The Video Advertising Bureau today circulated a 13-page analysis of Google's 2025 advertising enforcement data, setting the 3.3 million US advertiser accounts Google suspended last year against the 86,800 brand advertisers that ran on national and local broadcast television in the same period - a ratio of roughly 38 to 1 - and arguing that Google's own disclosures leave open how much harmful advertising reached consumers before it was caught.

In Short

Google blocked or removed more than 8 billion ads and shut down about 25 million advertiser accounts in 2025, and a trade group that represents TV companies has now repackaged those numbers to argue that Google's ad systems attract a lot of bad actors. This matters to brands and publishers because the group says Google does not tell anyone how long the bad ads that slipped through were live, how many people saw them, or where they ran. Nothing about Google's rules changes, but the dispute over whether TV or Google is the safer place to spend ad money now has a new set of talking points on the TV side.

What VAB published

The document, titled "Bad Actors: Examining How Much Objectionable Advertising Google Takes Action Against," belongs to the trade body's Fast Facts series and was distributed to media in a weekly newsletter on September 15, 2026. It carries no separate publication date on its pages. Every slide bears a notice restricting the material to VAB members and qualified marketers, and the landing page offers the deck as both a PowerPoint and a PDF.

No new data collection sits behind it. According to VAB, the analysis draws on two Google documents released in April 2026: the 2025 Global Ads Safety Report and the 2025 Ads Safety Report for the United States. The only original figures in the deck come from a VAB analysis of Nielsen Ad Intel covering January 1 to December 31, 2025, which supplies the television side of the comparison.

VAB describes its scope in four parts: the number of ads blocked, removed or restricted, the number of advertiser accounts suspended, the volume of web pages blocked or restricted, and the number of publisher sites actioned. The landing page frames the purpose more bluntly, describing it as "Analyzing the findings from Google's 2025 brand safety report."

PPC Land covered the underlying Google data when Google's 2025 Ads Safety Report documented 8.3 billion ads blocked or removed in April. The additive element in the VAB version is not the numbers themselves but the frame placed around them, and the US-specific figures, which received less attention at the time.

The global numbers

According to the deck, Google blocked or removed more than 8.3 billion ads and suspended more than 24.9 million advertiser accounts worldwide in 2025. Enforcement against publishers reached more than 480 million web pages blocked or restricted and more than 245,000 publisher sites actioned. VAB's footnote defines those publisher actions as ad demonetization, ad termination and account suspension.

Scams account for a defined slice. According to VAB's reading of the Google report, 602 million ads were removed for policy violations most closely associated with scams, and more than 4 million accounts were suspended for scam-related activity. Those figures work out to roughly 7% of all blocked or removed ads but about 16% of all suspended accounts - a gap suggesting that scam enforcement leans more heavily on account-level action than the enforcement programme as a whole.

The category breakdown runs across 16 policy areas. Abusing the ad network leads at more than 1.29 billion ads. VAB's footnote, drawn from Google's definitions, says that category covers malicious or unwanted software, hacked websites, gaining an unfair advantage in the ad auction, disguising ad elements to avoid detection, creating new accounts to evade enforcement and violating Google Search spam policies.

Personalization violations follow at more than 755 million ads, a category covering ads or targeting that uses sensitive user data, personal traits or intimate life situations. Legal requirements account for 646.7 million, misrepresentation for 421.5 million and trademark for 372.7 million. The remaining categories, in descending order:

Dating and companionship reached 354.2 million, financial services 327.8 million and sexual content 321 million. Gambling and games came to 270.7 million, copyright to 229.4 million and inappropriate content to 147 million. Healthcare and medicines stood at 87.8 million and alcohol at 20.6 million, while enabling dishonest behaviour and dangerous products or services each came to 13.1 million. Counterfeit goods closed the list at 513,000.

The dating and companionship definition is among the more detailed in the deck. It includes prostitution, sex tourism, sugar dating, deceptive dating services built on AI and deepfakes, exploitative dating aimed at victims of crisis and people with disabilities, and mail-order spouses.

What the category figures do not add up to

The 16 listed categories total approximately 5.27 billion ads by PPC Land's calculation, about 63% of the 8.3 billion headline figure. The deck does not explain what accounts for the remaining 3 billion or so. Google's policy set includes more categories than the chart displays, and a single ad can in principle be actioned under more than one policy, so the gap need not indicate an error. It does mean the chart cannot be read as a complete distribution.

A second inconsistency sits in the deck's own conclusions. The "Key Marketer Takeaways" slide lists trademark and copyright infringement, sexual content, fraudulent actions and counterfeit goods among the top reasons ads were removed. By VAB's own chart, counterfeit goods ranked last of the 16 categories at 513,000 ads, and copyright ranked tenth. Personalization violations and legal requirements, the second and third largest categories, do not appear in that takeaway list at all.

The PDF also carries a stray line in its text layer on the takeaways page, a sentence about adults over 50 accounting for 36% of the population, which does not appear on the rendered slide and has no connection to the subject. It appears to be residue from a template.

Restricted ads present a third gap. VAB lists ads "blocked, removed or restricted" in its scope, but the deck contains no restricted-ad figure. Google's global report put restricted ads at 4.8 billion for 2025, down from 9.1 billion in 2024, as PPC Land reported in April. Restricted ads are permitted to run under conditions such as limits on targeting or placement, which is a different outcome from removal.

Publisher pages: mostly a restricted category

The 480 million pages figure invites the reading that nearly half a billion pages were banned from monetization. The detail in VAB's own slide points somewhere else. Sexual content accounted for more than 409 million pages, about 85% of the total. According to the footnote, that is a restricted category covering partial nudity, sexual merchandise and nudity, "where ads may run in limited scenarios."

The remaining categories are far smaller. Dangerous and derogatory content reached 20.5 million pages, shocking content 15 million, weapons promotion and sales 12.8 million and online gambling 9.7 million. Alcohol and tobacco each came to 5 million. Sexually explicit content, which VAB's footnote describes as pornography and not allowed, and malware or unwanted software each came to 2 million pages.

Those nine categories sum to about 481 million pages, consistent with the headline. PPC Land's April coverage added that AI-driven systems contributed to detecting more than 467 million of those pages, according to Google.

For advertisers, the distinction matters in practice. A page restricted for sexual content can still carry ads under Google's rules, subject to limits, which puts the burden of brand suitability decisions back on the buyer's own controls rather than on Google's enforcement tallies.

The United States figures

The US report supplied the numbers VAB uses for its television comparison. According to the deck, Google blocked or removed 1.7 billion ads and suspended 3.3 million advertiser accounts in the US in 2025. Against the global totals, the US represented about 20% of blocked or removed ads and about 13% of suspensions.

VAB presents only a ranking, not volumes, for the top five US policy violations. Abusing the ad network ranked first, followed by misrepresentation, sexual content, personalization violations and dating and companionship. The ordering differs from the global chart, where legal requirements and trademark sit ahead of sexual content. Misrepresentation, the category most directly associated with scams, climbs from fourth globally to second in the US.

The television comparison

The slide that gives the deck its argument sets Google's US enforcement against linear television. On one side sit Google's 1.7 billion blocked or removed ads and 3.3 million suspended accounts. On the other sit 146 million total ad units aired on national TV and local broadcast TV, and 86,800 brand advertisers on the same outlets, according to VAB's analysis of Nielsen Ad Intel.

VAB's headline for the slide says the comparison "illustrates just how pervasive and habitual 'bad actors' are across Google's ad ecosystem." The takeaways page goes further: "The pervasiveness of 'bad actors' across Google's ad ecosystem far exceeds the number of legitimate advertisers in brand-safe TV environments."

The arithmetic is accurate. Google's US suspensions are about 38 times the number of TV brand advertisers, and its US blocked or removed ads are about 11.6 times the number of TV ad units aired. Whether the comparison measures what the headline claims is less clear, and the footnotes themselves supply the reasons.

The TV side counts ad units that actually aired, on an unequivalized basis, meaning a 15-second and a 60-second spot each count once. The Google side counts ads that were blocked or removed, most of which, according to Google, were never served at all. The TV advertiser count uses Nielsen's brand variants, while Google counts accounts, and a single operator can open many accounts. VAB's own definition of the largest category includes creating new accounts to evade enforcement. The TV scope covers network broadcast, Spanish-language network broadcast, national cable, national Spanish-language cable, syndication and spot local TV, and excludes local cable.

The comparison also involves systems with very different entry requirements. Television inventory is sold through negotiated relationships with a limited set of sellers, while Google Ads accepts self-serve sign-ups at global scale. A higher count of suspended accounts on an open system is partly a function of openness and partly a function of how aggressively the platform enforces. The deck does not separate the two.

Then there is the direction of travel. Google's suspensions fell from 39.2 million in 2024 to 24.9 million in 2025, a decline of roughly 36%, which Google attributed to more precise targeting of genuinely malicious accounts. In November 2025, Google said it had cut incorrect advertiser suspensions by more than 80%. The VAB deck presents the 2025 suspension figure as evidence of pervasive bad actors without noting that it is lower than the year before, or that some share of suspended accounts belonged to legitimate advertisers.

The 99% question

The deck's most substantive point concerns what Google does not disclose. According to VAB, Google reports that its Gemini-powered systems stopped more than 99% of policy-violating ads in 2025 before they were served. VAB writes that the report "provides limited visibility into the ads that were not stopped before serving, including how long they remained active, how many people saw them or where they appeared."

The landing page compresses the argument to a single line: "1% can be millions of ads."

The two organisations do not describe the 99% figure in quite the same way. Google's report, as PPC Land reported in April, says more than 99% of the ads it blocked or removed were stopped before anyone saw them. VAB's deck describes the figure as a share of all policy-violating ads. The difference is material. Google's version measures its enforcement queue. It says nothing about violating ads that were never detected at all, which by definition sit outside any enforcement count.

Taken on Google's own terms, less than 1% of 8.3 billion still allows for up to roughly 83 million ads that served before they were removed. That is a ceiling, not an estimate, and neither document supplies the actual figure. What VAB frames as a question - "how much unsafe, deceptive or brand-inappropriate advertising reaches consumers before it is identified and removed" - cannot be answered from either Google report, and the deck does not attempt to answer it either.

Recent cases show why the question is live. In July, Confiant documented the SourTrade malvertising operation, which assembles malware inside victims' browsers and whose landing pages contained conversion logic for Google Ads alongside Meta and X tracking. Such techniques are built specifically to pass the kind of pre-serving review that produces the 99% figure. Post-bid verification exists precisely because a bid-time or submission-time check is a claim about what will run, not an observation of what did.

Who is making the argument

VAB represents premium multiscreen television providers and distributors. Its commercial position is not neutral on whether advertisers perceive Google as a safe environment, and the "Bad Actors" deck fits a sustained pattern of output that PPC Land has tracked through 2026.

In February, VAB and TVision published The Impression Gap, which found premium video platforms ahead of YouTube on every connected TV attention metric measured. In March, VAB accused Nielsen of suppressing a February edition of The Gauge that would have shown linear television ahead of streaming, a dispute analysts later framed as a contest over who controls TV money. In July, the same Fast Facts format was applied to Meta, when a VAB analysis found Facebook banned 3.5 billion fake accounts in 2025. And earlier this month, VAB's leisure time infographic reported video at 51% of American leisure time for a fourth consecutive year.

The Google deck promotes that wider body of work directly. Its final content pages point to "Deleted Scenes," a VAB analysis of YouTube content removals, and to other VAB material on transparency and brand safety, including "Where do ad dollars go when they're spent with Google?" and "Friend or Frenemy?" on Facebook. The thumbnails reproduced in the deck show YouTube removing 219.3 million videos, 154.4 million channels and 30.3 billion comments between 2019 and 2025, with almost 41 million videos removed in 2025 alone, up 15% from 2024. VAB's rough estimate puts that at 965 years of video content in 2025, against 837 years in 2024. Those figures describe content moderation rather than advertising enforcement, a distinction the deck's closing headline blurs by folding ads and content into a single claim about objectionable material on Google platforms.

None of this makes the underlying data wrong. The numbers are Google's, and VAB reproduces them accurately where they can be checked against PPC Land's April reporting. The selection and framing, however, come from a party with a direct interest in the outcome of budget decisions between television and digital platforms.

Why the enforcement data matters to the marketing community

For advertisers and agencies, Google's enforcement scale cuts in two directions. The same systems that removed 8.3 billion ads also generate false positives, and the operational cost of those errors has shaped several policy moves this year. In July, PPC Land reported that Google Ads ended appeals for policy decisions more than six months old. In August, Google set out plans that will cost unqualified advertisers unlimited impressions by 2028 under its limited ad serving policy, a form of restriction that, unlike suspension, is neither visible nor appealable in the conventional sense.

Scam advertising has also become a legal and regulatory exposure for platforms, not only a reputational one. Reuters reported in November 2025 that Meta internally projected roughly 10% of its 2024 revenue from ads for scams and banned goods, and in April 2026 the Consumer Federation of America sued Meta on the strength of those disclosures. Google has published no comparable revenue figure for ads it later removed, and the VAB deck does not claim one exists.

For publishers, the 245,000 sites actioned and the dominance of the sexual content category indicate where Google's publisher enforcement concentrates. The figures give little sense of how many legitimate news or entertainment pages were caught in restricted classifications, a recurring complaint from publishers whose revenue depends on Google's network.

The deck's central question - how much violating advertising reaches people before removal - is a fair one. Google's reports measure the work its systems do, not the exposure that work fails to prevent. VAB has put the gap on the record. What it has not done is measure it, and its television comparison rests on counting methods that the footnotes show are not equivalent.

Timeline

Summary

Who: The Video Advertising Bureau, a trade organisation representing premium multiscreen television providers and distributors, produced the analysis. The data comes from Google's 2025 global and US Ads Safety Reports, with television figures from a VAB analysis of Nielsen Ad Intel.

What: A 13-page Fast Facts deck reproducing Google's 2025 enforcement figures - 8.3 billion ads blocked or removed, 24.9 million accounts suspended, 602 million scam-linked ads removed, 480 million pages and 245,000 publisher sites actioned globally, and 1.7 billion ads and 3.3 million accounts in the US - and comparing the US figures with 146 million TV ad units and 86,800 TV brand advertisers. The deck argues that Google discloses too little about violating ads that served before removal. Its category chart accounts for about 5.27 billion of the 8.3 billion ads, its takeaways misstate the rank of counterfeit goods and copyright, and its TV comparison relies on counting methods that are not equivalent.

When: VAB distributed the deck to media on September 15, 2026. The Google reports it analyses were released in April 2026 and cover calendar year 2025; the Nielsen data covers January 1 to December 31, 2025.

Where: The enforcement data covers Google's advertising and publisher network worldwide, with a separate US breakdown. The TV comparison covers US national broadcast, national cable, syndication and spot local TV, excluding local cable.

Why: The deck feeds a continuing contest over whether advertising budgets belong on television or on Google's platforms. For marketers, it raises a legitimate disclosure gap - how long violating ads ran and how many people saw them - while presenting Google's enforcement volume in a frame chosen by a party with a direct commercial stake in the answer.