisers/which-bid-strategy-is-right-for-my-campaign-objectives/ Accessed 2026-09-27. [139] What is Target CPA Conversion Bid Strategy?. Outbrain advertiser help center, undated. https://www.outbrain.com/help/advertisers/what-is-target-cpa-conversion-bid-strategy/ Accessed 2026-09-27. [140] What is Engagement Bid Strategy, and how do I use it?. Outbrain advertiser help center, undated. https://www.outbrain.com/help/advertisers/ebs/ Accessed 2026-09-27. [141] Introducing Engagement Bid Strategy, an AI-Based Campaign Optimization Tool Built for a Cookieless World. Outbrain blog, undated on page (shown only as a relative '4 years ago'; approx. 2021-2022). https://www.outbrain.com/blog/outbrain-engagement-bid-strategy/ Accessed 2026-09-27. [142] Smarter Optimization, Less Guesswork: Introducing CPA Tune. MGID blog, 2025-06-25. https://www.mgid.com/blog/introducing-cpa-tune Accessed 2026-09-27. [143] Target CPA Bidding Strategy (MGID Help Center). MGID help center, undated (copyright 2026). https://help.mgid.com/target-cpa-bidding-strategy Accessed 2026-09-27. [144] MaxConversions Bidding Strategy (MGID Help Center). MGID help center, undated. https://help.mgid.com/maxconversions-bidding-strategy Accessed 2026-09-27. [145] Setting Your Bid Strategy: Maximize Conversions (Realize Advertiser Help Center). Taboola / Realize help center, 2026-08-02 (last updated). https://realize.com/help/en/articles/8936549-setting-your-bid-strategy-maximize-conversions Accessed 2026-09-27. [146] Bidding Strategies and Budgets (Outbrain DSP Help Center). Outbrain DSP (Zemanta) help center on Intercom, undated. https://intercom.help/outbrain_dsp/en/articles/8225540-bidding-strategies-and-budgets Accessed 2026-09-27. [147] Setting Your Bid Strategy: Enhanced CPC (Realize Advertiser Help Center). Taboola / Realize help center, 2026-07-23 (last updated). https://realize.com/help/en/articles/8936522-setting-your-bid-strategy-enhanced-cpc Accessed 2026-09-27. Access: not recorded (added by verifier). [148] Creative Expedition: Launch New Creative Assets Faster (Realize Advertiser Help Center). Taboola / Realize help center, 2025-04-27 (last updated). https://realize.com/help/en/articles/5660598-creative-expedition-launch-new-creative-assets-faster Accessed 2026-09-27. Access: not recorded (added by verifier). [149] Programmatic Activation (Realize Advertiser Help Center). Taboola / Realize help center, 2025-04-27 (last updated). https://realize.com/help/en/articles/10441723-programmatic-activation Accessed 2026-09-27. Access: not recorded (added by verifier). [150] OpenRTB 2.6 specification (BidRequest object: at, tmax; Native object). IAB Tech Lab (GitHub InteractiveAdvertisingBureau/openrtb2.x, main branch), 2.6 (living document on main branch; version date not captured). https://raw.githubusercontent.com/InteractiveAdvertisingBureau/openrtb2.x/main/2.6.md Accessed 2026-09-27. [151] GenAI Ad Maker (Realize Advertiser Help Center). Taboola / Realize help center, 2025-05-22 (last updated). https://realize.com/help/en/articles/10449136-genai-ad-maker Accessed 2026-09-27. [152] Taboola Launches Realize+, an Agentic AI System Turning Advertiser Goals Into Outcomes and Expands Agentic Capabilities by Opening its Platform to Claude Skills. Taboola press release, 2026-04-23. https://www.taboola.com/press-releases/realize-plus/ Accessed 2026-09-27. [153] Taboola Launches Realize+, an Agentic AI System Turning Advertiser Goals Into Outcomes and Expands Agentic Capabilities by Opening its Platform to Claude Skills. Taboola (via GlobeNewswire), 2026-04-23. https://www.globenewswire.com/news-release/2026/04/23/3279988/31297/en/taboola-launches-realize-an-agentic-ai-system-turning-advertiser-goals-into-outcomes-and-expands-agentic-capabilities-by-opening-its-platform-to-claude-skills.html Accessed 2026-09-27. [154] Unleashing creativity: exploring MGID's generative AI features. MGID, 2024-02-15. https://www.mgid.com/blog/unleashing-creativity-exploring-mgid-s-generative-ai-features Accessed 2026-09-27. [155] Performance Prediction in MGID Ads (MGID Help Center). MGID help center, undated. https://help.mgid.com/performance-prediction-in-mgid-ads Accessed 2026-09-27. [156] Teads Launches AI-Powered Solutions to Help Advertisers and Publishers Achieve Full-Funnel Success. Teads press release, 2023-10-16. https://www.teads.com/teads-connects-ai-all-screens/ Accessed 2026-09-27. [157] Teads Predictive AI: Shaping the Future of Creative Effectiveness. Teads blog, 2025-11-18. https://www.teads.com/blog/teads-predictive-ai-shaping-the-future-of-creative-effectiveness/9284/ Accessed 2026-09-27. [158] Advertisers - Teads Ad Manager (product page). Teads (Teads Holding Co.), undated (accessed 2026-09-27). https://www.teads.com/teads-ad-manager/ Accessed 2026-09-27. [159] New Study: AI Ads Match Human Creative in Major Report from Columbia University, Harvard University, Technical University of Munich, and Carnegie Mellon University; Taboola Data Shows AI Wins by Appearing Authentically Human. Taboola press release, 2026-01-28. https://www.taboola.com/press-releases/genai-ads-study-2026/ Accessed 2026-09-27. [160] AI in disguise: How AI-generated ads' visual cues shape consumer perception and performance (working paper; SSRN 5096969). Yannick Exner, Jochen Hartmann, Oded Netzer, Shunyuan Zhang (TUM, Columbia, Harvard), 2025-01-14. https://business.columbia.edu/sites/default/files-efs/citation_file_upload/ssrn-5096969.pdf Accessed 2026-09-27. [161] How Taboola Handles Deep Learning Training-Serving Features Discrepancies (Sherlock). Itai Barel, Taboola Engineering blog, 2025-01-14. https://www.taboola.com/engineering/discrepancy-solution/ Accessed 2026-09-27. [162] The Unfavorable Economics of Measuring the Returns to Advertising. Randall A. Lewis; Justin M. Rao — Quarterly Journal of Economics 130(4): 1941-1973, 2015-07-06. https://doi.org/10.1093/qje/qjv023 Accessed 2026-09-27. Access: abstract only. [163] A Comparison of Approaches to Advertising Measurement: Evidence from Big Field Experiments at Facebook. Brett R. Gordon; Florian Zettelmeyer; Neha Bhargava; Dan Chapsky — Marketing Science 38(2): 193-225, 2019-03. https://ideas.repec.org/a/inm/ormksc/v38y2019i2p193-225.html Accessed 2026-09-27. Access: abstract only. [164] Close Enough? A Large-Scale Exploration of Non-Experimental Approaches to Advertising Measurement. Brett R. Gordon; Robert Moakler; Florian Zettelmeyer — Marketing Science 42(4), 2023; arXiv:2201.07055, 2022-01-18 (arXiv v1); revised 2022-10-04; journal 2023. https://arxiv.org/abs/2201.07055 Accessed 2026-09-27. Access: abstract only. [165] Here, There, and Everywhere: Correlated Online Behaviors Can Lead to Overestimates of the Effects of Advertising. Randall A. Lewis; Justin M. Rao; David H. Reiley — Yahoo! Research; WWW 2011 (Hyderabad), ACM 978-1-4503-0632-4/11/03, 2011-03-28. https://davidreiley.com/papers/HereThereEverywhere.pdf Accessed 2026-09-27. [166] Brand Lift Study: Taboola Outperforms at Boosting Mid-Funnel KPIs (Taboola–Kantar meta-analysis). Taboola (Editorial Team) with Kantar, 2023-01-11. https://www.taboola.com/taboola-kantar-brand-lift-study/ Accessed 2026-09-27. [167] Outbrain True Engagement Brand Lift Study: Premium automotive brand increases brand awareness by 23 percent with Outbrain. Outbrain Inc. with PHD Germany; survey by Nielsen commissioned by Outbrain, 2018-02. https://www.outbrain.com/blog/wp-content/uploads/2018/02/OB-Brandlift-PHD-Germany-UK-1.pdf Accessed 2026-09-27. [168] Taboola Launches Realize ID to Overcome Data Fragmentation and Drive Incremental Growth for Performance Marketers; Advertisers See 2.4x Uplift in Conversion Efficiency. Taboola Inc. (press release), 2026-09-22. https://www.taboola.com/press-releases/realizeid/ Accessed 2026-09-27. [169] Incrementality: How Do We Measure It? (Realize marketing hub). Realize (Taboola), 2025-10-27. https://realize.com/marketing-hub/incrementality/ Accessed 2026-09-27. [170] Incrementality Measurement with Measured and Taboola To Inform Marketing Attribution (integration page). Measured (measurement vendor), undated (current as of access). https://www.measured.com/integration/taboola/ Accessed 2026-09-27. [171] Guidelines for Incremental Measurement in Commerce Media. IAB and IAB Europe (IAB Commerce Board; Incrementality Task Force), 2025-11-03. https://www.iab.com/guidelines/guidelines-for-incremental-measurement-in-commerce-media/ Accessed 2026-09-27. [172] How to confidently open a new channel using testing and marketing mix model (MMM). Adrian White, Violet Growth, undated (describes a February 2024 test). https://www.violetgrowth.com/playbooks/how-to-confidently-open-a-new-channel-using-testing-and-marketing-mix-model-mmm-oyfrg Accessed 2026-09-27. [173] An Experimental Investigation of the Effects of Retargeted Advertising: The Role of Frequency and Timing. Navdeep S. Sahni; Sridhar Narayanan; Kirthi Kalyanam — Journal of Marketing Research 56(3), 2019, 2019 (online 2018). https://doi.org/10.1177/0022243718813987 Accessed 2026-09-27. Access: abstract only. [174] When Does Retargeting Work? Information Specificity in Online Advertising. Anja Lambrecht; Catherine E. Tucker — Journal of Marketing Research 50(5), 2013, 2013. https://doi.org/10.1509/jmr.11.0503 Accessed 2026-09-27. Access: abstract only. [175] Consumer Heterogeneity and Paid Search Effectiveness: A Large Scale Field Experiment (NBER WP 20171; Econometrica 83(1): 155-174, 2015). Thomas Blake; Chris Nosko; Steven Tadelis — NBER / Econometrica, 2014-05 (NBER); 2015 (Econometrica). https://www.nber.org/papers/w20171 Accessed 2026-09-27. Access: abstract only. [176] TV Advertising Effectiveness and Profitability: Generalizable Results From 288 Brands. Bradley T. Shapiro; Günter J. Hitsch; Anna E. Tuchman — Econometrica 89(4), 2021, 2021. https://doi.org/10.3982/ecta17674 Accessed 2026-09-27. Access: abstract only. [177] Ghost Ads: Improving the Economics of Measuring Online Ad Effectiveness. Garrett A. Johnson; Randall A. Lewis; Elmar I. Nubbemeyer — Journal of Marketing Research 54(6): 867-884, 2017-12. https://doi.org/10.1509/jmr.15.0297 Accessed 2026-09-27. Access: abstract only. [178] About conversion windows — Google Ads Help. Google, undated help page (current as of access). https://support.google.com/google-ads/answer/3123169 Accessed 2026-09-27. [179] Understand your conversion tracking data (view-through conversions) — Google Ads Help. Google, undated help page (current as of access). https://support.google.com/google-ads/answer/6270625 Accessed 2026-09-27. [180] Meta restricts attribution windows and data retention in Ads Insights API. Luis Rijo — PPC Land (reporting on Meta Developer Blog post by Chris Cutlip, 2025-10-16), 2025-10 (13-16 October 2025). https://ppc.land/meta-restricts-attribution-windows-and-data-retention-in-ads-insights-api/ Accessed 2026-09-27. [181] Defining conversions (Taboola Pixel): attribution windows and conversion types. Taboola developer documentation, 2026-02-26. https://developers.taboola.com/pixel/docs/defining-conversions Accessed 2026-09-27. [182] The Cyber Week Incrementality Report: How CTV, YouTube, and Paid Social Drive ROI. Ike Armstrong; Chandler Dutton — Haus (measurement vendor), 2025-12-11. https://www.haus.io/blog/the-cyber-week-incrementality-report-how-ctv-youtube-and-paid-social-drive-roi Accessed 2026-09-27. [183] MRC Outcomes and Data Quality Standards (Final). Media Rating Council; sponsoring associations ANA, 4A’s and ACA, 2022-09. https://www.mediaratingcouncil.org/sites/default/files/Standards/MRC%20Outcomes%20and%20Data%20Quality%20Standards%20%28Final%29.pdf Accessed 2026-09-27. [184] IAB/MRC Retail Media Measurement Guidelines (Understanding MRC Measurement Standards and Business Requirements for Retail Media Organizations). IAB; Media Rating Council (MRC), 2024-01 (PDF metadata last modified 2023-09-07; cover shows September 2023 / January 2024). https://www.iab.com/wp-content/uploads/2024/01/IAB_Retail_Media_Measurement_Guidelines_January2024.pdf Accessed 2026-09-27. [185] Pixel Conversion Tracking. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/conversion-tracking Accessed 2026-09-27. [186] Nativo Conversion Pixels (advertiser docs). Nativo, 2026. https://resources.nativo.com/docs/nativo-conversion-pixels Accessed 2026-09-27. [187] Meridian documentation — ROI priors and calibration. Google (Meridian open-source MMM), undated documentation (current as of access). https://developers.google.com/meridian/docs/advanced-modeling/roi-priors-and-calibration Accessed 2026-09-27. [188] Meridian overview (Meridian GeoX; calibrate MMMs with experiments). Google, undated (current as of access). https://developers.google.com/meridian Accessed 2026-09-27. [189] Robyn documentation — Features (calibration, MAPE.LIFT, adstock, ridge regression) and home page. Meta Marketing Science (Robyn open-source MMM), undated documentation (current as of access). https://facebookexperimental.github.io/Robyn/docs/features Accessed 2026-09-27. [190] Calibrating Marketing Mix Modeling with Incrementality Experiments for Cross-Channel Understanding. Meta for Business (news/insights page), 2025-10-29. https://www.facebook.com/business/news/calibrating-MMM-with-incrementality Accessed 2026-09-27. [191] Modernizing MMM: Best Practices for Marketers. IAB (shaped with MMM providers and practitioners), 2025-12. https://www.iab.com/wp-content/uploads/2025/12/IAB_Modernizing_MMM_Best_Practices_for_Marketers_December_2025.pdf Accessed 2026-09-27. [192] 75% of marketers say their measurement systems are falling short (coverage of IAB / BWG Global 'State of Data 2026'). Constantine von Hoffman — MarTech, 2026-02-03. https://martech.org/75-of-marketers-say-their-measurement-systems-are-falling-short/ Accessed 2026-09-27. [193] With Retail Media Experiencing Meteoric Growth, ANA Survey Finds Marketers Are Optimistic But Cautious (report: Retail Media Networks: Optimism Tempered with Caution). Association of National Advertisers (ANA), 2024-07-23. https://www.ana.net/content/show/id/pr-2024-07-rmn Accessed 2026-09-27. [194] Spam policies for Google web search (site reputation abuse; link spam). Google Search Central, 2026-08-28 (last updated). https://developers.google.com/search/docs/essentials/spam-policies Accessed 2026-09-27. [195] Taboola Acquiring Connexity, Bringing Personalized e-Commerce Recommendations To The Open Web (press release). Taboola, 2021-07-23. https://www.taboola.com/press-releases/taboola-acquires-connexity/ Accessed 2026-09-27. [196] Time to launch e-commerce section run by suggested content platform Taboola. Press Gazette, 2023-02-14. https://pressgazette.co.uk/marketing/time-taboola-ecommerce-deal/ Accessed 2026-09-27. [197] Taboola Turnkey Commerce to Power New AP Buyline Offering From The Associated Press (press release). Taboola, 2024-03-12. https://www.taboola.com/press-releases/apbuyline/ Accessed 2026-09-27. [198] Skimlinks homepage. Skimlinks (a Taboola company), 2026 (undated page). https://skimlinks.com/ Accessed 2026-09-27. [199] Payments, commissions and costs: how publishers make money with Skimlinks. Skimlinks (a Taboola company), undated. https://www.skimlinks.com/insights/faq-skimlinks-payments/ Accessed 2026-09-27. [200] How much does Skimlinks charge?. Skimlinks (Taboola), 2026-09-07 (updated). https://support.skimlinks.com/hc/en-us/articles/223835508-How-much-does-Skimlinks-charge Accessed 2026-09-27. [201] Taboola.com Ltd. FY2025 10-K, XBRL page R6 (Consolidated Statements of Comprehensive Income (Loss), partial). Taboola.com Ltd. / SEC EDGAR, 2026-02-25. https://www.sec.gov/Archives/edgar/data/1840502/000184050226000004/R6.htm Accessed 2026-09-27. Access: partial extract. [202] Teads Holding Co. Form 10-Q for quarter ended March 31, 2026. Teads Holding Co. / SEC EDGAR, 2026-05. https://www.sec.gov/Archives/edgar/data/1454938/000145493826000037/tead-20260331.htm Accessed 2026-09-27. [203] Pentaleap and Teads Partner to Break New Ground in Programmatic Retail Media (press release). Pentaleap Inc. and Teads, 2025-07-24. https://www.pentaleap.com/press-release/pentaleap-and-teads-partner-to-break-new-ground-in-programmatic-retail-media Accessed 2026-09-27. [204] Pentaleap Insights page (news and press listing). Pentaleap Inc., 2026 (undated listing). https://www.pentaleap.com/insights Accessed 2026-09-27. [205] Criteo Reports Fourth Quarter 2025 Results (press release). Criteo S.A., 2026-02-11. https://criteo.investorroom.com/2026-02-11-CRITEO-REPORTS-FOURTH-QUARTER-2025-RESULTS Accessed 2026-09-27. [206] Criteo S.A. Form 10-K for fiscal year ended December 31, 2025 (business section). Criteo S.A. / SEC EDGAR, 2026 (exact filing date not read). https://www.sec.gov/Archives/edgar/data/1576427/000157642726000014/crto-20251231.htm Accessed 2026-09-27. Access: partial extract. [207] Commerce Grid - The world's first commerce SSP, built by Criteo (product page). Criteo, 2026 (undated page). https://www.criteo.com/platform/commerce-grid/ Accessed 2026-09-27. [208] Amazon Publisher Cloud (product page). Amazon Publisher Services, undated. https://aps.amazon.com/aps/amazon-publisher-cloud/ Accessed 2026-09-27. [209] Amazon Publisher Services homepage. Amazon Publisher Services, undated. https://aps.amazon.com/aps/index.html Accessed 2026-09-27. [210] Pinterest, BuzzFeed, and more apps and websites will start to show Amazon Sponsored Product ads. Amazon (aboutamazon.com newsroom), 2023-08 (date not shown on page; dated 2023-08-17 by Marketing Dive). https://www.aboutamazon.com/news/company-news/pinterest-buzzfeed-and-more-apps-will-show-amazon-sponsored-product-ads-to-discover-relevant-products Accessed 2026-09-27. Access: not recorded (added by verifier). [211] Sponsored Products (product page). Amazon Ads, undated. https://advertising.amazon.com/solutions/products/sponsored-products Accessed 2026-09-27. [212] Amazon Adds Offsite Controls to Sponsored Products: What You Need to Know. Feedvisor (Amazon ad-tech vendor blog), 2025-08-19. https://feedvisor.com/resources/amazon-trends/amazon-adds-offsite-controls-to-sponsored-products/ Accessed 2026-09-27. [213] Amazon's Sponsored Products Now Go Off-Site: What You Need to Know. JumpFly (agency blog), 2025-07-15. https://www.jumpfly.com/blog/amazons-sponsored-products-now-go-off-site-what-you-need-to-know/ Accessed 2026-09-27. [214] Amazon Marketing Cloud: Discover advertising insights (product page). Amazon Ads, undated. https://advertising.amazon.com/solutions/products/amazon-marketing-cloud Accessed 2026-09-27. [215] People Incorporated Reports Q2 2026 (Exhibit 99.1 press release). People Incorporated (formerly IAC Inc.) / SEC EDGAR, 2026-08-03. https://www.sec.gov/Archives/edgar/data/1800227/000162828026051836/ex_991q22026ppli-pressrele.htm Accessed 2026-09-27. [216] People Inc. (IAC Inc.) Form 10-Q for quarter ended June 30, 2026. People Incorporated / SEC EDGAR, 2026-08. https://www.sec.gov/Archives/edgar/data/0001800227/000162828026051881/ppli-20260630.htm Accessed 2026-09-27. Access: partial extract. [217] BuzzFeed, Inc. Reports Q4 and Full Year 2025 Results (press release). BuzzFeed, Inc., 2026-03-12. https://www.buzzfeed.com/buzzfeedpress/buzzfeed-inc-reports-q4-and-full-year-2025-results Accessed 2026-09-27. [218] BuzzFeed, Inc. Form 10-K for fiscal year ended December 31, 2025 (revenue definitions, risk factors). BuzzFeed, Inc. / SEC EDGAR, 2026-03-12. https://www.sec.gov/Archives/edgar/data/1828972/000182897226000030/bzfd-20251231.htm Accessed 2026-09-27. Access: partial extract. [219] Future plc FY 2025 Full year results presentation (4 December 2025). Future plc, 2025-12-04. https://futureplc.com/wp-content/uploads/2025/12/FINAL-FY-2025-results-for-upload.pdf Accessed 2026-09-27. [220] How to import conversions to Outbrain (help article). Outbrain / Teads, undated. https://www.outbrain.com/help/advertisers/offline-conversions/ Accessed 2026-09-27. [221] How does Outbrain's pixel track conversions? (help article). Outbrain / Teads, undated. https://www.outbrain.com/help/advertisers/how-does-outbrains-pixel-track-conversions/ Accessed 2026-09-27. [222] Teads Conversion Pixel (Shopify App Store listing). Outbrain (developer) / Shopify, 2025-11-20. https://apps.shopify.com/teads-conversion-pixel Accessed 2026-09-27. [223] How do I set up server-to-server integrations? (Outbrain help). Outbrain / Teads, undated. https://www.outbrain.com/help/advertisers/server2server-integrations/ Accessed 2026-09-27. Access: not recorded (added by verifier). [224] Teads advertiser conversion API template for server-side Google Tag Manager. Teads (GitHub), undated repository. https://github.com/teads/teads-advertiser-conversion-api-gtm-template Accessed 2026-09-27. [225] The S2S postback URL (Taboola Pixel documentation). Taboola developers, undated. https://developers.taboola.com/pixel/docs/the-postback-url Accessed 2026-09-27. Access: not recorded (added by verifier). [226] Taboola: Track CRM-based conversions using S2S (Pixel documentation). Taboola developers, undated. https://developers.taboola.com/pixel/docs/s2s-track-crm-conversions Accessed 2026-09-27. [227] Taboola Manual S2S integration (Pixel documentation). Taboola developers, undated. https://developers.taboola.com/pixel/docs/s2s-manual-integration Accessed 2026-09-27. [228] Taboola S2S Overview (Pixel documentation). Taboola developers, undated. https://developers.taboola.com/pixel/docs/s2s-overview Accessed 2026-09-27. [229] Meta Conversions API overview. Meta for Developers, undated. https://developers.facebook.com/docs/marketing-api/conversions-api Accessed 2026-09-27. [230] Meta Conversions API parameters. Meta for Developers, undated. https://developers.facebook.com/docs/marketing-api/conversions-api/parameters Accessed 2026-09-27. [231] Taboola Backstage API documentation index (llms.txt). Taboola developers, undated. https://developers.taboola.com/backstage-api/llms.txt Accessed 2026-09-27. [232] Taboola Campaign Summary report dimensions. Taboola developers, undated. https://developers.taboola.com/backstage-api/docs/campaign-summary-dimensions.md Accessed 2026-09-27. Access: partial extract. [233] Taboola first-party audience onboarding overview (Backstage API). Taboola developers, undated. https://developers.taboola.com/backstage-api/docs/onboarding-overview.md Accessed 2026-09-27. [234] How to use the Outbrain API (Amplify API help page). Outbrain / Teads, undated. https://www.outbrain.com/help/advertisers/amplify-api/ Accessed 2026-09-27. [235] Taboola Backstage API reference welcome page. Taboola developers, undated. https://developers.taboola.com/backstage-api/reference/welcome Accessed 2026-09-27. [236] Guides Concerning the Use of Endorsements and Testimonials in Advertising, 88 FR 48092 (final rule notice). Federal Trade Commission / GovInfo, 2023-07-26. https://www.govinfo.gov/content/pkg/FR-2023-07-26/html/2023-14795.htm Accessed 2026-09-27. Access: access error; text or metadata via another route. [237] The FTC's Endorsement Guides: What People Are Asking (business guidance). Federal Trade Commission, 2023 (revised; undated on page). https://www.ftc.gov/business-guidance/resources/ftcs-endorsement-guides-what-people-are-asking Accessed 2026-09-27. [238] Federal Trade Commission Announces Final Rule Banning Fake Reviews and Testimonials (press release). Federal Trade Commission, 2024-08-14. https://www.ftc.gov/news-events/news/press-releases/2024/08/federal-trade-commission-announces-final-rule-banning-fake-reviews-testimonials Accessed 2026-09-27. [239] The Consumer Reviews and Testimonials Rule: Questions and Answers. Federal Trade Commission, 2024 (undated on page). https://www.ftc.gov/business-guidance/resources/consumer-reviews-testimonials-rule-questions-answers Accessed 2026-09-27. [240] Our approach to advertising and expanding access (OpenAI) — read via r.jina.ai text proxy. OpenAI, not visible in proxied text. https://openai.com/index/our-approach-to-advertising-and-expanding-access/ Accessed 2026-09-27. Access: access error; text or metadata via another route. [241] Testing ads in ChatGPT. OpenAI, 2026-02-09 (originally published; updated through 2026-08-11). https://openai.com/index/testing-ads-in-chatgpt/ Accessed 2026-09-27. [242] OpenAI's ChatGPT ads hit $1 billion as Israel remains outside the global rollout. Calcalistech (ctech), 2026-09-01. https://www.calcalistech.com/ctechnews/article/skt1hw4ofl Accessed 2026-09-27. [243] ChatGPT (Wikipedia article, rendered and raw wikitext). Wikipedia editors, 2026-09 (live article). https://en.wikipedia.org/wiki/ChatGPT Accessed 2026-09-27. [244] Reimagining advertising with AI. OpenAI, 2026-09-16. https://openai.com/index/reimagining-advertising-with-ai/ Accessed 2026-09-27. [245] Sponsored Agents in ChatGPT Ads (help centre). OpenAI, about 2026-09-17 (page showed "Updated: 10 days ago"). https://help.openai.com/en/articles/20001524-sponsored-agents-in-chatgpt-ads Accessed 2026-09-27. [246] OpenAI Sponsored Agent ads: ChatGPT click-to-chat pilot. Relevant Audience (citing Digiday, 14 September 2026), 2026-09. https://www.relevantaudience.com/ai/openai-sponsored-agent-pilot-chatgpt-ads-click-to-chat/ Accessed 2026-09-27. [247] Perplexity brings ads to its platform. TechCrunch, 2024-11-12. https://techcrunch.com/2024/11/12/perplexity-brings-ads-to-its-platform/ Accessed 2026-09-27. Access: not recorded (added by verifier). [248] Perplexity abandons AI advertising strategy over trust worries. MacRumors (citing Financial Times), 2026-02-18. https://www.macrumors.com/2026/02/18/perplexity-abandons-ai-advertising/ Accessed 2026-09-27. [249] Perplexity AI (Wikipedia article, rendered and raw wikitext). Wikipedia editors, 2026-09 (live article). https://en.wikipedia.org/wiki/Perplexity_AI Accessed 2026-09-27. [250] A User Study on the Acceptance of Native Advertising in Generative IR. Ines Zelch, Matthias Hagen, Martin Potthast (Webis); ACM CHIIR 2024, pp. 142-152, 2024-03-08. https://doi.org/10.1145/3627508.3638316 Accessed 2026-09-27. Access: not recorded (added by verifier). [251] Detecting Generated Native Ads in Conversational Search. Sebastian Schmidt, Ines Zelch, Janek Bevendorff, Benno Stein, Matthias Hagen, Martin Potthast (Webis group); arXiv 2402.04889, 2024-02-07 (revised 2024-04-30). https://arxiv.org/abs/2402.04889 Accessed 2026-09-27. Access: abstract only. [252] Teads Conversational AI SDK release notes. Teads developer documentation, 2025-11-12. https://developers.teads.com/docs/ai/News/release-notes/ Accessed 2026-09-27. [253] Dianomi plc: Partnership with AI Company Dappier. Dianomi plc via RNS, 2026-03-16. https://www.investegate.co.uk/announcement/rns/dianomi-plc--dnm/partnership-with-ai-company-dappier/9474700 Accessed 2026-09-27. [254] 73% higher CTRs: why advertisers need to pay attention to conversational AI. Microsoft Advertising, 2025-08-06. https://about.ads.microsoft.com/en/blog/post/august-2025/73-higher-ctrs-why-advertisers-need-to-pay-attention-to-conversational-ai Accessed 2026-09-27. [255] Sponsored Products prompts and Sponsored Brands prompts. Amazon Ads, 2026-03-10. https://advertising.amazon.com/resources/whats-new/unboxed-2025-sponsored-products-and-sponsored-brands-prompts Accessed 2026-09-27. Access: not recorded (added by verifier). [256] How to connect with shoppers using AI agents. Amazon Ads, 2026-08-05. https://advertising.amazon.com/library/expert-advice/brands-ai-agent-shopping Accessed 2026-09-27. Access: not recorded (added by verifier). [257] Google Marketing Live 2025 (The Keyword, Ads & Commerce). Google, 2025-05-21. https://blog.google/products/ads-commerce/google-marketing-live-2025/ Accessed 2026-09-27. [258] Google Marketing Live 2026 (The Keyword, Ads & Commerce). Google, 2026-05-20. https://blog.google/products/ads-commerce/google-marketing-live-2026/ Accessed 2026-09-27. [259] agentic-commerce-protocol/agentic-commerce-protocol (GitHub README). OpenAI and Stripe (founding maintainers), 2026-04-17 (latest spec release). https://github.com/agentic-commerce-protocol/agentic-commerce-protocol Accessed 2026-09-27. [260] OpenAI debuts new ChatGPT 'buy' button and open source Agentic Commerce Protocol. VentureBeat, 2025-09-29. https://venturebeat.com/ai/openai-debuts-new-chatgpt-buy-button-and-open-source-agentic-commerce Accessed 2026-09-27. [261] OpenAI Commerce - Get started guide. OpenAI, 2026. https://developers.openai.com/commerce/guides/get-started Accessed 2026-09-27. [262] OpenAI Commerce developer documentation (index). OpenAI, 2026. https://developers.openai.com/commerce/ Accessed 2026-09-27. [263] Dianomi plc 2025 Annual Report and Financial Statements. Dianomi plc (audited by BDO LLP), 2026-05-29. https://www.dianomi.com/wp-content/uploads/178114-Dianomi-RA.pdf Accessed 2026-09-27. [264] Setting Up Publishers & Placements Targeting (Outbrain DSP Help Center). Outbrain DSP (Teads), 2025-01-13. https://intercom.help/outbrain_dsp/en/articles/8225749-setting-up-publishers-placements-targeting Accessed 2026-09-27. Access: not recorded (added by verifier). [265] Teads Launches its AI-Driven Operating System, Teads EngageOS, to Maximize Publisher Total Revenue. Teads (via GlobeNewswire), 2026-06-11. https://www.globenewswire.com/news-release/2026/06/11/3310407/0/en/teads-launches-its-ai-driven-operating-system-teads-engageos-to-maximize-publisher-total-revenue.html Accessed 2026-09-27. [266] Taboola for Advertisers (Realize product page). Taboola, 2026. https://www.taboola.com/advertisers Accessed 2026-09-27. [267] MGID for Advertisers (product page). MGID, 2026. https://www.mgid.com/advertisers Accessed 2026-09-27. [268] Outbrain Minimum Budget: What a Real Test Costs. OpenAdLibrary Team (openadlibrary.com), 2026-07-22. https://openadlibrary.com/blog/outbrain-minimum-budget/ Accessed 2026-09-27. [269] Help Center: Pricing and Billing Model. MGID, Inc., 2026. https://help.mgid.com/mgids-pricing-and-billing-model Accessed 2026-09-27. [270] How Much Does Native Advertising Cost in 2026? CPC & CPM Benchmarks. MGID, Inc., 2026-08. https://www.mgid.com/blog/how-much-does-native-advertising-cost Accessed 2026-09-27. [271] Teads Holding Co. Announces Second Quarter 2026 Results (Form 8-K, Exhibit 99.1). Teads Holding Co. / SEC EDGAR, 2026-08-06. https://www.sec.gov/Archives/edgar/data/1454938/000145493826000051/q22026er-exhibit991.htm Accessed 2026-09-27. [272] (Author's work) Performance & Native Playbook (performance.astro source; market claims last validated 2026-06-06). Evgeny Popov / No Fluff Advisory, 2026-06-06. https://nofluffadvisory.com/services/playbooks/performance/ Accessed 2026-09-27. [273] Outbrain Completes Change of Corporate Name to Teads (Form 8-K, Exhibit 99.1). Teads Holding Co. / SEC EDGAR, 2025-06-09. https://www.sec.gov/Archives/edgar/data/1454938/000095015725000478/ex99-1.htm Accessed 2026-09-27. [274] Omnicom Completes Acquisition of Interpublic, Forming the World's Leading Marketing and Sales Company (Form 8-K, Exhibit 99.1). Omnicom Group Inc. / SEC EDGAR, 2025-11-26. https://www.sec.gov/Archives/edgar/data/29989/000119312525300788/d24786dex991.htm Accessed 2026-09-27. [275] CEO tells staff 'the future of WPP Media is here' as GroupM rebrand confirmed. The Drum, 2025-05-28. https://www.thedrum.com/news/2025/05/28/ceo-tells-staff-the-future-wpp-media-here-groupm-rebrand-confirmed Accessed 2026-09-27. [276] WPP Media corporate website (homepage). WPP Media Limited, 2026. https://www.wppmedia.com/ Accessed 2026-09-27. [277] Life360 Completes Acquisition of Nativo and Surpasses 50 Million U.S. MAU. Life360, Inc. (via GlobeNewswire), 2026-01-05. https://www.globenewswire.com/news-release/2026/01/05/3212816/0/en/Life360-Completes-Acquisition-of-Nativo-and-Surpasses-50-Million-U-S-MAU.html Accessed 2026-09-27. [278] Reddit Opens Up 'Max' Campaigns To All Advertisers. MediaPost, 2026-09-04. https://www.mediapost.com/publications/article/417702/reddit-opens-up-max-campaigns-to-all-advertisers.html Accessed 2026-09-27. Access: not recorded (added by verifier). [279] Reddit Introduces Max Campaigns, Its New Automated Campaign Type. Search Engine Journal, 2026-01-05. https://www.searchenginejournal.com/reddit-introduces-max-campaigns-its-new-automated-campaign-type/564361/ Accessed 2026-09-27. [280] The Trade Desk Introduces Kokai Zuma, the Latest Release of Kokai. The Trade Desk (press room), 2026-08-27. https://www.thetradedesk.com/press-room/the-trade-desk-introduces-kokai-zuma-the-latest-release-of-kokai Accessed 2026-09-27. [281] Omnicom Media Officially Launches Hearts United. Omnicom Group (omc.com newsroom), 2026-08-28. https://www.omc.com/newsroom/omnicom-media-officially-launches-hearts-united/ Accessed 2026-09-27. [282] Omnicom Media corporate website (homepage). Omnicom Media, 2026. https://www.omnicommedia.com/ Accessed 2026-09-27. [283] Omnicom Media Retires Kinesso & Annalect: Here's What It Means To Aussie Market. B&T, 2026-07-17. https://www.bandt.com.au/omnicom-media-retires-kinesso-annalect-heres-what-it-means-to-aussie-market/ Accessed 2026-09-27. Access: not recorded (added by verifier). [284] Criteo Completes Redomiciliation from France to Luxembourg; Board Approves Next Step Toward U.S. Domicile. Criteo S.A. (investor room), 2026-07-29. https://criteo.investorroom.com/2026-07-29-Criteo-Completes-Redomiciliation-from-France-to-Luxembourg-Board-Approves-Next-Step-Toward-U-S-Domicile Accessed 2026-09-27. [285] Criteo Attracts Takeover Bids From Equity, Investment Firms. MediaPost (citing Bloomberg), 2026-07-06. https://www.mediapost.com/publications/article/416328/criteo-attracts-takeover-bids-from-equity-investm.html Accessed 2026-09-27. Access: not recorded (added by verifier). [286] Media - Publicis Groupe UK. Publicis Groupe UK, 2026. https://www.publicisgroupeuk.com/what-we-do/media/ Accessed 2026-09-27. [287] CitrusAd website (redirecting readers to Epsilon Retail Media). CitrusAd / Epsilon, 2026. https://www.citrusad.com/ Accessed 2026-09-27. [288] Programmatic Transparency Benchmark: Q1 2026 Findings (14-page PDF). ANA, TAG TrustNet, Fiducia (MediaPost-hosted mirror), 2026-05. https://s3.amazonaws.com/media.mediapost.com/uploads/ANAq1ProggrammaticTransparencyBenchmarkReport.pdf Accessed 2026-09-27. Access: not recorded (added by verifier). [289] ANA Programmatic Transparency Benchmark Reveals Widening Gap Between Top and Bottom Advertisers (press release). ANA, 2026-05-27. https://www.ana.net/content/show/id/pr-2026-05-programmatic Accessed 2026-09-27. Access: not recorded (added by verifier). [290] The ANA Q3 2025 Programmatic Transparency Benchmark (press release). ANA, 2025-11-05. https://www.ana.net/content/show/id/pr-2025-11-transparency Accessed 2026-09-27. [291] Why MFA Is Ad Tech's Biggest Problem - But Also Its Easiest To Solve. AdExchanger (James Hercher), quoting Chris Kane, Jounce Media, 2023-10-02. https://www.adexchanger.com/programmatic/why-mfa-is-ad-techs-biggest-problem-but-also-its-easiest-to-solve/ Accessed 2026-09-27. [292] Which Publishers Am I On? (Outbrain advertiser help). Outbrain/Teads help center, undated. https://www.outbrain.com/help/advertisers/publishers-list/ Accessed 2026-09-27. [293] How To Filter and Exclude Publishers (Outbrain advertiser help). Outbrain/Teads help center, undated. https://www.outbrain.com/help/advertisers/exclude-publishers/ Accessed 2026-09-27. [294] What Is the Publisher & Section Exclusion Feature and How Do I Use It? (Outbrain advertiser help). Outbrain/Teads help center, undated. https://www.outbrain.com/help/advertisers/what-is-the-publisher-section-exclusion-feature-and-how-do-i-use-it/ Accessed 2026-09-27. [295] SpendGuard (Realize help center). Taboola Inc. (Realize), 2025-04-27. https://realize.com/help/en/articles/10466464-spendguard Accessed 2026-09-27. [296] Setting Up a New Campaign (Realize help center). Taboola Inc. (Realize), 2026-07-23. https://realize.com/help/en/articles/10473049-setting-up-a-new-campaign Accessed 2026-09-27. [297] Media selection: updated features and how to use them (support). Readpeak, 2025-07-31 (sitemap lastmod). https://www.readpeak.com/support/campaign-setup/media-selection-updated-features-and-how-to-use-them Accessed 2026-09-27. [298] Native Advertising Platform (product page). Readpeak, 2026-09-15. https://www.readpeak.com/native-advertising-platform Accessed 2026-09-27. [299] Looking to Elevate Ad Inventory Quality? Taboola Select Offers a Premium Solution. Taboola (company blog), 2025-01-16. https://www.taboola.com/taboola-updates/mfa-sites-explained/ Accessed 2026-09-27. [300] Made-for-Advertising (MFA) Sites (Realize marketing hub). Taboola Inc. (Realize), 2026-07-03. https://realize.com/marketing-hub/made-for-advertising/ Accessed 2026-09-27. [301] Trust & Authenticity: Our Advertising Guidelines (Outbrain / Teads advertiser guidelines). Outbrain Inc. (renamed Teads Holding Co.) and Outbrain UK Limited, undated. https://www.outbrain.com/advertisers/guidelines/ Accessed 2026-09-27. [302] Landing Page Policies (Realize / Taboola Advertiser Help Center). Taboola Inc. (Realize help center), 2025-04-27. https://realize.com/help/en/articles/3878205-landing-page-policies Accessed 2026-09-27. [303] Affiliate Advertising Guidelines (2-page PDF). Taboola, undated. https://www.taboola.com/affiliate-advertising-guidelines Accessed 2026-09-27. [304] Outbrain For Publishers - Publisher Guidelines. Outbrain Inc. (Teads Holding Co.) and Outbrain UK Limited, undated. https://www.outbrain.com/publishers/guidelines/ Accessed 2026-09-27. [305] Media Transparency: Prescriptions, Principles, and Processes for Marketers - Overview. ANA, 2016-07 (template v2.0 2018-07). https://www.ana.net/content/show/id/industry-initiative-recommendations-overview Accessed 2026-09-27. [306] The ANA And The 4A's Clash Over Media Transparency Guidelines. AdExchanger, 2016-01-29. https://www.adexchanger.com/ad-exchange-news/ana-4as-clash-media-transparency-guidelines/ Accessed 2026-09-27. [307] Vevo Partners with Adelaide to Launch 'Attention Guaranteed'. Vevo (press page), 2026-01-27. https://www.hq.vevo.com/press/vevo-partners-with-adelaide-to-launch-attention-guaranteed Accessed 2026-09-27. [308] Adelaide Releases 2026 Outcomes Guide. Adelaide Metrics, 2026-01-20. https://www.adelaidemetrics.com/blog/adelaide-releases-2026-outcomes-guide Accessed 2026-09-27. [309] Why Attention Is No Longer Just a Metric - It's a Guarantee. Teads (blog), 2025-05-07. https://www.teads.com/blog/guaranteeing-attention-connecting-outcomes/6467/ Accessed 2026-09-27. Access: not recorded (added by verifier). [310] Taboola Data Reveals 76% Advertisers See Improved Performance From Agentic AI. ExchangeWire, 2026-05-13. https://www.exchangewire.com/blog/2026/05/13/taboola-data-reveals-76-advertisers-see-improved-performance-from-agentic-ai/ Accessed 2026-09-27. [311] Taboola's Content Chum Boxes Also Spread Disinformation. Vice, 2021-01-12. https://www.vice.com/en/article/taboolas-content-chum-boxes-also-spread-disinformation/ Accessed 2026-09-27. [312] Media Briefing: Cheap reach inventory is the next target from the consultancy that spearheaded the MFA crackdown. Digiday, 2024-03-07. https://digiday.com/media/media-briefing-cheap-reach-inventory-is-the-next-target-from-the-consultancy-that-spearheaded-the-mfa-crackdown/ Accessed 2026-09-27. [313] FTC Issues Enforcement Policy Statement Addressing 'Native' Advertising and Deceptively Formatted Advertisements (press release). US Federal Trade Commission, 2015-12-22. https://www.ftc.gov/news-events/news/press-releases/2015/12/ftc-issues-enforcement-policy-statement-addressing-native-advertising-deceptively-formatted Accessed 2026-09-27. [314] FTC Approves Final Consent Orders Settling Endorsement and Deceptive Native Advertising Charges against Creaxion Corporation and Inside Publications, LLC. US Federal Trade Commission, 2019-02-08. https://www.ftc.gov/news-events/news/press-releases/2019/02/ftc-approves-final-consent-orders-settling-endorsement-deceptive-native-advertising-charges-against Accessed 2026-09-27. Access: not recorded (added by verifier). [315] Lord & Taylor, LLC, In the Matter of (File No. 152 3181; Docket C-4576). US Federal Trade Commission, 2016-05-23. https://www.ftc.gov/legal-library/browse/cases-proceedings/152-3181-lord-taylor-llc-matter Accessed 2026-09-27. [316] TruHeight (Vanilla Chip LLC), In the Matter of (File No. 242 3093; Docket C-4837). US Federal Trade Commission, 2026-07-15. https://www.ftc.gov/legal-library/browse/cases-proceedings/242-3093-truheight-vanilla-chip-llc-matter Accessed 2026-09-27. [317] Concurring Statement of Chairman Andrew N. Ferguson joined by Commissioner Mark R. Meador, TruHeight, Matter No. 2423093. US Federal Trade Commission, 2026-04-13. https://www.ftc.gov/system/files/ftc_gov/pdf/ferguson-meador-truheight-statement.pdf Accessed 2026-09-27. [318] FTC Legal Library, Cases and Proceedings, search 'fake reviews' (39 results). US Federal Trade Commission, 2026-09-27. https://www.ftc.gov/legal-library/browse/cases-proceedings?search=fake%20reviews Accessed 2026-09-27. Access: partial extract. [319] Text of Regulations, Title 11 Div. 6 Ch. 1 California Consumer Privacy Act Regulations (approved text, 127 pp.). California Privacy Protection Agency, 2025-09. https://cppa.ca.gov/regulations/pdf/ccpa_updates_cyber_risk_admt_appr_text.pdf Accessed 2026-09-27. [320] California Civil Code section 1798.140 (CCPA definitions: cross-context behavioral advertising, sell, share, nonpersonalized advertising). California Legislative Information, 2025. https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml?lawCode=CIV§ionNum=1798.140 Accessed 2026-09-27. [321] California Civil Code section 1798.135 (CCPA/CPRA: Do Not Sell or Share link; opt-out preference signals). California Legislative Information, 2025-01-01. https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml?lawCode=CIV§ionNum=1798.135 Accessed 2026-09-27. [322] CCPA Updates, Cybersecurity Audits, Risk Assessments, Automated Decisionmaking Technology (ADMT), and Insurance Regulations (rulemaking page). California Privacy Protection Agency, 2025-09-22. https://cppa.ca.gov/regulations/ccpa_updates.html Accessed 2026-09-27. [323] California Consumer Privacy Act regulations: rulemaking history. California Privacy Protection Agency, 2023-03-29. https://cppa.ca.gov/regulations/consumer_privacy_act.html Accessed 2026-09-27. [324] AB 566, California Consumer Privacy Act of 2018: opt-out preference signal (California Opt Me Out Act), Chapter 465, Statutes of 2025. California Legislative Information, 2025-10-08. https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260AB566 Accessed 2026-09-27. [325] US State Privacy Legislation Tracker 2025: Comprehensive Consumer Privacy Bills (chart PDF). IAPP, 2025-11-24. https://iapp.org/media/pdf/resource_center/State_Comp_Privacy_Law_Chart.pdf Accessed 2026-09-27. [326] US State Privacy Legislation Tracker (web page). IAPP, 2026-09-08. https://iapp.org/resources/article/us-state-privacy-legislation-tracker/ Accessed 2026-09-27. Access: partial extract. [327] Major Data Privacy Bill Signed into Law (SB 546). Oklahoma House of Representatives, 2026-03-23. https://www.okhouse.gov/posts/news-20260323_2 Accessed 2026-09-27. Access: not recorded (added by verifier). [328] Alabama Becomes 21st State With Comprehensive Consumer Privacy Law. Hunton Andrews Kurth (Privacy and Cybersecurity Law Blog), 2026-04-21. https://www.hunton.com/privacy-and-cybersecurity-law-blog/alabama-becomes-21st-state-with-comprehensive-consumer-privacy-law Accessed 2026-09-27. Access: not recorded (added by verifier). [329] Directive 2005/29/EC (Unfair Commercial Practices Directive), consolidated text of 28 May 2022. EUR-Lex / European Union, 2022-05-28. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:02005L0029-20220528 Accessed 2026-09-27. [330] The Digital Services Act: questions and answers. European Commission (DG CNECT), 2025-12-19. https://digital-strategy.ec.europa.eu/en/faqs/digital-services-act-questions-and-answers Accessed 2026-09-27. [331] The Digital Services Act package (policy page). European Commission (DG CNECT), 2026-09-09. https://digital-strategy.ec.europa.eu/en/policies/digital-services-act-package Accessed 2026-09-27. [332] Regulation (EU) 2022/2065 (Digital Services Act), Articles 19, 26 and 28. European Union, Official Journal L 277, 2022-10-27. https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=OJ%3AL%3A2022%3A277%3AFULL Accessed 2026-09-27. [333] EU AI Act, Article 50: transparency obligations for providers and deployers of certain AI systems (consolidated text). Future of Life Institute, artificialintelligenceact.eu (reproducing Regulation (EU) 2024/1689), 2024-07-12 (regulation published in the Official Journal). https://artificialintelligenceact.eu/article/50/ Accessed 2026-09-27. [334] Legislative Train Schedule: Digital Omnibus on AI. European Parliament, 2026-08-01. https://www.europarl.europa.eu/legislative-train/package-digital-package/file-digital-omnibus-on-ai Accessed 2026-09-27. Access: not recorded (added by verifier). [335] EU AI Act Omnibus Agreement: Postponed High-Risk Deadlines and Other Key Changes. Gibson Dunn, 2026-05-27. https://www.gibsondunn.com/eu-ai-act-omnibus-agreement-postponed-high-risk-deadlines-and-other-key-changes/ Accessed 2026-09-27. Access: not recorded (added by verifier). [336] Commission starts enforcing AI Act rules and new transparency requirements on 2 August (press release). European Commission, 2026-07-31. https://digital-strategy.ec.europa.eu/en/news/commission-starts-enforcing-ai-act-rules-and-new-transparency-requirements-2-august Accessed 2026-09-27. [337] AI Act (Regulatory framework for AI) policy page. European Commission (DG CNECT), 2026-08-03. https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai Accessed 2026-09-27. [338] Guidelines on transparency obligations for providers and deployers of certain AI systems (policy page). European Commission (AI Office), 2026-08-06. https://digital-strategy.ec.europa.eu/en/policies/guidelines-transparency-ai-generated-content Accessed 2026-09-27. [339] EU agrees to simplify AI rules, boost innovation and ban nudification apps to protect citizens (news). European Commission, 2026-05-07. https://digital-strategy.ec.europa.eu/en/news/eu-agrees-simplify-ai-rules-boost-innovation-and-ban-nudification-apps-protect-citizens Accessed 2026-09-27. [340] AI Omnibus enters into force (news). European Commission, 2026-07-27. https://digital-strategy.ec.europa.eu/en/news/ai-omnibus-enters-force Accessed 2026-09-27. [341] CAP Code (UK Code of Non-broadcast Advertising and Direct & Promotional Marketing), Section 02 Recognition of marketing communications. Committee of Advertising Practice / ASA, 2025. https://www.asa.org.uk/type/non_broadcast/code_section/02.html Accessed 2026-09-27. [342] Digital Markets, Competition and Consumers Act 2024, Schedule 20: Commercial practices which are in all circumstances considered unfair. UK legislation.gov.uk, 2024-05-24. https://www.legislation.gov.uk/ukpga/2024/13/schedule/20 Accessed 2026-09-27. [343] ASA Ruling on Health Bridge Ltd t/a Zava (and Mumsnet Ltd), A25-1309865. Advertising Standards Authority, 2026-02-18. https://www.asa.org.uk/rulings/health-bridge-ltd-a25-1309865-health-bridge-ltd.html Accessed 2026-09-27. [344] ASA Ruling on Xinyu International Trading Ltd t/a consumertestreports.org, A26-1327770. Advertising Standards Authority, 2026-04-22. https://www.asa.org.uk/rulings/xinyu-international-trading-ltd-a26-1327770-xinyu-international-trading-ltd-.html Accessed 2026-09-27. [345] ASA rulings search: 'advertorial'. Advertising Standards Authority, 2026-09-27. https://www.asa.org.uk/codes-and-rulings/rulings.html?q=advertorial Accessed 2026-09-27. Access: partial extract. [346] Teads Holding Co. Reports First Quarter 2026 Results (Form 8-K Exhibit 99.1). Teads Holding Co., 2026-05-07. https://www.sec.gov/Archives/edgar/data/1454938/000145493826000034/q12026er-exhibit991.htm Accessed 2026-09-27. [347] Taboola Reports First Quarter 2026 Results (Form 8-K Exhibit 99.1). Taboola.com Ltd., 2026-05-06. https://www.sec.gov/Archives/edgar/data/1840502/000114036126019164/ef20071189_ex99-1.htm Accessed 2026-09-27. [348] Taboola Reports Strong Fourth Quarter & Full Year 2025 Financial Results (Exhibit 99.1 to Form 8-K). Taboola.com Ltd. (SEC EDGAR), 2026-02-25. https://www.sec.gov/Archives/edgar/data/1840502/000114036126006649/ef20064253_ex99-1.htm Accessed 2026-09-27. [349] Taboola Reports Fourth Quarter and Full Year 2023 Results (Form 8-K Exhibit 99.1). Taboola.com Ltd., 2024-02-28. https://www.sec.gov/Archives/edgar/data/1840502/000114036124009863/ef20022363_ex99-1.htm Accessed 2026-09-27. [350] Taboola Reports First Quarter 2024 Results (Form 8-K Exhibit 99.1). Taboola.com Ltd., 2024-05-08. https://www.sec.gov/Archives/edgar/data/1840502/000114036124024773/ef20028228_ex99-1.htm Accessed 2026-09-27. [351] Taboola Q4 and FY2024 results press release (Form 8-K, EX-99.1). Taboola.com Ltd. (SEC EDGAR), 2025-02-26. https://www.sec.gov/Archives/edgar/data/1840502/000114036125005947/ef20044101_ex99-1.htm Accessed 2026-09-27. [352] Taboola (TBLA) Q2 2026 Earnings Call Transcript. The Motley Fool (third-party transcript of Taboola call; speakers Adam Singolda, Steve Walker), 2026-08-12 (call held 2026-08-05). https://www.fool.com/earnings/call-transcripts/2026/08/12/taboola-tbla-q2-2026-earnings-call-transcript/ Accessed 2026-09-27. Access: not recorded (added by verifier). [353] EDGAR full-text search: phrase 'AI Overviews' in Taboola.com Ltd. filings (CIK 0001840502) — zero hits; phrase 'Microsoft Start' in Taboola 10-K/10-Q/8-K — zero hits. U.S. SEC EDGAR full-text search, 2026-09-27 (query date). https://efts.sec.gov/LATEST/search-index?q=%22AI%20Overviews%22&ciks=0001840502 Accessed 2026-09-27. [354] People Inc. Form 8-K (recast financial information; name change from IAC Inc.; Search segment discontinued; Care.com sale). People Inc., 2026-09-11. https://www.sec.gov/Archives/edgar/data/1800227/000162828026061531/iaci-20260911_d2.htm Accessed 2026-09-27. [355] IAC Reports Q4 2025 Results (Form 8-K Exhibit 99.1). IAC Inc. (now People Inc.), 2026-02-03. https://www.sec.gov/Archives/edgar/data/1800227/000162828026004988/ex_991q42025iac-pressrelea.htm Accessed 2026-09-27. [356] IAC Reports Q1 2026 Results (Form 8-K Exhibit 99.1). IAC Inc. (now People Inc.), 2026-05-04. https://www.sec.gov/Archives/edgar/data/1800227/000162828026029796/ex_991q12026iac-pressrelea.htm Accessed 2026-09-27. [357] Publisher ad supply fell by up to 40% in Q2 as AI search choked the open web. Digiday (Jessica Davies), reporting Ozone benchmarking data, 2026-07-15. https://digiday.com/media/publisher-ad-supply-fell-by-up-to-40-in-q2-as-ai-search-choked-the-open-web/ Accessed 2026-09-27. [358] Dianomi maintains revenue resilience as AI initiatives support second-half recovery (DNM) — full-year 2025 results. ADVFN (summarising Dianomi plc RNS), 2026-05-29. https://uk.advfn.com/market-news/article/16518/dianomi-maintains-revenue-resilience-as-ai-initiatives-support-second-half-recovery-dnm Accessed 2026-09-27. [359] Dianomi swings back to growth as first-half revenue edges up 2% — H1 2026 results. ADVFN (summarising Dianomi plc RNS), 2026-09-07. https://uk.advfn.com/market-news/article/22418/dianomi-swings-back-to-growth-as-first-half-revenue-edges-up-2 Accessed 2026-09-27. [360] Equativ and Sharethrough Will Now Operate Under Equativ Brand, Solidifying Global Position as Leading End-to-End Media Platform (press release). Equativ (via GlobeNewswire), 2025-06-09. https://www.globenewswire.com/news-release/2025/06/09/3095845/0/en/equativ-and-sharethrough-will-now-operate-under-equativ-brand-solidifying-global-position-as-leading-end-to-end-media-platform.html Accessed 2026-09-27. [361] Microsoft Start (Wikipedia article; cites Windows Central, 'Microsoft revives MSN brand ... retires Microsoft Start branding', 2024-11-11). Wikipedia contributors, 2025 (retrieved 2026-09-27). https://en.wikipedia.org/wiki/Microsoft_Start Accessed 2026-09-27. [362] Sixth Amendment to Amended and Restated Loan and Security Agreement (Silicon Valley Bank / Outbrain Inc.), Exhibit 10.5 to Outbrain Form S-1, incl. Perfection Certificate listing Ligatus GmbH. Outbrain Inc., 2021-06-29. https://www.sec.gov/Archives/edgar/data/1454938/000110465921086945/tm2113258d11_ex10-5.htm Accessed 2026-09-27. [363] EDGAR full-text search: 'Ligatus' in Outbrain/Teads Holding Co. filings (CIK 0001454938), forms S-1 and 10-K — 14 hits incl. Exhibit 21.1 subsidiary lists for FY2021-FY2025. U.S. SEC EDGAR full-text search, 2026-09-27 (query date). https://efts.sec.gov/LATEST/search-index?q=%22Ligatus%22&ciks=0001454938&forms=S-1,10-K Accessed 2026-09-27. [364] Life360 to Acquire Nativo, Accelerating Growth and Expanding Its Advertising Platform (press release, Form 8-K exhibit). Life360, Inc., 2025-11-10. https://www.sec.gov/Archives/edgar/data/1581760/000158176025000225/nexuspressrelease_11725x.htm Accessed 2026-09-27. [365] Life360 Completes Acquisition of Nativo and Surpasses 50 Million U.S. MAU (press release, Form 8-K Exhibit 99.1). Life360, Inc., 2026-01-05. https://www.sec.gov/Archives/edgar/data/1581760/000119312526001632/d81240dex991.htm Accessed 2026-09-27. [366] TripleLift Quietly Lays Off A Double-Digit Percentage Of Its Workforce. AdExchanger (Allison Schiff), 2025-07-11. https://www.adexchanger.com/publishers/triplelift-quietly-lays-off-a-double-digit-percentage-of-its-workforce/ Accessed 2026-09-27. [367] RevContent corporate website (homepage). RevContent, LLC, 2026 (copyright). https://www.revcontent.com/ Accessed 2026-09-27. [368] ZergNet website (homepage). ZergNet, 2026 (accessed). https://www.zergnet.com/ Accessed 2026-09-27. [369] Adyoulike corporate website (homepage). Adyoulike, 2026 (copyright). https://www.adyoulike.com/ Accessed 2026-09-27. [370] engageya.com page record (captured discontinuation notice). Website Informer (third-party capture), undated capture. https://website.informer.com/engageya.com Accessed 2026-09-27. Access: not recorded (added by verifier). [371] StartupTicker: +++ Sofatutor +++ Forget Finance +++ Waterglass +++ Plista +++ Solaris +++. deutsche-startups.de, 2024-03-11. https://www.deutsche-startups.de/tag/plista/ Accessed 2026-09-27. Access: not recorded (added by verifier). [372] Domain resolution checks (WebFetch) for plista.com / www.plista.com, www.content.ad and www.engageya.com. Research agent observation, 2026-09-27. https://plista.com/ Accessed 2026-09-27. [373] EDGAR full-text search: phrase 'Adyoulike' across all filers, 2022-01-01 to 2026-09-27 — zero hits. U.S. SEC EDGAR full-text search, 2026-09-27 (query date). https://efts.sec.gov/LATEST/search-index?q=%22Adyoulike%22&dateRange=custom&startdt=2022-01-01&enddt=2026-09-27 Accessed 2026-09-27. [374] Outside magazine built its own recommendation widgets to replace Outbrain. Digiday, 2017-05-04. https://digiday.com/media/outside-magazine-built-recommendation-widgets-replace-outbrain/ Accessed 2026-09-27. [375] AdCP - Ad Context Protocol (home page). AgenticAdvertising.org, 2026. https://adcontextprotocol.org/ Accessed 2026-09-27. [376] adcontextprotocol/adcp - Releases. AgenticAdvertising.org (GitHub org adcontextprotocol), 2026-09-24. https://github.com/adcontextprotocol/adcp/releases Accessed 2026-09-27. Access: partial extract. [377] AdCP documentation (docs home). AgenticAdvertising.org, 2026. https://docs.adcontextprotocol.org/ Accessed 2026-09-27. [378] IAB Tech Lab Standards index. IAB Tech Lab, 2026. https://iabtechlab.com/standards/ Accessed 2026-09-27. [379] AAMP - Agentic Advertising Management Protocols (standard page). IAB Tech Lab (Agentic Task Force), 2026-09-22 (page 'last updated'). https://iabtechlab.com/standards/aamp-agentic-advertising-management-protocols/ Accessed 2026-09-27. [380] IABTechLab/AAMP - hub repository. IAB Tech Lab (GitHub), 2026. https://github.com/IABTechLab/AAMP Accessed 2026-09-27. [381] IAB Tech Lab Introduces AAMP 3.0 with OpenProposal. IAB Tech Lab, 2026-09-22. https://iabtechlab.com/press-releases/iab-tech-lab-introduces-aamp-3-0-with-openproposal/ Accessed 2026-09-27. Access: not recorded (added by verifier). [382] prebid/salesagent - README. Prebid.org (GitHub), 2026. https://github.com/prebid/salesagent Accessed 2026-09-27. [383] Prebid Sales Agent (product page). Prebid.org, 2026. https://prebid.org/product-suite/sales-agent/ Accessed 2026-09-27. [384] adcontextprotocol/adcp - README. AgenticAdvertising.org, 2026. https://github.com/adcontextprotocol/adcp Accessed 2026-09-27. [385] AgenticAdvertising.org (home page). AgenticAdvertising.org, 2026. https://agenticadvertising.org/ Accessed 2026-09-27. [386] Realize - The Performance Platform Built for Advertisers (home page). Taboola (Realize), 2026. https://realize.com/ Accessed 2026-09-27. [387] Teads - The omnichannel outcomes platform (home page). Teads Holding Co., 2026. https://www.teads.com/ Accessed 2026-09-27. [388] Canonical-formats roadmap beyond 3.1 (Issue #4663). AgenticAdvertising.org (adcontextprotocol/adcp GitHub), 2026-05-17. https://github.com/adcontextprotocol/adcp/issues/4663 Accessed 2026-09-27. Access: not recorded (added by verifier). [389] Creative Formats (AdCP docs, v3.1.24). AgenticAdvertising.org, 2026. https://docs.adcontextprotocol.org/docs/creative/formats Accessed 2026-09-27. Access: not recorded (added by verifier). [390] Apostra (home page; scope3.com redirects here). Apostra (formerly Scope3), 2026. https://apostra.com/ Accessed 2026-09-27. [391] Introducing Apostra. Apostra (formerly Scope3), 2026-09-23. https://apostra.com/blog/introducing-apostra Accessed 2026-09-27. [392] Abby - KI-Assistenz fuer Werbetreibende (Taboola product/press page, German language). Taboola, 2024-10-17. https://www.taboola.com/abby Accessed 2026-09-27. [393] Taboola press releases page (navigation and footer). Taboola, 2026. https://www.taboola.com/press-releases Accessed 2026-09-27. Access: partial extract. [394] Meet MIA: The AI Assistant Behind Smarter Campaign Decisions. MGID, 2026-07-16. https://www.mgid.com/blog/meet-mia-the-ai-agent-behind-smarter-campaign-decisions Accessed 2026-09-27. [395] unBoxed 2025: Introducing Ads Agent. Amazon Ads, 2025-11-11. https://advertising.amazon.com/resources/whats-new/unboxed-2025-introducing-ads-agent Accessed 2026-09-27. [396] unBoxed 2025: Creative Agent. Amazon Ads, 2025-11-11. https://advertising.amazon.com/resources/whats-new/unboxed-2025-creative-agent Accessed 2026-09-27. [397] Amazon Ads - What's new (index). Amazon Ads, 2026. https://advertising.amazon.com/resources/whats-new Accessed 2026-09-27. [398] Reach your audience in new ways with August's Demand Gen Drop. Google, 2026-08-27. https://blog.google/products/ads-commerce/demand-gen-drop-august-2026/ Accessed 2026-09-27. [399] Boost your holiday sales with these agentic commerce updates. Google, 2026-09-16. https://blog.google/products-and-platforms/products/shopping/google-shopping-updates-holiday-shopping Accessed 2026-09-27. [400] How Businesses Are Driving Results with Meta's AI-Powered Ads. Meta (Meta for Business news), 2026-09-03. https://www.facebook.com/business/news/businesses-driving-results-with-meta-ai-ads Accessed 2026-09-27. [401] The Trade Desk Q4 and FY2025 results press release (Form 8-K, EX-99.1). The Trade Desk, Inc. (SEC EDGAR), 2026-02-25. https://www.sec.gov/Archives/edgar/data/1671933/000167193326000009/ttd-20260225x8kexx991.htm Accessed 2026-09-27. [402] Realize MCP (Backstage API reference). Taboola Dev Center, 2026. https://developers.taboola.com/backstage-api/reference/realize-mcp.md Accessed 2026-09-27. Access: not recorded (added by verifier). [403] Realize MCP Server (Realize Advertiser Help Center). Taboola / Realize, 2026-07-23. https://realize.com/help/en/articles/15565551-realize-mcp-server Accessed 2026-09-27. Access: not recorded (added by verifier). [404] Amendments No. 5, 6 and 7 to the Digital Property and Demand Services Agreement between Taboola, Inc. and Yahoo Inc. (exhibit to Form 10-Q for Q1 2026). Taboola.com Ltd. / Yahoo Inc., 2026-05-06. https://www.sec.gov/Archives/edgar/data/1840502/000184050226000009/yamendments.htm Accessed 2026-09-27. Access: full text with redactions. [405] Amendment No. 4 to the Digital Property and Demand Services Agreement (Taboola / Yahoo) (Exhibit 10.1 to Form 10-Q for Q3 2025). Taboola.com Ltd. / Yahoo Inc., 2025-11-05. https://www.sec.gov/Archives/edgar/data/1840502/000184050225000081/amendmentno4tothedigitalpr.htm Accessed 2026-09-27. Access: full text with redactions. [406] Digital Property and Demand Services Agreement (eff. 2022-11-28) with Programmatic Addendum (eff. 2023-05-01) and Payment, Data and Security Amendment (eff. 2023-07-14) (Exhibit 10.1 to Form 10-Q for Q3 2024). Taboola, Inc. / Taboola.com Ltd. / Yahoo Inc., 2024-11-07. https://www.sec.gov/Archives/edgar/data/1840502/000114036124045484/ef20034533_ex10-1.htm Accessed 2026-09-27. Access: full text with redactions. [407] Revcontent Acquired by Star Mountain Capital and Capital Dynamics. Capital Dynamics (news release), 2021-06-02. https://www.capdyn.com/news/revcontent-acquired-by-star-mountain-capital-and-capital-dynamics/ Accessed 2026-09-27. [408] Voland Partners and Readpeak join forces. Readpeak (company news), 2025-04-09. https://www.readpeak.com/articles/news/voland-partners-and-readpeak-join-forces Accessed 2026-09-27. [409] Backstage API: Bidding strategies overview and Maximize conversions. Taboola Developer Center, 2026. https://developers.taboola.com/backstage-api/docs/bid-strategy-overview.md Accessed 2026-09-27. [410] Conversion Tracking Overview (available integrations). Taboola Developer Center, 2026. https://developers.taboola.com/pixel/docs/conversion-tracking-overview Accessed 2026-09-27. [411] Backstage API: Publisher Targeting (Blocking). Taboola Developer Center, 2026. https://developers.taboola.com/backstage-api/docs/publisher-targeting.md Accessed 2026-09-27. [412] Conversion Bid Strategy (help centre). Outbrain / Teads, undated. https://www.outbrain.com/help/advertisers/conversion-bid-strategy/ Accessed 2026-09-27. [413] Teads Holding Co. FY2025 10-K, XBRL page R5 (Consolidated Statements of Operations). Teads Holding Co. / SEC EDGAR, 2026-03. https://www.sec.gov/Archives/edgar/data/1454938/000145493826000015/R5.htm Accessed 2026-09-27. [414] Turning CTV Into a Performance Powerhouse: From Impressions to Outcomes. Teads (blog), 2025-10-31. https://www.teads.com/blog/ctv-performance-outcomes-measurable-action/6470/ Accessed 2026-09-27. [415] CTV Performance Is Now Generally Available: Turning the Living Room Into an Outcomes Engine. Dylan Forman, Teads (blog), 2026-07-02. https://www.teads.com/blog/ctv-performance-is-now-generally-available-turning-the-living-room-into-an-outcomes-engine/11159/ Accessed 2026-09-27. [416] Teads Enterprise advertisers page. Teads, 2026. https://www.teads.com/advertisers/enterprise/ Accessed 2026-09-27. [417] Teads Launches its AI-Driven Operating System, Teads EngageOS. Teads (press release on blog), 2026-06-11. https://www.teads.com/blog/teads-launches-engage-os-ai-publisher-feed-system-with-magnite/11014/ Accessed 2026-09-27. [418] Teads Expands Access to Walmart Connect Sponsored Search and Onsite Display. Teads (press release on blog), 2026-06-16. https://www.teads.com/blog/teads-walmart-connect-retail-media-convergence/11044/ Accessed 2026-09-27. [419] What is Outbrain Amplify? (help centre). Outbrain / Teads, undated. https://www.outbrain.com/help/advertisers/what-is-outbrain-amplify/ Accessed 2026-09-27. [420] Help Center: Tracking Options in MGID Ads. MGID, Inc., 2026. https://help.mgid.com/available-tracking-options-in-mgid-ads Accessed 2026-09-27. [421] Help Center: Advertisers (article catalogue). MGID, Inc., 2026. https://help.mgid.com/advertisers Accessed 2026-09-27. [422] Help Center: Sources Optimization and Blocklist Management in MGID Ads. MGID, Inc., 2026. https://help.mgid.com/sources-optimization-and-blocklist-management-in-mgid-ads Accessed 2026-09-27. [423] Help Center: MGID Advertiser verification program. MGID, Inc., 2026. https://help.mgid.com/mgid-advertiser-verification-program Accessed 2026-09-27. [424] Help Center: How MGID Detects and Prevents Invalid Traffic (IVT). MGID, Inc., 2026. https://help.mgid.com/how-mgid-detects-and-prevents-invalid-traffic-ivt Accessed 2026-09-27. [425] MGID for Publishers. MGID, Inc., 2026. https://www.mgid.com/publishers Accessed 2026-09-27. [426] MGID Google MCM programme page. MGID, Inc., 2026. https://google-mcm.mgid.com/ Accessed 2026-09-27. [427] MGID homepage. MGID Inc., 2026 (copyright year). https://www.mgid.com/ Accessed 2026-09-27. [428] MGID Has Achieved the TAG 'Certified Against Fraud' Seal. MGID via PR Newswire, 2018-12-20. https://www.prnewswire.com/news-releases/mgid-has-achieved-the-tag-certified-against-fraud-seal-867097708.html Accessed 2026-09-27. [429] MGID sellers.json. MGID, Inc., 2026-09-27. https://www.mgid.com/sellers.json Accessed 2026-09-27. [430] About MGID. MGID, Inc., 2026. https://www.mgid.com/about Accessed 2026-09-27. [431] How Clickbait Ads Weaponize Brazilian News Websites to Spread Health Misinformation. Aos Fatos (Bortolon, Favero, Mangabeira, Bartels, Sousa), 2024-12-02. https://www.aosfatos.org/noticias/how-clickbait-ads-weaponize-brazilian-news-websites-to-spread-health-misinformation/ Accessed 2026-09-27. Access: not recorded (added by verifier). [432] Online Advertising in Ukraine and Russia During the 2022 Russian Invasion. Yeung, Iqbal, Tsipenyuk O'Neil, Kohno, Roesner (University of Washington; ACM Web Conference 2023), 2023-04. https://www.franziroesner.com/pdf/yeung-ads-webconf23.pdf Accessed 2026-09-27. Access: not recorded (added by verifier). [433] Help Center: Can I change the categories of ads on my website?. MGID, Inc., 2026. https://help.mgid.com/can-i-change-the-categories-of-ads-on-my-website Accessed 2026-09-27. [434] Mediaite Selects MGID to Drive Monetization Strategy. MGID, Inc., 2026-09-03. https://www.mgid.com/blog/mediaite-selects-mgid-to-drive-monetization-strategy Accessed 2026-09-27. [435] MGID blog: Press category listing. MGID, Inc., 2026-09. https://www.mgid.com/blog/category/press Accessed 2026-09-27. [436] MGID Partners With IAS to Strengthen Third-Party Measurement and Optimization. MGID, Inc., 2025-04-25. https://www.mgid.com/blog/mgid-partners-with-ias-to-strengthen-third-party-measurement-and-optimization Accessed 2026-09-27. [437] Strossle — now Azerion Native, Powered by MGID. Azerion Group / Strossle, undated (accessed 2026-09-27). https://strossle.com/ Accessed 2026-09-27. [438] MGID Awarded TAG Brand Safety Certification. MGID, Inc., 2022-05-04. https://www.mgid.com/blog/mgid-awarded-tag-brand-safety-certification Accessed 2026-09-27. Access: not recorded (added by verifier). [439] Widget Targeting and Bidding. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/widget-targeting Accessed 2026-09-27. [440] Widget Blocking and Un-Blocking. RevContent, LLC, 2024-01. https://help.revcontent.com/knowledge/widget-blocking-and-un-blocking Accessed 2026-09-27. [441] Demand-side Platform (DSP) Common Questions. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/demand-side-platform-common-questions Accessed 2026-09-27. [442] Supply-Side Platform (SSP) Common Questions. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/common-ssp-onboarding-questions Accessed 2026-09-27. [443] Prebid.js bidder documentation — Revcontent. Prebid.org, n.d.. https://docs.prebid.org/dev-docs/bidders/revcontent.html Accessed 2026-09-27. [444] Widget Catalog. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/widget-catalog Accessed 2026-09-27. [445] Configuring Apple News Widgets. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/apple-news-widget-integration Accessed 2026-09-27. [446] Email Monetization Overview. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/email-monetization-overview Accessed 2026-09-27. [447] Landing Page Policies. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/landing-page-policies Accessed 2026-09-27. [448] Prohibited Content. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/prohibited-content Accessed 2026-09-27. [449] Creating a Campaign. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/guide-to-creating-a-campaign Accessed 2026-09-27. [450] Server to Server (S2S) Conversion Tracking. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/server-to-server-s2s-conversion-tracking Accessed 2026-09-27. [451] How does RevContent measure Impressions?. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/impression-reporting Accessed 2026-09-27. [452] Invalid Traffic (IVT) Filters. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/invalid-traffic-ivt-filters Accessed 2026-09-27. [453] RevContent Earns Verification as Participant in TAG Registry. RevContent, LLC (blog; archived original on blog.revcontent.com), 2018-05-11 (CMS 'Updated On' 2025-02-21). https://www.revcontent.com/blog/revcontent-earns-verification-as-participant-in-tag-registry-to-improve-transparency-and-brand-safety-in-digital-advertising Accessed 2026-09-27. Access: not recorded (added by verifier). [454] For Advertisers. RevContent, LLC, n.d.. https://www.revcontent.com/advertisers Accessed 2026-09-27. [455] RevContent sellers.json. RevContent, LLC, 2026-09-26. https://assets.revcontent.com/sellers/master/sellers.json Accessed 2026-09-27. [456] RevContent Advertiser Agreement. RevContent, LLC, 2024-11-13. https://help.revcontent.com/knowledge/advertiser-agreement Accessed 2026-09-27. [457] RevContent and Mula Announce Strategic Partnership to Launch Agentic Content Monetization OS, Bringing Article-Level Intelligence to Native Advertising. RevContent / Mula via Newswire.com, 2026-03-09. https://www.newswire.com/news/revcontent-and-mula-announce-strategic-partnership-to-launch-agentic-content Accessed 2026-09-27. [458] RevContent and Mula Announce Strategic Partnership to Launch Agentic Content Monetization OS. RevContent / Mula (press release via Yahoo Finance), 2026-03-09. https://finance.yahoo.com/news/revcontent-mula-announce-strategic-partnership-123000094.html Accessed 2026-09-27. [459] Help Center — Advertisers category index. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/advertisers Accessed 2026-09-27. [460] Dianomi: Publishers. Dianomi plc, 2026. https://www.dianomi.com/who-we-help/publishers/ Accessed 2026-09-27. [461] Taboola to Acquire Dianomi, Significantly Bolstering its Ability to Provide Performance Advertisers with a Highly Specialized, Premium Ad Network Focused on Finance, Powered by Taboola's Performance Advertising Platform Realize (GlobeNewswire, syndicated on Yahoo Finance). Taboola.com Ltd, 2026-09-18. https://finance.yahoo.com/media-advertising/articles/taboola-acquire-dianomi-significantly-bolstering-080000597.html Accessed 2026-09-27. [462] Dianomi plc Half-year Financial Report, six months ended 30 June 2026. Dianomi plc via RNS, 2026-09-07. https://www.investegate.co.uk/announcement/rns/dianomi-plc--dnm/half-year-financial-report/9757536 Accessed 2026-09-27. [463] Dianomi plc Trading Update (H1 2026). Dianomi plc via RNS, 2026-07-27. https://www.investegate.co.uk/announcement/rns/dianomi-plc--dnm/trading-update/9687603 Accessed 2026-09-27. [464] Optimizing for Conversions. Dianomi Help Centre (Zendesk), 2025-04-28. https://support.dianomi.com/hc/en-gb/articles/33329397822609-Optimizing-for-Conversions Accessed 2026-09-27. Access: not recorded (added by verifier). [465] What is Conversion Tracking?. Dianomi Help Centre (Zendesk), 2026-05-05. https://support.dianomi.com/hc/en-gb/articles/30722512067473-What-is-Conversion-Tracking Accessed 2026-09-27. Access: not recorded (added by verifier). [466] Use DoubleVerify with Dianomi campaigns / Use Integral Ad Science (IAS) with Dianomi campaigns. Dianomi Help Centre (Zendesk), 2026-04-02. https://support.dianomi.com/hc/en-gb/articles/30614395383185-Use-DoubleVerify-with-Dianomi-campaigns Accessed 2026-09-27. Access: not recorded (added by verifier). [467] Dianomi plc Final Results for the year ended 31 December 2024. Dianomi plc via RNS, 2025-05-19. https://www.investegate.co.uk/announcement/rns/dianomi-plc--dnm/final-results/8883818 Accessed 2026-09-27. [468] Dianomi: Advertisers. Dianomi plc, 2026. https://www.dianomi.com/who-we-help/advertisers/ Accessed 2026-09-27. [469] Dianomi homepage. Dianomi (UK Company Reg. 4513809), 2026 (copyright year). https://www.dianomi.com/ Accessed 2026-09-27. [470] Recommended Acquisition of Dianomi plc by Taboola Europe Limited (Rule 2.7 announcement). Dianomi plc / Taboola Europe Limited via RNS, 2026-09-18. https://www.investegate.co.uk/announcement/rns/dianomi-plc--dnm/recommended-acquisition-of-dianomi-plc/9778532 Accessed 2026-09-27. [471] Readpeak Oy - company information and key figures. Finder.fi (Fonecta), 2026 (FY2025/06 accounts shown). https://www.finder.fi/Mainostoimisto/Readpeak+Oy/Helsinki/yhteystiedot/2923564 Accessed 2026-09-27. Access: paywalled excerpt. [472] Tracking scripts: technical documentation (support). Readpeak, 2025-07-31 (sitemap lastmod). https://www.readpeak.com/support/campaign-management/tracking-scripts-technical-documentation Accessed 2026-09-27. [473] Server-to-server tracking (support). Readpeak, 2025-04-28 (sitemap lastmod). https://www.readpeak.com/support/tracking/server-to-server-tracking Accessed 2026-09-27. [474] Account and users (support). Readpeak, 2026-09-01. https://www.readpeak.com/support/general/account-and-users Accessed 2026-09-27. [475] Creative guidelines (support). Readpeak, 2026-06-11. https://www.readpeak.com/support/creatives/creative-guidelines Accessed 2026-09-27. [476] Product launch: Video Native. Readpeak, 2025-06-12. https://www.readpeak.com/articles/news/product-launch-video-native Accessed 2026-09-27. [477] Basic campaign settings (support). Readpeak, 2025-04-09 (sitemap lastmod). https://www.readpeak.com/support/campaign-setup/basic-campaign-settings Accessed 2026-09-27. [478] Conversion optimization algorithm (support). Readpeak, 2025-11-28. https://www.readpeak.com/support/campaign-management/conversion-optimization-algorithm Accessed 2026-09-27. [479] Targeting features (support). Readpeak, 2025-04-30 (sitemap lastmod). https://www.readpeak.com/support/targeting/targeting-features Accessed 2026-09-27. [480] Case study: BMW Nordics. Readpeak, 2026-02-06 (sitemap lastmod; campaign dates not stated). https://www.readpeak.com/case-studies/bmw-nordics Accessed 2026-09-27. [481] Case study: Verman with Kinesso Finland. Readpeak, 2026-02-06 (sitemap lastmod; campaign dates not stated). https://www.readpeak.com/case-studies/kinesso Accessed 2026-09-27. [482] Case study: Native advertising vs display ads (OMD Denmark). Readpeak / OMD Denmark, 2026-04-14 (sitemap lastmod; study date not stated). https://www.readpeak.com/case-studies/native-advertising-vs-display-ads Accessed 2026-09-27. [483] About Readpeak. Readpeak, 2026-09 (sitemap lastmod 2026-09-11). https://www.readpeak.com/about Accessed 2026-09-27. [484] Voland Partners - portfolio (Readpeak entry) and fund facts. Voland Partners, 2026 (as accessed). https://voland.fi/ Accessed 2026-09-27. [485] Readpeak page sitemap. Readpeak, 2026-09-25 (index lastmod). https://www.readpeak.com/page-sitemap.xml Accessed 2026-09-27. [486] Sami Pippuri joins as new CTO. Readpeak, 2026-09-11. https://www.readpeak.com/articles/news/sami-pippuri-joins-as-new-cto Accessed 2026-09-27. [487] Growing our European presence. Readpeak, 2026-02-02. https://www.readpeak.com/articles/news/growing-our-european-presence Accessed 2026-09-27. [488] Readpeak - LinkedIn company page. Readpeak (LinkedIn), 2026-09-27 (as accessed). https://www.linkedin.com/company/readpeak Accessed 2026-09-27. [489] Life360 Reports Record Q1 2026 Results (media release, EX-99.1 to 8-K). Life360, Inc., 2026-05-11. https://www.sec.gov/Archives/edgar/data/1581760/000158176026000079/a991-life360incq126media.htm Accessed 2026-09-27. [490] Life360 Reports Record Q2 2026 Results (media release, EX-99.1 to 8-K). Life360, Inc., 2026-08-10. https://www.sec.gov/Archives/edgar/data/1581760/000158176026000142/life360incq226mediarelea.htm Accessed 2026-09-27. [491] Life360 Q2 2026 Results Presentation (EX-99.3 to 8-K). Life360, Inc., 2026-08-10. https://www.sec.gov/Archives/edgar/data/1581760/000158176026000142/a360q226resultspresentat.htm Accessed 2026-09-27. [492] Prebid bidder documentation: Nativo. Prebid.org, 2026. https://docs.prebid.org/dev-docs/bidders/nativo.html Accessed 2026-09-27. [493] Solutions: SSP. Nativo / Life360 Ads, 2026. https://ads.life360.com/solutions/ssp Accessed 2026-09-27. [494] Life360 Ads (Nativo) homepage. Life360, Inc. / Nativo, Inc., 2026 (copyright year). https://ads.life360.com/ Accessed 2026-09-27. [495] Solutions: Curation. Nativo / Life360 Ads, 2026. https://ads.life360.com/solutions/curation Accessed 2026-09-27. [496] Solutions: SaaS (publishers). Nativo / Life360 Ads, 2026. https://ads.life360.com/solutions/saas Accessed 2026-09-27. [497] Solutions: Self Serve. Nativo / Life360 Ads, 2026. https://ads.life360.com/solutions/self-serve Accessed 2026-09-27. [498] Solutions: SPARC (Nativo Predictive Audiences). Nativo / Life360 Ads, 2026. https://ads.life360.com/solutions/sparc Accessed 2026-09-27. [499] Life360 Q2 2026 Shareholder Letter (EX-99.2 to 8-K). Life360, Inc., 2026-08-10. https://www.sec.gov/Archives/edgar/data/1581760/000158176026000142/q226life360_shareholderl.htm Accessed 2026-09-27. [500] Solutions: Brand Rank. Nativo / Life360 Ads, 2026. https://ads.life360.com/solutions/brand-rank Accessed 2026-09-27. [501] Form 10-Q for quarter ended 2026-06-30 (Nativo acquisition note; disaggregation). Life360, Inc., 2026-08-10. https://www.sec.gov/Archives/edgar/data/1581760/000158176026000141/lifx-20260630.htm Accessed 2026-09-27. [502] Life360 (LIF) Q4 2025 Earnings Call Transcript. The Motley Fool (transcript of Life360 call), 2026-03-02. https://www.fool.com/earnings/call-transcripts/2026/03/02/life360-lif-q4-2025-earnings-call-transcript/ Accessed 2026-09-27. Access: not recorded (added by verifier). [503] Life360 Reports Record Q4 2025 Results (media release, EX-99.1 to 8-K). Life360, Inc., 2026-03-02. https://www.sec.gov/Archives/edgar/data/1581760/000158176026000017/life360incq425mediarelea.htm Accessed 2026-09-27. [504] Inside Life360's Plan To Build An Ad Platform On Top Of Nativo. AdExchanger (Allison Schiff), 2026-03-06. https://www.adexchanger.com/platforms/inside-life360s-plan-to-build-an-ad-platform-on-top-of-nativo/ Accessed 2026-09-27. [505] TripleLift sellers.json (snapshot). TripleLift Inc., 2026-09-27. https://triplelift.com/sellers.json Accessed 2026-09-27. [506] documentation/ortb2.3.tl.md (TripleLift Native RTB integration spec). TripleLift (GitHub), undated. https://github.com/triplelift/documentation/blob/master/ortb2.3.tl.md Accessed 2026-09-27. [507] TripleLift Introduces the World's First Programmatic RTB-Enabled Platform for Native Advertising. TripleLift (PRWeb), 2014-12-16. https://www.prweb.com/releases/triplelift_introduces_the_world_s_first_programmatic_rtb_enabled_platform_for_native_advertising/prweb12397787.htm Accessed 2026-09-27. [508] Native Request Object - v1.2. TripleLift developer docs, 2026-09-14. https://docs.triplelift.com/docs/native-request-object Accessed 2026-09-27. [509] TripleLift bidder adapter documentation. Prebid.org, undated. https://docs.prebid.org/dev-docs/bidders/triplelift.html Accessed 2026-09-27. [510] Responsive eCommerce creative (REC) on 3p Native Inventories. Amazon Ads (What's new), 2023-12-08. https://advertising.amazon.com/resources/whats-new/responsive-ecommerce-creative-on-3p-native-inventories Accessed 2026-09-27. Access: not recorded (added by verifier). [511] TripleLift Joins Amazon's DSP Certified Supply Exchange Program. TripleLift, 2025-10-14. https://triplelift.com/press/triplelift-joins-amazon-certified-supply-exchange-program/ Accessed 2026-09-27. [512] Vista Equity Partners Acquires Majority Stake in TripleLift in a Deal Valued at $1.4B. AlleyWatch, 2021-03-30. https://www.alleywatch.com/2021/03/triplelift-adtech-native-programmatic-acquired-vista-equity-partners/ Accessed 2026-09-27. [513] TripleLift Closes a Breakthrough 2025 with Global Expansion, Creative Innovation and Strategic Partnerships Across CTV, Retail Media, and Addressability. TripleLift (PR Newswire; same text on triplelift.com/press), 2025-12-11. https://www.prnewswire.com/news-releases/triplelift-closes-a-breakthrough-2025-with-global-expansion-creative-innovation-and-strategic-partnerships-across-ctv-retail-media-and-addressability-302639190.html Accessed 2026-09-27. [514] AdTech Leader TripleLift Announces Majority Investment from Vista Equity Partners. TripleLift / Vista Equity Partners (PR Newswire), 2021-03-29. https://www.prnewswire.com/news-releases/adtech-leader-triplelift-announces-majority-investment-from-vista-equity-partners-301257437.html Accessed 2026-09-27. [515] Ads observed on made-for-advertising sites in January 2024. Adalytics, 2024-01. https://adalytics.io/blog/ads-observed-on-made-for-advertising-sites-in-january-2024 Accessed 2026-09-27. [516] TripleLift Creative Formats Delivered 272% ROI, According to Total Economic Impact Study. TripleLift (PR Newswire); study by Forrester Consulting, 2026-02-26. https://www.prnewswire.com/news-releases/triplelift-creative-formats-delivered-272-roi-according-to-total-economic-impact-study-302698325.html Accessed 2026-09-27. Access: partial extract. [517] TripleLift CTV Campaigns Deliver Up to 26x Industry Benchmarks, Proving the Power of Non-Standard Formats. TripleLift (PR Newswire), 2026-06-23. https://www.prnewswire.com/news-releases/triplelift-ctv-campaigns-deliver-up-to-26x-industry-benchmarks-proving-the-power-of-non-standard-formats-302807396.html Accessed 2026-09-27. [518] Platform page. TripleLift, 2026-09-27. https://triplelift.com/platform/ Accessed 2026-09-27. [519] TripleLift Knowledge Base (public landing view). TripleLift, 2026-09-27. https://support.triplelift.com/ Accessed 2026-09-27. Access: partial extract. [520] TripleLift Launches TL Direct, Self-Serve Platform for Sell-Side Orchestration Across Web, Mobile, and CTV. TripleLift (PR Newswire), 2026-06-10. https://www.prnewswire.com/news-releases/triplelift-launches-tl-direct-self-serve-platform-for-sell-side-orchestration-across-web-mobile-and-ctv-302796586.html Accessed 2026-09-27. [521] TripleLift Launches TL Spark, Introducing an Intelligence Layer for Outcome-Driven Advertising. TripleLift (PR Newswire), 2026-04-21. https://www.prnewswire.com/news-releases/triplelift-launches-tl-spark-introducing-an-intelligence-layer-for-outcome-driven-advertising-302748652.html Accessed 2026-09-27. [522] TripleLift and Multilocal Partner to Put Creative into the Curation Equation. TripleLift (PR Newswire), 2026-09-22. https://www.prnewswire.com/news-releases/triplelift-and-multilocal-partner-to-put-creative-into-the-curation-equation-302886017.html Accessed 2026-09-27. [523] TripleLift Sees Record Breaking Results for Prime Day 2026 as Offsite Commerce Media Reaches Inflection Point. TripleLift (PR Newswire), 2026-07-02. https://www.prnewswire.com/news-releases/triplelift-sees-record-breaking-results-for-prime-day-2026-as-offsite-commerce-media-reaches-inflection-point-302817014.html Accessed 2026-09-27. [524] TripleLift CEO Dave Clark Abruptly Exits After Setting The SSP On A New Trajectory. AdExchanger (James Hercher), 2024-07-15. https://www.adexchanger.com/online-advertising/triplelift-ceo-dave-clark-abruptly-exits-after-setting-the-ssp-on-a-new-trajectory/ Accessed 2026-09-27. [525] TripleLift Appoints Dave Simon as Chief Revenue Officer, Completing Its Executive Leadership Team. TripleLift, 2026-08-10. https://triplelift.com/press/triplelift-appoints-dave-simon-as-cro-completing-its-executive-leadership-team/ Accessed 2026-09-27. [526] Sharethrough sellers.json (www.sharethrough.com/sellers.json redirects to iab.sharethrough.com/sellers.json). Sharethrough Inc., 2026-09-27. https://iab.sharethrough.com/sellers.json Accessed 2026-09-27. [527] Equativ sellers.json (www.equativ.com/sellers.json redirects to ced-ns.sascdn.com/iab/sellers.json). Equativ, 2026-09-25. https://ced-ns.sascdn.com/iab/sellers.json Accessed 2026-09-27. [528] Sharethrough Ad Exchange Supply Policy. Sharethrough, 2022-06-28. https://www.sharethrough.com/sharethrough-supply-policy Accessed 2026-09-27. [529] Open RTB API integration: bid request specification. Equativ Help Center, 2026-09-24. https://help.equativ.com/open-rtb-api-integration-bid-request-specification Accessed 2026-09-27. [530] Prebid bidder documentation: Sharethrough. Prebid.org, 2026. https://docs.prebid.org/dev-docs/bidders/sharethrough.html Accessed 2026-09-27. [531] Equativ - Programmatic Offering - Creative Specs by Channel & Enhancements. Equativ / Sharethrough support centre (Zendesk), 2026-06-01. https://support.sharethrough.com/hc/en-us/articles/16543531373851-Creative-Specs Accessed 2026-09-27. [532] Equativ bans MFA. Equativ (press release), 2025-05-16. https://www.equativ.com/press/equativ-bans-mfa Accessed 2026-09-27. [533] Equativ Helps Advertisers Reclaim Viewer Attention with Creative Enhancements. Equativ (press release), 2026-04-07. https://www.equativ.com/press/equativ-helps-advertisers-reclaim-viewer-attention-with-creative-enhancements Accessed 2026-09-27. [534] Display Native (Creative Solutions). Equativ, 2026. https://www.equativ.com/advertisers/creative-solutions/display-native Accessed 2026-09-27. [535] Publisher Solutions page. Equativ, 2026. https://www.equativ.com/publishers Accessed 2026-09-27. [536] Equativ Brings ChatGPT Ads to Brands as Advertising Partner. Equativ (press release; syndicated on GlobeNewswire 2026-09-22), 2026-09-22. https://www.equativ.com/press/equativ-brings-chatgpt-ads-to-brands-as-advertising-partner Accessed 2026-09-27. [537] New ways to buy ChatGPT ads. OpenAI, 2026-05-05. https://openai.com/index/new-ways-to-buy-chatgpt-ads/ Accessed 2026-09-27. [538] Equativ Unveils Maestro by Equativ: Its Most Advanced End-to-End Curation Platform .... Equativ (GlobeNewswire), 2025-01-22. https://www.globenewswire.com/news-release/2025/01/22/3013463/0/en/Equativ-Unveils-Maestro-by-Equativ-Its-Most-Advanced-End-to-End-Curation-Platform-for-Enhanced-Campaign-Performance-and-Greater-Control-for-Advertisers.html Accessed 2026-09-27. Access: partial extract. [539] Equativ Files Lawsuit Against Google .... Equativ (press release), 2025-10-13. https://www.equativ.com/press/equativ-files-lawsuit-against-google-in-pursuit-of-more-open-and-competitive-digital-advertising-markets Accessed 2026-09-27. [540] Taboola.com Ltd. Form 10-Q for the quarter ended March 31, 2026 (accession 0001840502-26-000009). Taboola.com Ltd. / SEC EDGAR, 2026-05-06. https://www.sec.gov/Archives/edgar/data/1840502/000184050226000009/tbla-20260331.htm Accessed 2026-09-27. [541] Taboola.com Ltd. Annual Report on Form 10-K for fiscal year 2024. Taboola.com Ltd., 2025-02-26. https://www.sec.gov/Archives/edgar/data/1840502/000184050225000019/tbla-20241231.htm Accessed 2026-09-27. Access: partial extract. [542] Apple turns to Taboola to sell ads. Marketing Brew (Ryan Barwick), 2024-07-17. https://www.marketingbrew.com/stories/2024/07/17/apple-turns-to-taboola-to-sell-ads Accessed 2026-09-27. [543] Taboola Expands Beyond Native Ads with Realize; New Technology Platform Allows Taboola to Serve The Entire Performance Advertising Market. Taboola (press release), 2025-02-26. https://www.taboola.com/press-release/taboola-realize Accessed 2026-09-27. [544] Taboola Dev Center home (Backstage API, Conversion Tracking, Dynamic Creative, Mobile SDK, Web Integrations, Newsroom). Taboola Developer Center, 2026. https://developers.taboola.com/ Accessed 2026-09-27. [545] Abby: Gen-AI Performance Expert (Realize Advertiser Help Center). Taboola / Realize help center, 2026-02-02. http://realize.com/help/en/articles/11205912-abby-gen-ai-performance-expert Accessed 2026-09-27. [546] Taboola Launches Realize+, an Agentic AI System Turning Advertiser Goals Into Outcomes and Expands Agentic Capabilities by Opening its Platform to Claude Skills. Taboola (press release), 2026-04-23. https://www.taboola.com/press-release/realize-plus Accessed 2026-09-27. [547] Taboola Introduces Abby, An Industry First Generative AI Assistant. Taboola (press release), 2024-10-15. https://www.taboola.com/press-release/taboola-introduces-abby Accessed 2026-09-27. [548] Budget Allocator (Realize Advertiser Help Center). Taboola / Realize help center, 2026-07-28. http://realize.com/help/en/articles/15920115-budget-allocator Accessed 2026-09-27. [549] Realize Advertiser Help Center: home and collection indexes (Optimize Performance; Policy & Content Review). Taboola / Realize help center, 2026. http://realize.com/help/en/collections/11915685-optimize-performance Accessed 2026-09-27. Access: partial extract. [550] Campaign Reporting (Realize Advertiser Help Center). Taboola / Realize help center, 2026. http://realize.com/help/en/articles/10466790-campaign-reporting Accessed 2026-09-27. [551] Taboola's Trust Center (Realize Advertiser Help Center). Taboola / Realize help center, 2026. http://realize.com/help/en/articles/3878192-taboola-s-trust-center Accessed 2026-09-27. [552] A Content Checklist: How to Avoid Campaign Rejection (Realize Advertiser Help Center). Taboola / Realize help center, 2025-05-26. http://realize.com/help/en/articles/8965770-a-content-checklist-how-to-avoid-campaign-rejection Accessed 2026-09-27. [553] After going public via a SPAC, Taboola acquires e-commerce marketing network Connexity for $800M. TechCrunch, 2021-07-23. https://techcrunch.com/2021/07/23/after-going-public-via-a-spac-taboola-acquires-e-commerce-marketing-network-connexity-for-800m/ Accessed 2026-09-27. [554] Teads advertisers page. Teads, 2026. https://www.teads.com/advertisers/ Accessed 2026-09-27. [555] Pentaleap and Teads Partner to Break New Ground in Programmatic Retail Media. Teads (press release on blog), 2025-07-24. https://www.teads.com/blog/pentaleap-and-teads-partner-to-break-new-ground-in-programmatic-retail-media/ Accessed 2026-09-27. [556] Outbrain homepage (Outbrain Direct Response by Teads). Teads / Outbrain, 2026. https://www.outbrain.com/ Accessed 2026-09-27. [557] Teads Launches CTV Ensemble for Unified Omnichannel Orchestration. Teads (press release on blog), 2026-06-18. https://www.teads.com/blog/teads-launches-ctv-ensemble-for-unified-omnichannel-orchestration/11116/ Accessed 2026-09-27. [558] Teads elevates Travel Performance advertising with Teads Conversions. Teads (blog), 2024-09-16. https://www.teads.com/blog/travel-advertising-growth-performance-conversions/2971/ Accessed 2026-09-27. [559] Teads and Adelaide Partner to Launch High-Attention Omnichannel Private Marketplaces. Teads (press release on blog), 2023-08-21. https://www.teads.com/blog/teads-adelaide-high-attention-omnichannel-private-marketplaces/2594/ Accessed 2026-09-27. [560] Teads and Lumen Research Bring Proven Attention Science to CTV for Omnichannel Impact. Teads (press release on blog), 2026-05-06. https://www.teads.com/blog/teads-and-lumen-research-bring-proven-attention-science-to-ctv-for-omnichannel-impact/10475/ Accessed 2026-09-27. [561] Teads Launches First-to-Market Innovation with TVision and Adelaide to Maximize Media Quality Across the Glass. Teads (press release on blog), 2024-04-29. https://www.teads.com/blog/first-to-market-media-quality-innovation-across-glass/2793/ Accessed 2026-09-27. [562] Outbrain advertiser help centre index. Outbrain / Teads, undated. https://www.outbrain.com/help/advertisers/ Accessed 2026-09-27. [563] Zemanta — now part of the Teads offering. Teads (Zemanta), undated (page metadata 2022–2024; accessed 2026-09-27). https://www.zemanta.com/ Accessed 2026-09-27. [564] Teads press releases listing page. Teads, 2026-09. https://www.teads.com/press-releases/ Accessed 2026-09-27. [565] FY2025 10-K XBRL R116: Segment and Geographic Information - Geographical Revenue (Details). Teads Holding Co. / SEC EDGAR rendering, 2026-03-16. https://www.sec.gov/Archives/edgar/data/1454938/000145493826000015/R116.htm Accessed 2026-09-27. [566] Help Center: Requirements for Publishers. MGID, Inc., 2026. https://help.mgid.com/requirements-to-publishers Accessed 2026-09-27. [567] MGID Native Advertising product page. MGID, Inc., 2026. https://www.mgid.com/native-advertising Accessed 2026-09-27. [568] Help Center: OpenRTB DSP Integration Guide. MGID, Inc., 2026. https://help.mgid.com/openrtb-dsp-integration-guide Accessed 2026-09-27. [569] Prebid.js bidder documentation: Mgid. Prebid.org (bidder-authored page), 2026. https://docs.prebid.org/dev-docs/bidders/mgid.html Accessed 2026-09-27. [570] Prebid.js bidder documentation: MgidX. Prebid.org (bidder-authored page), 2026. https://docs.prebid.org/dev-docs/bidders/mgidX.html Accessed 2026-09-27. [571] Help Center: Compliance Risk Levels. MGID, Inc., 2026. https://help.mgid.com/compliance-risk-levels Accessed 2026-09-27. [572] Help Center: Ad Safety Rankings. MGID, Inc., 2026. https://help.mgid.com/ad-safety-rankings Accessed 2026-09-27. [573] MGID Advertisers Terms of Service. MGID, Inc., 2026-07-29. https://www.mgid.com/services/advertisers-tos Accessed 2026-09-27. [574] MGID Help Center home. MGID, Inc., 2026. https://help.mgid.com/ Accessed 2026-09-27. [575] Help Center: AI Image Generator for Ad Campaigns. MGID, Inc., 2026. https://help.mgid.com/ai-image-generator-for-ad-campaigns Accessed 2026-09-27. [576] Help Center: Which content is restricted or prohibited on MGID?. MGID, Inc., 2026. https://help.mgid.com/which-content-is-restricted-or-prohibited-on-mgid Accessed 2026-09-27. [577] Help Center: CPA Tune Overview. MGID, Inc., 2026. https://help.mgid.com/cpa-tune-overview Accessed 2026-09-27. [578] Help Center: Budget Type Settings. MGID, Inc., 2026. https://help.mgid.com/budget-type-settings Accessed 2026-09-27. [579] MGID Launches CPA Tune, Delivering 155%+ Improvement. MGID, Inc., 2025-06-26. https://www.mgid.com/blog/mgid-launches-cpa-tune Accessed 2026-09-27. [580] MGID Upgrades CPA Tune to Boost Campaign Conversion Rate by Up to 237% Over Manual Optimization. MGID, Inc., 2026-09-16. https://www.mgid.com/blog/mgid-upgrades-cpa-tune-to-boost-campaign-conversion-rate-by-up-to-237-over-manual-optimization Accessed 2026-09-27. [581] Help Center: Fraud Prevention (index). MGID, Inc., 2026. https://help.mgid.com/fraud-prevention Accessed 2026-09-27. [582] MGID Make Only 100% Viewable Native Advertisements Billable. MGID via PR Newswire, 2020-02-12. https://www.prnewswire.com/news-releases/mgid-make-only-100-viewable-native-advertisements-billable-301002972.html Accessed 2026-09-27. [583] Help Center: Publishers (article catalogue). MGID, Inc., 2026. https://help.mgid.com/publishers Accessed 2026-09-27. [584] Help Center: What is the payout threshold?. MGID, Inc., 2026. https://help.mgid.com/what-is-the-payout-threshold Accessed 2026-09-27. [585] Help Center: Search Arbitrage. MGID, Inc., 2026. https://help.mgid.com/search-arbitrage Accessed 2026-09-27. [586] Prebid.js source: modules/mgidBidAdapter.js. Prebid.js repository (adapter contributed by MGID), 2026. https://raw.githubusercontent.com/prebid/Prebid.js/master/modules/mgidBidAdapter.js Accessed 2026-09-27. [587] Help Center: DSP (index). MGID, Inc., 2026. https://help.mgid.com/dsp Accessed 2026-09-27. [588] Help Center: SSP (index). MGID, Inc., 2026. https://help.mgid.com/ssp Accessed 2026-09-27. [589] IAB Europe TCF Global Vendor List v3 (JSON). IAB Europe, 2026-09-24. https://vendor-list.consensu.org/v3/vendor-list.json Accessed 2026-09-27. Access: partial extract. [590] PR Newswire: MGID news releases (organisation page). PR Newswire / MGID, 2017-2020. https://www.prnewswire.com/news/mgid/ Accessed 2026-09-27. [591] Help Center: Ad Quality Policies and Creative Guidelines (index). MGID, Inc., 2026. https://help.mgid.com/ad-quality-policies-and-creative-guidelines Accessed 2026-09-27. [592] MGID Partners with Digiseg to Enhance Privacy-First Audience Targeting Across Global Markets. MGID, Inc., 2025-05-30. https://www.mgid.com/blog/mgid-partners-with-digiseg-to-enhance-privacy-first-audience-targeting-across-global-markets Accessed 2026-09-27. [593] MGID Case Studies (index). MGID, Inc., 2026. https://www.mgid.com/case-studies Accessed 2026-09-27. Access: partial extract. [594] Revcontent Sees 1,100% Increase in Profitability, Bolsters Client Roster and Product Offerings Under New Leadership. Revcontent via PR Newswire, 2020-09-23. https://www.prnewswire.com/news-releases/revcontent-sees-1-100-increase-in-profitability-bolsters-client-roster-and-product-offerings-under-new-leadership-301136350.html Accessed 2026-09-27. [595] Help Center — Publishers category index. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/publishers Accessed 2026-09-27. [596] Revcontent Launches Industry-First Video Marketplace. Revcontent via PR Newswire, 2020-01-21. https://www.prnewswire.com/news-releases/revcontent-launches-industry-first-video-marketplace-300990490.html Accessed 2026-09-27. [597] Getting Started with RevContent's API. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/publisher-advertiser-api-requests Accessed 2026-09-27. [598] Personal Guarantee. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/personal-guarantee Accessed 2026-09-27. [599] Using RevGPT for AI-Generated Headlines. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/revgpt Accessed 2026-09-27. [600] RevContent Expands Access to GenAI Headline Creation Tech, RevGPT. RevContent, LLC (blog, Stephen Flee), 2025-03-18. https://www.revcontent.com/blog/revcontent-expands-access-to-genai-headline-creation-tech-revgpt Accessed 2026-09-27. [601] Native vCPM Best Practices. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/native-vcpm-best-practices Accessed 2026-09-27. [602] Native Advertising: What's Hot in 2026?. RevContent, LLC (blog, Patrick Brady), 2026-04-28. https://www.revcontent.com/blog/native-advertising-whats-hot-in-2026 Accessed 2026-09-27. [603] Publisher Account Tour. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/publisher-rep-dashboard-tour Accessed 2026-09-27. [604] Maximizing Reach with Brand Safe Content (Brand Safety Seal). RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/brand-safety-seal Accessed 2026-09-27. [605] Advertising Disclosures (site guidelines). RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/advertising-disclosures Accessed 2026-09-27. [606] Real Time Bidding (RTB) for Publishers. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/rtb-for-publishers Accessed 2026-09-27. [607] RevContent + Flipp: Discovering Weekly Specials from Local Grocers Just Got Easier. RevContent, LLC (blog, Kaelyn Keane), 2025-03-18. https://www.revcontent.com/blog/revcontent-flipp-br-discovering-weekly-specials-from-local-grocers-just-got-easier Accessed 2026-09-27. [608] Prebid.js modules/revcontentBidAdapter.md. Prebid.js repository (GitHub), n.d.. https://raw.githubusercontent.com/prebid/Prebid.js/master/modules/revcontentBidAdapter.md Accessed 2026-09-27. [609] Publisher Prebid Integration. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/publisher-prebid Accessed 2026-09-27. [610] Mobile SDKs. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/mobile-sdks Accessed 2026-09-27. [611] Revcontent Expands to Brazil, Offering Competitive Revenue to Publishers and Premium Inventory Offerings to Advertisers. Revcontent via PR Newswire, 2021-03-30. https://www.prnewswire.com/news-releases/revcontent-expands-to-brazil-offering-competitive-revenue-to-publishers-and-premium-inventory-offerings-to-advertisers-301258061.html Accessed 2026-09-27. [612] Campaign Targeting Options. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/campaign-targeting-options Accessed 2026-09-27. [613] Publisher Payment Requirements & Methods. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/publisher-payment-requirements-methods Accessed 2026-09-27. [614] Restricted Content. RevContent, LLC, n.d.. https://help.revcontent.com/knowledge/restricted-content Accessed 2026-09-27. [615] Revcontent Helps Tapstone Exceed Revenue Goals by 40% With Its Granular and Powerful Advertiser Capabilities. Revcontent via PR Newswire, 2021-02-02. https://www.prnewswire.com/news-releases/revcontent-helps-tapstone-exceed-revenue-goals-by-40-with-its-granular-and-powerful-advertiser-capabilities-301219667.html Accessed 2026-09-27. [616] Turning Ads into Experiences: Introducing Dianomi Interactive. Dianomi plc (blog), 2026-04-07. https://www.dianomi.com/cms/turning-ads-into-experiences-introducing-dianomi-interactive/ Accessed 2026-09-27. [617] How to Find and Deploy Image Conversion Tracking Pixel. Dianomi Help Centre (Zendesk), 2026-04-24. https://support.dianomi.com/hc/en-gb/articles/30687389969425-How-to-Find-and-Deploy-Image-Conversion-Tracking-Pixel Accessed 2026-09-27. Access: not recorded (added by verifier). [618] Third-party tracking. Dianomi Help Centre (Zendesk), 2025-12-15. https://support.dianomi.com/hc/en-gb/articles/30602644788753-Third-party-tracking Accessed 2026-09-27. Access: not recorded (added by verifier). [619] How can I see which publishers I'm appearing on?. Dianomi Help Centre (Zendesk), 2025-04-28. https://support.dianomi.com/hc/en-gb/articles/33131164416785-How-can-I-see-which-publishers-I-m-appearing-on Accessed 2026-09-27. Access: not recorded (added by verifier). [620] Supply Manager. Dianomi Help Centre (Zendesk), 2026-09-04. https://support.dianomi.com/hc/en-gb/articles/49750675787025-Supply-Manager Accessed 2026-09-27. Access: not recorded (added by verifier). [621] Publisher Bidding. Dianomi Help Centre (Zendesk), 2026-08-12. https://support.dianomi.com/hc/en-gb/articles/30604477563921-Publisher-Bidding Accessed 2026-09-27. Access: not recorded (added by verifier). [622] Dianomi plc Half-year Report, six months ended 30 June 2025. Dianomi plc via RNS, 2025-09-24. https://www.investegate.co.uk/announcement/rns/dianomi-plc--dnm/half-year-report/9127411 Accessed 2026-09-27. [623] Buying Dianomi inventory programmatically. Dianomi Help Centre (Zendesk), 2026-04-24. https://support.dianomi.com/hc/en-gb/articles/30722729171473-Buying-Dianomi-inventory-programmatically Accessed 2026-09-27. Access: not recorded (added by verifier). [624] Dianomi plc: Admission to AIM and First Day of Dealings. Dianomi plc via RNS, 2021-05-24. https://www.investegate.co.uk/announcement/rns/dianomi-plc--dnm/admission-to-aim-and-first-day-of-dealings/6480829 Accessed 2026-09-27. [625] Readpeak — Programmatic Native Advertising Platform. Readpeak, 2026 (copyright year). https://www.readpeak.com/ Accessed 2026-09-27. [626] Readpeak Privacy Policy. Readpeak, 2026-08. https://www.readpeak.com/readpeak-privacy-policy Accessed 2026-09-27. [627] Introduction to Readpeak: your self-serve platform for programmatic native advertising. Readpeak, 2024-09-11. https://www.readpeak.com/articles/product/introduction-to-readpeak-your-self-serve-platform-for-programmatic-native-advertising Accessed 2026-09-27. [628] Introducing AI-powered ad creation: a new feature on Readpeak. Readpeak, 2024-09-11. https://www.readpeak.com/articles/product/introducing-ai-powered-ad-creation-a-new-feature-on-readpeak Accessed 2026-09-27. [629] Readpeak support sitemap (help-article inventory). Readpeak, 2026-09-15. https://www.readpeak.com/support-sitemap.xml Accessed 2026-09-27. [630] Automatic campaign pacing (support). Readpeak, 2026-05-27. https://www.readpeak.com/support/readpeak-product/automatic-pacing Accessed 2026-09-27. [631] Understanding our time spent measurement and how it compares to GA4 (support). Readpeak, 2026-01-14. https://www.readpeak.com/support/campaign-setup/understanding-our-time-spent-measurement Accessed 2026-09-27. [632] DoubleVerify integration (support). Readpeak, 2025-05-05. https://www.readpeak.com/support/campaign-management/doubleverify-integration Accessed 2026-09-27. [633] TAG Brand Safety Certified in the UK. Readpeak, 2025-09-01. https://www.readpeak.com/articles/news/tag-certified-in-uk Accessed 2026-09-27. [634] Readpeak Brand Safety Policy. Readpeak, 2025-05-20 (sitemap lastmod). https://www.readpeak.com/readpeak-brand-safety-policy Accessed 2026-09-27. [635] Readpeak enters France. Readpeak, 2025-11-27. https://www.readpeak.com/articles/news/france-entry Accessed 2026-09-27. [636] READPEAK LTD - Companies House overview. Companies House (UK), 2026-09-27 (live register). https://find-and-update.company-information.service.gov.uk/company/15479295 Accessed 2026-09-27. [637] Readpeak AB - company page. allabolag.se, 2026 (FY2025/06 shown). https://www.allabolag.se/foretag/readpeak-ab/stockholm/reklambyr%C3%A5er/2KHKZ9MI5YGSK Accessed 2026-09-27. Access: partial extract. [638] Readpeak Norway NUF - company page (and proff.no search listing Readpeak AS). proff.no, 2026 (FY2025/06 shown). https://www.proff.no/selskap/readpeak-norway-nuf/bekkestua/reklamebyr%C3%A5er/IFD6C0U01DI Accessed 2026-09-27. Access: partial extract. [639] Readpeak SAS - North Data record. North Data, 2026. https://www.northdata.de/Readpeak Accessed 2026-09-27. Access: partial extract. [640] The State of Native Advertising in the Nordics 2024. Readpeak, 2024-11-26. https://www.readpeak.com/articles/news/the-state-of-native-advertising-in-the-nordics-2024 Accessed 2026-09-27. Access: partial extract. [641] Advertisers page. Nativo / Life360 Ads, 2026. https://ads.life360.com/advertisers Accessed 2026-09-27. [642] Nativo Ad Formats (publisher docs). Nativo, 2026. https://pub-resources.nativo.com/docs/nativo-ad-formats Accessed 2026-09-27. [643] Nativo Advertiser Resource Center (index). Nativo, 2026. https://resources.nativo.com/ Accessed 2026-09-27. [644] Self-serve changelog: Release 4.6.1. Nativo, undated. https://resources.nativo.com/changelog/release-461.md Accessed 2026-09-27. [645] Life360 Q3 2025 Investor Presentation (EX-99.2 to 8-K), slides 27-30. Life360, Inc., 2025-11-10. https://www.sec.gov/Archives/edgar/data/1581760/000158176025000225/a11102025ex992investorpr.htm Accessed 2026-09-27. [646] Nativo API documentation: Introduction. Nativo, 2026. https://api-docs.nativo.com/docs/introduction Accessed 2026-09-27. [647] Nativo Inventory Delivers 40% Higher AU Scores Than Industry Benchmarks (Adelaide study). Nativo (study by Adelaide), 2025-08-26. https://ads.life360.com/newsroom/new-adelaide-research-nativo-inventory-delivers-40-higher-au-scores-than-industry-benchmarks Accessed 2026-09-27. [648] Publishers page. Nativo / Life360 Ads, 2026. https://ads.life360.com/publishers Accessed 2026-09-27. [649] Nativo Publisher Resource Center (index). Nativo, 2026. https://pub-resources.nativo.com/ Accessed 2026-09-27. [650] Safety Blocking (publisher docs). Nativo, 2026. https://pub-resources.nativo.com/docs/safety-blocking Accessed 2026-09-27. [651] Developers page (SDK, API, resource portals). Nativo / Life360 Ads, 2026. https://ads.life360.com/developers Accessed 2026-09-27. [652] Nativo's Programmatic Revenue Soars Over 120% Year-Over-Year in Q2. Nativo, 2025-07-01. https://ads.life360.com/newsroom/nativos-programmatic-revenue-soars-over-120-year-over-year-in-q2 Accessed 2026-09-27. [653] High-Quality Inventory (TripleLift SSP page). TripleLift, 2026-09-27. https://triplelift.com/triplelift-ssp/ Accessed 2026-09-27. [654] TripleLift homepage. TripleLift, 2026 (copyright year). https://triplelift.com/ Accessed 2026-09-27. [655] Native advertising page ('Creative SSP for the Orchestration Era'). TripleLift, 2026-09-27. https://triplelift.com/native-advertising/ Accessed 2026-09-27. [656] TL Spark product page. TripleLift, 2026-09-27. https://triplelift.com/tl-spark/ Accessed 2026-09-27. [657] Retail media page. TripleLift, 2026-09-27. https://triplelift.com/retail-media/ Accessed 2026-09-27. [658] TripleLift Partners with Adelaide Launching First-Ever Attention-Based Buying Guarantee for CTV. TripleLift (PR Newswire), 2023-09-13. https://www.prnewswire.com/news-releases/triplelift-partners-with-adelaide-launching-first-ever-attention-based-buying-guarantee-for-ctv-301926477.html Accessed 2026-09-27. [659] TripleLift Is The First Supply-Side Platform (SSP) to Enable Responsive eCommerce Native Supply For Amazon Ads Buyers. TripleLift blog, 2024-05-15. https://triplelift.com/resources/blog/triplelift-amazon-ads/ Accessed 2026-09-27. [660] Amazon Ads deepens relationships with leading SSPs (Certified Supply Exchange). Amazon Ads, 2024-06-13. https://advertising.amazon.com/library/news/certified-supply-exchange Accessed 2026-09-27. [661] Global AdTech Company TripleLift Names Dave Clark as New Chief Executive Officer. TripleLift (PR Newswire), 2022-10-03. https://www.prnewswire.com/news-releases/global-adtech-company-triplelift-names-dave-clark-as-new-chief-executive-officer-301639066.html Accessed 2026-09-27. [662] TripleLift Unveils TripleLift Audiences, a Cross-Site Targeting Solution Without the Reach Limits and Data Costs of ID-Based Alternatives. TripleLift (PR Newswire), 2024-04-25. https://www.prnewswire.com/news-releases/triplelift-unveils-triplelift-audiences-a-cross-site-targeting-solution-without-the-reach-limits-and-data-costs-of-id-based-alternatives-302126806.html Accessed 2026-09-27. [663] Curation Platform (Maestro by Equativ) page. Equativ, 2026. https://www.equativ.com/advertisers/curation Accessed 2026-09-27. [664] Prebid bidder documentation: Equativ. Prebid.org, 2026. https://docs.prebid.org/dev-docs/bidders/equativ.html Accessed 2026-09-27. [665] Native (RTB) creative template. Equativ Help Center, 2025-12-19. https://help.equativ.com/native-rtb Accessed 2026-09-27. Access: partial extract. [666] Equativ Launches Media Planning Agent .... Equativ (press release), 2025-12-09. https://www.equativ.com/press/equativ-launches-media-planning-agent-to-enhance-buyer-efficiency-and-boost-campaign-outcomes Accessed 2026-09-27. [667] Bidding, Targeting, and Optimizations. Equativ / Sharethrough support centre, 2022-11-11. https://support.sharethrough.com/hc/en-us/articles/211834783-Bidding-Targeting-and-Optimizations Accessed 2026-09-27. [668] Sharethrough Marketplace Quality Policy. Sharethrough, 2024-04-29. https://www.sharethrough.com/sharethrough-marketplace-quality-policy Accessed 2026-09-27. [669] Equativ Launches Shoppable Ads to Bring Brands Closer to the Point of Purchase. Equativ (press release), 2025-10-28. https://www.equativ.com/press/equativ-launches-shoppable-ads-to-bring-brands-closer-to-the-point-of-purchase Accessed 2026-09-27. [670] Connect bidder to Equativ SSP: supported fields. Equativ Help Center, 2026-07-02. https://help.equativ.com/connect-bidder-to-equativ-ssp-supported-fields Accessed 2026-09-27. Access: partial extract. [671] Leading European Media, Data and Technology Companies Launch the European Media Marketplace. Equativ (press release), 2026-07-07. https://www.equativ.com/press/leading-european-media-data-and-technology-companies-launch-the-european-media-marketplace Accessed 2026-09-27. [672] Equativ announces Bridgepoint as new majority shareholder to accelerate scale and growth of the independent adtech company. Equativ (press release), 2023-02-23. https://www.equativ.com/press/equativ-announces-bridgepoint-as-new-majority-shareholder-to-accelerate-scale-and-growth-of-the-independent-adtech-company Accessed 2026-09-27. [673] Overview of STX. Equativ / Sharethrough support centre, 2022-11-11. https://support.sharethrough.com/hc/en-us/articles/207464003-Overview-of-STX Accessed 2026-09-27. [674] Taboola homepage. Taboola, undated (accessed 2026-09-27). https://www.taboola.com/ Accessed 2026-09-27. [675] EDGAR company record and 8-K list — Taboola.com Ltd. (CIK 0001840502). SEC EDGAR, 2026-09-27 (live index). https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&company=taboola&type=8-K&dateb=&owner=include&count=10 Accessed 2026-09-27. [676] EDGAR company record and 8-K list — Teads Holding Co. (CIK 0001454938). SEC EDGAR, 2026-09-27 (live index). https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&company=teads+holding&type=8-K&dateb=&owner=include&count=20 Accessed 2026-09-27. [677] Life360 Completes Acquisition of Nativo and Surpasses 50 Million U.S. MAU. Life360, Inc., 2026-01-05. https://ads.life360.com/newsroom/life360-completes-acquisition-of-nativo-and-surpasses-50-million-us-mau Accessed 2026-09-27. [678] Kargo homepage. Kargo, undated (accessed 2026-09-27). https://www.kargo.com/ Accessed 2026-09-27. [679] StackAdapt homepage. StackAdapt Inc., 2014–2026 (copyright range). https://www.stackadapt.com/ Accessed 2026-09-27. [680] StackAdapt — Native Advertising. StackAdapt Inc., undated (accessed 2026-09-27). https://www.stackadapt.com/native-advertising Accessed 2026-09-27. [681] StackAdapt — ChatGPT Ads. StackAdapt Inc., undated (accessed 2026-09-27). https://www.stackadapt.com/chatgpt-ads Accessed 2026-09-27. [682] Yieldmo homepage. Yieldmo, 2026 (copyright year). https://www.yieldmo.com/ Accessed 2026-09-27. [683] Media.net homepage. Media.net, 2025 (copyright year). https://www.media.net/ Accessed 2026-09-27. [684] Online Digital X (redirect target of plista.de). Online Digital X, Frankfurt, 2026 (copyright year). https://www.online-digitalx.de/ Accessed 2026-09-27. [685] ZergNet — About. ZergNet (Static Media), 2026 (copyright year). https://www.zergnet.com/about Accessed 2026-09-27. [686] NewsBreak for Business. Particle Media, Inc., 2026 (copyright year). https://biz.newsbreak.com/ Accessed 2026-09-27. [687] About SmartNews (EN). SmartNews, Inc., undated (news items dated 2025). https://about.smartnews.com/en/ Accessed 2026-09-27. [688] Primis homepage. Primis, 2026 (copyright year). https://www.primis.tech/ Accessed 2026-09-27. [689] JWX homepage (formerly JWP Connatix). JWP, Inc. (JWX), 2026 (copyright year). https://jwx.com/ Accessed 2026-09-27. [690] Seedtag homepage. Seedtag, undated (accessed 2026-09-27). https://www.seedtag.com/ Accessed 2026-09-27. [691] GumGum homepage. GumGum, Inc., 2025 (copyright year in footer). https://gumgum.com/ Accessed 2026-09-27. [692] Perion homepage (redirect target of undertone.com). Perion Network Ltd., 2026 (copyright year). https://perion.com/ Accessed 2026-09-27. [693] Sovrn homepage. Sovrn, Inc., undated (accessed 2026-09-27). https://www.sovrn.com/ Accessed 2026-09-27. [694] Ezoic homepage. Ezoic Inc., 2026 (copyright year). https://www.ezoic.com/ Accessed 2026-09-27. [695] PropellerAds homepage. Propeller Ads Ltd (Cyprus, HE353513), undated (accessed 2026-09-27). https://propellerads.com/ Accessed 2026-09-27. [696] Adsterra homepage. AD MARKET LIMITED (Cyprus), 2013–2026 (copyright range). https://adsterra.com/ Accessed 2026-09-27. [697] AdNow homepage. RENODO MEDIA LTD (Plovdiv, Bulgaria), 2026 (copyright year). https://adnow.com/ Accessed 2026-09-27. [698] BidVertiser homepage. BidVertiser / Bpath, LTD, undated ('Serving you Since 2008'). https://www.bidvertiser.com/ Accessed 2026-09-27. [699] Google AdSense Help — native ad formats (In-feed, In-article, Multiplex). Google, undated help article. https://support.google.com/adsense/answer/9183549 Accessed 2026-09-27. [700] Google Ad Manager Help — Native ads overview. Google, undated help article. https://support.google.com/admanager/answer/6366845 Accessed 2026-09-27. [701] Amazon Ads — Sponsored Display (now 'display ads'). Amazon Ads, undated (accessed 2026-09-27). https://advertising.amazon.com/solutions/products/sponsored-display Accessed 2026-09-27. [702] Amazon Publisher Services homepage. A9.com LLC (Amazon), 2025 (copyright year). https://aps.amazon.com/ Accessed 2026-09-27. [703] Microsoft Advertising — Display and Native ads (Audience ads). Microsoft, undated (accessed 2026-09-27). https://about.ads.microsoft.com/en/solutions/microsoft-audience-network Accessed 2026-09-27. [704] Meta Audience Network — Native Ads (developer docs). Meta Platforms, Inc., undated developer doc. https://developers.facebook.com/docs/audience-network/guides/ad-formats/native/ Accessed 2026-09-27. [705] AppLovin MAX — Native Ads (Android). AppLovin, undated developer doc (references Q2 2025 deprecation). https://support.applovin.com/en/max/android/ad-formats/native-ads/ Accessed 2026-09-27. [706] Unity Grow documentation hub. Unity, undated developer hub. https://docs.unity.com/en-us/grow Accessed 2026-09-27. [707] Native ads integration for Android (LevelPlay). Unity, undated developer doc. https://docs.unity.com/en-us/grow/levelplay/sdk/android/native-ads-integration Accessed 2026-09-27. Access: not recorded (added by verifier). [708] Digital Turbine investor relations — press release index. Digital Turbine, Inc., 2026-09-27 (live index; latest release 2026-08-04). https://ir.digitalturbine.com/ Accessed 2026-09-27. [709] Meta Reports Second Quarter 2026 Results. Meta Platforms, Inc., 2026-07-29. https://investor.atmeta.com/investor-news/press-release-details/2026/Meta-Reports-Second-Quarter-2026-Results/default.aspx Accessed 2026-09-27. [710] TikTok for Business (US). TikTok, 2026 (copyright year). https://ads.tiktok.com/business/en-US/ Accessed 2026-09-27. [711] LinkedIn Marketing Solutions — Native Advertising (Sponsored Content). LinkedIn (Microsoft), undated (accessed 2026-09-27). https://business.linkedin.com/marketing-solutions/native-advertising Accessed 2026-09-27. [712] Access-attempt record — X (formerly Twitter) advertising pages. this stream (fetch log), 2026-09-27. https://business.x.com/en/advertising Accessed 2026-09-27. Access: access error. [713] EDGAR company record and 8-K list — Reddit, Inc. (CIK 0001713445). SEC EDGAR, 2026-09-27 (live index). https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001713445&type=8-K&dateb=&owner=include&count=5 Accessed 2026-09-27. [714] Pinterest Business — Advertise. Pinterest, undated (accessed 2026-09-27). https://business.pinterest.com/en-us/advertise/ Accessed 2026-09-27. [715] Snapchat for Business — Advertising. Snap Inc., undated (data footnotes dated 2024). https://forbusiness.snapchat.com/advertising Accessed 2026-09-27. [716] Google Ads Help — About Demand Gen campaigns. Google, undated help article (references June 2026 migration). https://support.google.com/google-ads/answer/13695777 Accessed 2026-09-27. [717] Criteo homepage — Commerce Media Platform. Criteo S.A., 2026 (copyright year). https://www.criteo.com/ Accessed 2026-09-27. [718] Amazon Ads — Amazon DSP. Amazon Ads, undated (accessed 2026-09-27). https://advertising.amazon.com/solutions/products/amazon-dsp Accessed 2026-09-27. [719] The Trade Desk homepage. The Trade Desk, Inc., 2026 (copyright year). https://www.thetradedesk.com/ Accessed 2026-09-27. [720] Google Marketing Platform — Display & Video 360. Google, undated (accessed 2026-09-27). https://marketingplatform.google.com/about/display-video-360/ Accessed 2026-09-27. [721] ChatGPT Ads expands across Europe. OpenAI (read via r.jina.ai text proxy), 2026-08-18 (updated 2026-08-31). https://openai.com/index/chatgpt-ads-expands-across-europe/ Accessed 2026-09-27. Access: not recorded (added by verifier). [722] A milestone in expanding access to AI. OpenAI (read via r.jina.ai text proxy), 2026 (date not captured). https://openai.com/index/expanding-access-to-ai-with-chatgpt-ads/ Accessed 2026-09-27. Access: not recorded (added by verifier). [723] Why we're experimenting with advertising (Perplexity) — read via r.jina.ai text proxy. Perplexity AI, not visible in proxied text. https://www.perplexity.ai/hub/blog/why-we-re-experimenting-with-advertising Accessed 2026-09-27. Access: access error; text or metadata via another route. [724] Perplexity Pulling Sponsored Answers From AI Platform. PYMNTS (citing Financial Times), 2026-02-18. https://www.pymnts.com/artificial-intelligence-2/2026/perplexity-pulling-sponsored-answers-from-ai-platform/ Accessed 2026-09-27. Access: not recorded (added by verifier). [725] A new generation of ads for the AI era of Search. Google (The Keyword), 2026-05-20. https://blog.google/products/ads-commerce/google-marketing-live-search-ads Accessed 2026-09-27. [726] Microsoft Advertising blog — index. Microsoft, 2026-09-17 (latest post visible). https://about.ads.microsoft.com/en/blog Accessed 2026-09-27. Access: partial extract. [727] How to use Amazon Rufus (About Amazon). Amazon, 2024 (updated; rename note dated 2026-05-13). https://www.aboutamazon.com/news/retail/amazon-rufus Accessed 2026-09-27. [728] Dable homepage (Korean). Dable Inc., 2026 (copyright year). https://www.dable.io/ Accessed 2026-09-27. [729] popIn Discovery DSP. popIn Corporation (Tokyo), undated (privacy policy revised 2019-11-08). https://discovery.popin.cc/ Accessed 2026-09-27. [730] LiveIntent homepage. LiveIntent, 2025 (copyright year). https://www.liveintent.com/ Accessed 2026-09-27. [731] Zeta Global Completes Acquisition of LiveIntent (8-K Exhibit 99.1). Zeta Global Holdings Corp. via SEC EDGAR, 2024-10-21. https://www.sec.gov/Archives/edgar/data/1851003/000095017024115991/zeta-ex99_1.htm Accessed 2026-09-27. Access: not recorded (added by verifier). [732] EX.CO homepage. EX.CO, undated (data points through mid-2025). https://ex.co/ Accessed 2026-09-27. [733] Bidtellect is now part of Simpli.fi. Simpli.fi, 2023 (copyright year). https://www.bidtellect.com/ Accessed 2026-09-27. [734] Understand your Conversion Lift based on users measurement data — Google Ads Help. Google, undated help page (current as of access). https://support.google.com/google-ads/answer/14102450 Accessed 2026-09-27. [735] Guides Concerning the Use of Endorsements and Testimonials in Advertising (final revised Guides), 88 FR, Vol. 88 No. 142, FR Doc. 2023-14795. US Federal Trade Commission via Federal Register / GovInfo, 2023-07-26. https://www.federalregister.gov/documents/2023/07/26/2023-14795/guides-concerning-the-use-of-endorsements-and-testimonials-in-advertising Accessed 2026-09-27. Access: access error; text or metadata via another route. [736] Trade Regulation Rule on the Use of Consumer Reviews and Testimonials (final rule), 89 FR, Vol. 89 No. 163, FR Doc. 2024-18519. US Federal Trade Commission via Federal Register / GovInfo, 2024-08-22. https://www.federalregister.gov/documents/2024/08/22/2024-18519/trade-regulation-rule-on-the-use-of-consumer-reviews-and-testimonials Accessed 2026-09-27. Access: access error; text or metadata via another route. [737] Legislative Train Schedule: Digital Fairness Act. European Parliament, 2026-08-01. https://www.europarl.europa.eu/legislative-train/theme-a-new-plan-for-europe-s-sustainable-prosperity-and-competitiveness/file-digital-fairness-act Accessed 2026-09-27. [738] Realize Help: Video, Title and Thumbnail Best Practices. Realize (Taboola), undated. https://realize.com/help/en/articles/8964660-video-title-and-thumbnail-best-practices Accessed 2026-09-27. [739] MGID Help: What widget formats do you have?. MGID, undated. https://help.mgid.com/what-widget-formats-do-you-have Accessed 2026-09-27. [740] Outbrain Direct (ODR by Teads): Ads Specs. Teads Holding Co., undated. https://www.outbrain.com/help/ads-specs/ Accessed 2026-09-27. [741] Publisher Monetization Solutions | Kargo. Kargo, n/a (living product page). https://www.kargo.com/publishers Accessed 2026-09-27. [742] ANA Media Practice - Programmatic Transparency (hub page). ANA, 2026 (rolling). https://www.ana.net/content/show/id/media-programmatic-transparency Accessed 2026-09-27. Access: partial extract. [743] ANA Provides 'First Look' at In-depth Programmatic Media Transparency Study (press release). ANA, 2023-06-19. https://www.ana.net/content/show/id/pr-2023-06-programmaticstudy Accessed 2026-09-27. [744] Sponsorship Disclosure: Effects of Duration on Persuasion Knowledge and Brand Responses. Sophie C. Boerman, Eva A. van Reijmersdal, Peter C. Neijens; Journal of Communication 62(6), 1047-1064, 2012-12. https://research.wur.nl/en/publications/sponsorship-disclosure-effects-of-duration-on-persuasion-knowledg/ Accessed 2026-09-27. Access: abstract only. [745] The Meta Report: Lessons from 640 Haus Incrementality Experiments. Tyler Horner, Haus (measurement vendor), 2025-07-28. https://www.haus.io/blog/the-meta-report-lessons-from-640-haus-incrementality-experiments Accessed 2026-09-27. [746] Realize Help: Motion Ads. Realize (Taboola), undated. https://realize.com/help/en/articles/10466720-motion-ads Accessed 2026-09-27. [747] Realize Help: Carousel Ads. Realize (Taboola), undated. https://realize.com/help/en/articles/10466667-carousel-ads Accessed 2026-09-27. [748] Realize Help: Vertical / Display Ads. Realize (Taboola), undated. https://realize.com/help/en/articles/10693567-vertical-display-ads Accessed 2026-09-27. [749] Realize Help: App Promotion. Realize (Taboola), undated. https://realize.com/help/en/articles/10441763-app-promotion Accessed 2026-09-27. [750] Realize: Ad Experiences. Realize (Taboola), undated. https://realize.com/ad-experiences/ Accessed 2026-09-27. [751] MGID Help: Video Player and VAST Integration. MGID, undated. https://help.mgid.com/video-player-and-vast-integration Accessed 2026-09-27. [752] MGID Help: Motion Ads Tutorial. MGID, undated. https://help.mgid.com/motion-ads-tutorial Accessed 2026-09-27. [753] Google Ad Manager Help: Create, traffic, and report on native video ads. Google, undated. https://support.google.com/admanager/answer/7025311 Accessed 2026-09-27. [754] Media.net: Publishers. Media.net, undated. https://www.media.net/publishers Accessed 2026-09-27. [755] Meta Audience Network: Native Banner Ads (Android). Meta, undated. https://developers.facebook.com/docs/audience-network/setting-up/ad-setup/android/native-banner/ Accessed 2026-09-27. [756] Microsoft Learn: Audience Ads. Microsoft, undated. https://learn.microsoft.com/en-us/advertising/guides/audience-ads?view=bingads-13 Accessed 2026-09-27. [757] NewsBreak Ads Help: Types of NewsBreak Ad Formats and Placements. NewsBreak, undated. https://adhelp.newsbreak.com/hc/en-us/articles/42301540100109-Types-of-NewsBreak-Ad-Formats-and-Placements Accessed 2026-09-27. [758] NewsBreak Ads Help: Video Ad Specs. NewsBreak, undated. https://adhelp.newsbreak.com/hc/en-us/articles/42301498732813-Video-Ad-Specs Accessed 2026-09-27. [759] NewsBreak Ads Help: Carousel Ads. NewsBreak, undated. https://adhelp.newsbreak.com/hc/en-us/articles/43947845477773-Carousel-Ads Accessed 2026-09-27. [760] Yahoo and Taboola Announce Closing of Deal, 30-Year Strategic Partnership Sees Taboola Power Recommendations for Yahoo. Taboola / Yahoo (press release), 2023-01-17. https://www.taboola.com/press-release/taboola-and-yahoo-close-deal Accessed 2026-09-27. [761] Taboola Form 8-K (Items 1.01, 1.02, 2.03): new $270M revolving credit facility; modification of Yahoo Share Repurchase Agreement. Taboola.com Ltd., 2025-03-19. https://www.sec.gov/Archives/edgar/data/1840502/000114036125009356/ef20045735_8k.htm Accessed 2026-09-27. [762] Taboola Reaches 10 Years in Serving Advertising on Microsoft Properties Including MSN, Edge Browsers and Windows Experiences; Taboola Now Serves Advertising Across Microsoft Productivity Suite, Including Outlook Mail and Office 365. Taboola (press release), 2025-03-18. https://www.taboola.com/press-release/microsoft-10years Accessed 2026-09-27. [763] Taboola Unveils DeeperDive, a Gen AI Answer Engine Built for the Open Web. Taboola (press release), 2025-06-11. https://www.taboola.com/press-release/deeperdive Accessed 2026-09-27. [764] Form 8-K dated November 10, 2025: privately negotiated repurchase of 6,387,648 shares from Pitango funds. Taboola.com Ltd. / SEC EDGAR, 2025-11-12. https://www.sec.gov/Archives/edgar/data/1840502/000114036125041711/ef20058675_8k.htm Accessed 2026-09-27. [765] EDGAR company filing index, Teads Holding Co., 8-K filings 2025-2026. U.S. SEC EDGAR, 2026-08-14. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001454938&type=8-K&dateb=&owner=include&count=40 Accessed 2026-09-27. [766] Form 8-K, Item 2.05 Costs Associated with Exit or Disposal Activities. Teads Holding Co., 2025-12-08. https://www.sec.gov/Archives/edgar/data/1454938/000145493825000142/outbrain-20251203.htm Accessed 2026-09-27. [767] Form 8-K, Item 8.01: Nasdaq confirms regained compliance with Minimum Bid Price Requirement. Teads Holding Co., 2026-06-08. https://www.sec.gov/Archives/edgar/data/1454938/000145493826000043/tead-20260605.htm Accessed 2026-09-27. [768] EDGAR company filing index, Teads Holding Co. (CIK 0001454938), 10-K filings. U.S. SEC EDGAR, 2026-03-16. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001454938&type=10-K&dateb=&owner=include&count=40 Accessed 2026-09-27. [769] Teads Files Lawsuit Against Google Seeking Financial Damages Following Federal Court Antitrust Ruling (8-K Exhibit 99.1). Teads Holding Co., 2026-08-03. https://www.sec.gov/Archives/edgar/data/1454938/000145493826000048/exh991pr.htm Accessed 2026-09-27. [770] outbrain.com/onyx/ (HTTP 301 redirect to teads.com). Teads / Outbrain, undated. https://www.outbrain.com/onyx/ Accessed 2026-09-27. Access: partial extract. [771] MGID Privacy Notice. MGID, Inc., 2025-11-03. https://www.mgid.com/services/privacy-policy Accessed 2026-09-27. [772] OpenCorporates search: MGID Inc. OpenCorporates, 2026. https://opencorporates.com/companies?q=MGID+Inc Accessed 2026-09-27. Access: partial extract. [773] DOU.ua company profile: MGID. DOU (Ukrainian IT job board), 2026. https://jobs.dou.ua/companies/mgid/ Accessed 2026-09-27. [774] Meet the Team (leadership, board, advisors). RevContent, LLC, n.d.. https://www.revcontent.com/resources/meet-the-team Accessed 2026-09-27. [775] Revcontent Acquired by Star Mountain Capital and Capital Dynamics. Revcontent via PR Newswire, 2021-06-02. https://www.prnewswire.com/news-releases/revcontent-acquired-by-star-mountain-capital-and-capital-dynamics-301304541.html Accessed 2026-09-27. [776] Google News RSS results for 'Revcontent' and 'RevContent Marques' (headline index). Google News (aggregating PR Newswire, Business Insider, Adweek, MediaPost, Sarasota Herald-Tribune, Editor & Publisher, Newswire.com, Business Observer), 2026-09-27. https://news.google.com/rss/search?q=Revcontent&hl=en-US&gl=US&ceid=US:en Accessed 2026-09-27. Access: partial extract. [777] RevContent Expands Board of Directors. RevContent, LLC (blog, Richard Marques), 2025-02-21. https://www.revcontent.com/blog/revcontent-expands-board-of-directors Accessed 2026-09-27. [778] RevContent Privacy Policy. RevContent, LLC, 2026-04-10. https://www.revcontent.com/privacy-policy Accessed 2026-09-27. [779] Companies House record: DIANOMI PLC 04513809. Companies House (UK registrar), 2026-09-03. https://find-and-update.company-information.service.gov.uk/company/04513809 Accessed 2026-09-27. [780] Dianomi: About. Dianomi plc, 2026. https://www.dianomi.com/who-we-are/about/ Accessed 2026-09-27. [781] Dianomi plc Trading Statement (FY2025). Dianomi plc via RNS, 2026-03-18. https://www.investegate.co.uk/announcement/rns/dianomi-plc--dnm/trading-statement/9479234 Accessed 2026-09-27. [782] Dianomi plc Final Results for the year ended 31 December 2025. Dianomi plc via RNS, 2026-05-29. https://www.investegate.co.uk/announcement/rns/dianomi-plc--dnm/final-results/9590881 Accessed 2026-09-27. [783] Taboola To Acquire Ad-Tech Company Targeting Affluent Audiences. Laurie Sullivan, MediaPost, 2026-09-18. https://www.mediapost.com/publications/article/418116/taboola-to-acquire-ad-tech-company-targeting-afflu.html Accessed 2026-09-27. [784] PRH open data API - company 2615511-5. Finnish Patent and Registration Office (PRH) / YTJ, 2026-09-27 (live register). https://avoindata.prh.fi/opendata-ytj-api/v3/companies?businessId=2615511-5 Accessed 2026-09-27. [785] Recap 2024 from our CEO Tomas Forsback. Readpeak, 2024-12-16. https://www.readpeak.com/articles/news/recap-2024-from-our-ceo-tomas-forsback Accessed 2026-09-27. [786] Readpeak expands to Switzerland. Readpeak, 2024-09-10. https://www.readpeak.com/articles/news/switzerland-expansion Accessed 2026-09-27. [787] Gearing up in Germany ahead of DMEXCO. Readpeak, 2025-09-05. https://www.readpeak.com/articles/news/gearing-up-in-germany-ahead-of-dmexco Accessed 2026-09-27. [788] Careers and Culture page. Nativo / Life360 Ads, 2026. https://ads.life360.com/careers-and-culture Accessed 2026-09-27. [789] Form 10-Q for quarter ended 2026-03-31 (Nativo acquisition note; revenue policy; disaggregation). Life360, Inc., 2026-05-11. https://www.sec.gov/Archives/edgar/data/1581760/000158176026000078/lifx-20260331.htm Accessed 2026-09-27. [790] Life360 Reports Record Q3 2025 Results (EX-99.1 to 8-K). Life360, Inc., 2025-11-10. https://www.sec.gov/Archives/edgar/data/1581760/000158176025000225/a11102025ex991life360inc.htm Accessed 2026-09-27. [791] Brought To You By The Consumers: TripleLift, Sticky Uncover Preferred Native Ad Disclosures. MediaPost (Tyler Loechner), 2015-05-04. https://www.mediapost.com/publications/article/249028/brought-to-you-by-the-consumers-triplelift-stick.html Accessed 2026-09-27. [792] Programmatic platform TripleLift sells majority stake for reported $1.4B. Marketing Dive (Robert Williams), 2021-03-30. https://www.marketingdive.com/news/programmatic-platform-triplelift-sells-majority-stake-for-reported-14b/597510/ Accessed 2026-09-27. [793] TripleLift Has Acquired 1plusX. TripleLift (CEO letter to employees), 2022-03-21. https://triplelift.com/press/triplelift-acquired-1plusx/ Accessed 2026-09-27. [794] TripleLift Snags 1plusX For $150 Million, Its First-Ever Acquisition. AdExchanger (James Hercher), 2022-03-21. https://www.adexchanger.com/ad-exchange-news/triplelift-snags-1plusx-for-150-million-its-first-ever-acquisition/ Accessed 2026-09-27. [795] Changes to Our Organization. TripleLift (Dave Clark, CEO), 2023-02-09. https://triplelift.com/press/changes-to-our-organization/ Accessed 2026-09-27. [796] TripleLift Names Dave Helmreich as Chief Executive Officer. TripleLift (PR Newswire), 2025-02-12. https://www.prnewswire.com/news-releases/triplelift-names-dave-helmreich-as-chief-executive-officer-302374788.html Accessed 2026-09-27. [797] TripleLift Appoints Timothy Jasionowski as CPTO, Adds Benjamin Felix to Leadership Team. TripleLift, 2026-05-05. https://triplelift.com/press/triplelift-appoints-timothy-jasionowski-as-cpto-and-benjamin-felix-as-cmo/ Accessed 2026-09-27. [798] TripleLift press archive (page 1). TripleLift, 2026-09-24. https://triplelift.com/press/ Accessed 2026-09-27. [799] TripleLift to Sell Majority Stake to Vista Equity Partners for $1.4 Billion. Adweek (Ronan Shields), 2021-03. https://www.adweek.com/media/triplelift-to-sell-majority-stake-to-vista-equity-partners-for-1-4-billion/ Accessed 2026-09-27. Access: paywalled excerpt. [800] District M & Sharethrough Have Entered a Definitive Agreement to Merge and Become One of the Leading Independent Omnichannel Ad Exchanges. PR Newswire (company release), 2021-02-09. https://www.prnewswire.com/news-releases/district-m--sharethrough-have-entered-a-definitive-agreement-to-merge-and-become-one-of-the-leading-independent-omnichannel-ad-exchanges-301224954.html Accessed 2026-09-27. Access: partial extract. [801] Sharethrough Bids Farewell To Top Execs Amid Shift To Programmatic. AdExchanger (James Hercher), 2017-06-16. https://www.adexchanger.com/online-advertising/sharethrough-bids-farewell-top-execs-amid-shift-programmatic/ Accessed 2026-09-27. [802] District M And Sharethrough Merge To Form Single Exchange. AdExchanger (Sarah Sluis), 2021-02-09. https://www.adexchanger.com/platforms/district-m-and-sharethrough-merge-to-form-single-exchange/ Accessed 2026-09-27. [803] Following Their Merger, District M & Sharethrough Will Now Operate Under Sharethrough As The Company Name Going Forward. GlobeNewswire (company release), 2021-05-10. https://www.globenewswire.com/news-release/2021/05/10/2226230/0/en/Following-Their-Merger-District-M-Sharethrough-Will-Now-Operate-Under-Sharethrough-As-The-Company-Name-Going-Forward.html Accessed 2026-09-27. Access: partial extract. [804] Equativ and Sharethrough Merge ... (Sharethrough copy). Sharethrough (press release), 2024-06-12. https://www.sharethrough.com/company/press/sharethrough-and-equativ-are-merging Accessed 2026-09-27. Access: partial extract. [805] Sharethrough Inc. completes merger with Equativ (matter page). McCarthy Tétrault LLP (Sharethrough's counsel), 2024-06-12. https://www.mccarthy.ca/en/experience/sharethrough-inc-completes-merger-with-equativ Accessed 2026-09-27. Access: partial extract. [806] Building a More Trusted, Transparent Web: Mozilla Ads Partners with Equativ. Equativ / Mozilla (press release), 2025-11-12. https://www.equativ.com/press/building-a-more-trusted-transparent-web-mozilla-ads-partners-with-equativ Accessed 2026-09-27. [807] Equativ Strengthens Americas and Global Leadership with Two Executive Moves. Equativ (press release), 2026-04-20. https://www.equativ.com/press/equativ-strengthens-americas-and-global-leadership-with-two-executive-moves Accessed 2026-09-27. [808] Exclusive: Equativ is acquiring Sharethrough to compete among undifferentiated SSPs. Adweek (Catherine Perloff), 2024-06. https://www.adweek.com/programmatic/exclusive-equativ-is-acquiring-sharethrough-to-compete-among-undifferentiated-ssps/ Accessed 2026-09-27. Access: paywalled excerpt.