The Video Advertising Bureau, the marketing body of the US television industry, has sent members and marketers a 25-page supplement titled "More Illusions of the Internet," adding 12 statistics on bad bots, fake social media followers, ad fraud, AI-generated content and consumer scams to the 20 it published on February 16, 2026. The deck carries no publication date and contains no new research by VAB. Every figure comes from Thales, SociaVault, Juniper Research, the Association of National Advertisers, the Federal Trade Commission, a survey for Ping Identity and an academic study, and the headline advertising number - between $530.1 million and $978.7 million of US programmatic spending on AI slop - is VAB's own multiplication of two of those sources.

In Short

A group that promotes TV advertising put together a slideshow of other organizations' statistics showing how much of the internet is fake, from robot visitors to scam ads to websites written by AI. That matters to companies buying online ads, because fake visitors and junk websites soak up ad money, and to ordinary people who lose money to scams that start online. The numbers are borrowed rather than new, and three of them do not add up when traced back to where they came from, so the deck works best as a pointer to the original studies.

A sequel built from other people's data

The supplement belongs to VAB's "Fast Facts" series, and every page carries the same footer: "This information is exclusively provided to VAB members and qualified marketers." It follows "The Illusions of the Internet: Uncovering the Fallacies & Realities of Audience, Advertising & Content," which, according to VAB, was released on February 16, 2026. That earlier guide listed 20 "realities" across four components of the internet: audience, advertising, products and services, and content. Its topics ran from opaque ad placements and app store ad fraud to counterfeit goods and pirated content.

VAB explains the timing in its introduction. "With news of investigative analyses, consumer lawsuits and advertising fraud being a nearly daily occurrence, we decided it was time to develop a supplement to our original report only nine months after its first release," the deck states. Nine months from mid-February points to mid-November. The supplement itself shows no date, and the most recent item it references is a New York Times article from August 17, 2026.

The tone is set on the second page. "The internet continues to face a crisis of reality which is only getting worse," according to VAB, which concludes that "combined, these two reports show an internet ecosystem only getting worse, not better, for both consumers and advertisers alike."

The cover promises realities "Across Audience, Advertising & Content," yet four of the 12 sit in a fourth section, Products & Services. The full list breaks down as follows:

  • Audience: growth of bad bots, bad bot attacks by country, AI-powered malicious bots, fake followers on social media
  • Advertising: digital ad fraud by region, AI slop in advertising
  • Products & Services: consumer fraud exposure by platform, cost of consumer fraud by platform, the consumer fraud trend on social media, fraud losses on social media by age
  • Content: "modern" scams, AI-generated or AI-assisted websites

VAB describes itself on the final page as playing "a dual role in the video advertising industry," adding: "We are leading the change to bring about a more innovative and transparent marketplace." Its recent output has followed a pattern that PPC Land has tracked through the year. In July, PPC Land reported a VAB analysis finding that Facebook banned 3.5 billion fake accounts in 2025. In August, VAB led an email to marketers with a 94% completion rate drawn from 2023 FreeWheel data. In September, a 13-page deck compared Google's 3.3 million US ad account suspensions with the 86,800 brands that advertised on TV. And on October 21, PPC Land reported on a VAB infographic that reused February study figures without their source or method. The new supplement follows the same model: a TV trade body repackaging published third-party data about digital platforms.

Audience: bots and fake followers

Bad bots at 40% of traffic

The supplement opens with Thales's "Bad Bot Report: Bad Bots in the Agentic Age, 2026 Edition." According to the chart VAB reproduces, malicious bots generated 40% of global internet traffic in 2025, so-called good bots 13% and humans 47%. The series runs back to 2015, when the split was 15% bad bots, 27% good bots and 58% humans. The bad bot share has risen every year since 2018, passing 20% in 2018, 24% in 2019, 26% in 2020, 28% in 2021, 30% in 2022, 33% in 2023 and 37% in 2024. The human share peaked at 63% in 2019.

VAB's headline claims that "non-human traffic has outpaced real humans for the last two years." The arithmetic supports it: bots combined accounted for 51% in 2024 against 49% for humans, and 53% in 2025 against 47%.

The deck's footnote defines bad bots as "malicious automated programs that exploit applications to extract data, commit fraud and disrupt services," enabling credential stuffing, scalping and distributed denial-of-service attacks. Good bots are "harmless automated traffic such as search engines and AI crawlers." According to Thales, which published the report in its 13th consecutive edition and presented it in a blog post dated April 30, 2026, the company blocked 17.2 trillion bad bot requests in 2025, and AI-driven bot attacks rose from 2 million to 25 million a day between 2024 and 2025.

Other measurements land in a similar range without matching it exactly. PPC Land reported in July that Imperva put automated traffic at 53% of the web, a figure consistent with the Thales total. Cloudflare Radar data for the week ending June 5, 2026, by contrast, put bots at 57.4% of requests for HTML content, with AI training crawlers alone at 50.6%. Each vendor counts traffic it can see, under its own definitions.

The United States as target

Two country charts follow. According to Thales, 59% of bad bot attacks targeted the United States in 2025, followed by Australia at 12% and the United Kingdom at 9%. France took 3%; Canada, Germany, Singapore, Spain and Colombia 2% each; Mexico 1%; and all other countries 6%. VAB sets each figure beside the country's share of world population, based on 2025 data from TheGlobalEconomy.com: 4.19% for the US, 0.34% for Australia and 0.85% for the UK. By PPC Land's calculation, the US draws attacks at roughly 14 times its population weight. The 59% figure is "five times more than the second most targeted country," according to VAB.

AI-powered attacks show the same skew at a slightly lower level. The US accounted for 49% of AI-driven bot attacks, according to the Thales data in the deck, eight times France's 6%. The United Kingdom, Japan, Indonesia and Germany took 5% each, Canada 4%, Australia 3%, and Italy and Spain 2% each, with 14% spread across all other countries.

These charts describe targets, not origins. A separate analysis by Decodo, covered by PPC Land in August, found that the US also originates 53.5% of global bot traffic, followed by Germany at 8.2% and the Netherlands at 5.6%. The supplement does not address where the attacks come from.

Fake followers and a $4.6 billion figure that does not compute

The fourth audience slide draws on SociaVault's "The Fake Follower Problem: 2026 State of Influencer Fraud." VAB's headline reads: "Over one-third of online influencers' followers are believed to be fake, costing marketers billions of dollars annually in wasted spend." Its pie chart, titled "Online Influencer Followers," shows a 37.2% "fraud rate," made up of 14.8% "likely fraudulent" and 22.4% "suspicious," against 62.8% "likely authentic." Instagram's combined rate is 41.8% (25.1% suspicious, 16.7% likely fraudulent) and TikTok's 32.6% (19.7% suspicious, 12.9% likely fraudulent).

SociaVault's report page describes a different unit of measurement. The company analyzed 100,000 influencer accounts - 50,000 on Instagram and 50,000 on TikTok - across five follower tiers and ten content niches, collecting the data between February 10 and 16, 2026 through its own API. Each account was scored on 12 indicators and then classified as a whole: "suspicious" with a score of 25 to 49 or three to four flags, "likely fraudulent" with a score of 50 or more or five or more flags. The percentages are therefore shares of influencer accounts showing signs of inauthentic followers or engagement, which is how SociaVault's executive summary phrases it, not a count of individual followers. SociaVault's own key finding uses the followers wording, and VAB repeats it. The 37.2% also includes the suspicious tier, not only accounts flagged as likely fraudulent.

VAB's slide pairs the rate with a dollar figure: $4.6 billion in "estimated annual brand spend wasted on fake follower reach." The footnote explains: "Estimated annual brand spend wasted on fake reach is based on 37.2% fraud rate applied to $24 billion total influencer spend in 2025." That calculation does not produce $4.6 billion. By PPC Land's calculation, 37.2% of $24 billion is about $8.9 billion; $4.6 billion corresponds to roughly 19.2%. SociaVault's own blog post of March 9, 2026 arrives at $4.6 billion a different way, applying a weighted rate of 19.2% to a projected $24 billion of influencer spending in 2026, and lists the inputs to its weighting model without publishing the weights. The footnote VAB reproduces, which also appears on SociaVault's report page, describes neither that rate nor that year.

Advertising: fraud by region and AI slop

Juniper's 2028 forecast

The advertising section begins with a forecast first published more than two years ago. According to Juniper Research, which in July 2024 analyzed data sets from 78,000 unique sources across 45 countries in eight regions, North America will account for 42% of global digital ad fraud in 2028, against 5% of world population (using United Nations World Population Prospects data from October 2024).

The full regional split, as VAB presents it, reads:

  • North America: 42% of fraud, 5% of population
  • Far East and China: 20% of fraud, 21% of population
  • Western Europe: 17% of fraud, 6% of population
  • Central and Eastern Europe: 7% of fraud, 4% of population
  • Rest of Asia-Pacific: 4% of fraud, 16% of population
  • Latin America: 4% of fraud, 8% of population
  • Indian Subcontinent: 3% of fraud, 21% of population
  • Africa and Middle East: 3% of fraud, 16% of population

Dollar values under each bar - $72.2 billion for North America, $34.4 billion for the Far East and China, $29.2 billion for Western Europe, $12.0 billion for Central and Eastern Europe, $6.9 billion each for the rest of Asia-Pacific and Latin America, and $5.2 billion each for the Indian Subcontinent and Africa and the Middle East - come from VAB applying the rounded shares to a $172 billion global fraud forecast for 2028. The numbers check out: 42% of $172 billion is $72.24 billion.

The forecast is a projection for a year still ahead, published before any of the 2025 data elsewhere in the deck. Estimates of current losses vary widely. Lunio put 2025 losses to invalid traffic at $63 billion based on 8.51% invalid clicks across 2.7 billion paid clicks, a figure that ad fraud researcher Augustine Fou has argued undercounts fraud on CPM-priced display ads, where a bot only needs to load an ad.

AI slop: VAB's own estimate

The supplement's one original calculation sits on page 13. "As much as almost $1 billion in ad dollars is estimated to be spent on AI slop within open web programmatic spend," according to VAB, which puts the range at $530.1 million to $978.7 million.

The underlying share comes from the ANA's "Programmatic Transparency Benchmark: AI Slop Findings from Q1 2026," which found that AI slop accounted for 1.3% to 2.4% of open web programmatic spend. The deck quotes the definition: "AI slop is generated primarily for monetization, with little or no human input, originality, or audience value. High-quality AI-assisted or AI-native content is not part of this definition." VAB then multiplies that share by eMarketer's projection of $40.78 billion in US programmatic digital display ad spending for 2025. The arithmetic is correct: 1.3% of $40.78 billion is $530.1 million, and 2.4% is $978.7 million.

Three caveats sit inside that calculation. The dollar range is VAB's, not the ANA's. It applies a first-quarter 2026 share to a full-year 2025 spending base. And the label shifts between the slide's caption, which calls the base "open web U.S. programmatic ad spend," and the footnote, which calls it "U.S. programmatic digital display ad spending." According to eMarketer, US programmatic ad spending was set to reach nearly $180 billion in 2025, so the $40.78 billion figure appears to cover a narrower segment, which the deck does not name.

The ANA work behind the share was more detailed than the slide suggests. The analysis by TAG, the ANA and Fiducia published on July 28 derived the 1.3% lower bound from page-level evaluation with human verification, with a 99% confidence interval of 0.98% to 1.56%, and the 2.4% upper bound from domain-level classification. The page-level pass covered 233 million URLs, 4.45 billion impressions and $14.84 million in measured spend. Slop inventory was graded premium more than 70% of the time. It posted an invalid traffic rate of 0.05% against 0.32% for clean supply, viewability of 77.2% against 74.9%, and a TrueCPM of $7.08 against $6.15. For comparison, made-for-advertising inventory accounted for 1.1% of spend in the same quarter. None of those findings - the ones that explain why slop is hard for buyers to avoid - appear in VAB's deck.

Instead, the slide surrounds the estimate with six press headlines: The Wall Street Journal on April 6 ("Brands Adopt 'No AI' Disclaimers to Stand Out Amid the Slop"), DesignRush on June 26 ("Ad Industry Leaders Call for End to AI Slop at Cannes Lions 2026"), Adweek on July 28 ("AI Slop Is Fooling Advertisers' Verification Tools"), NBC News on August 8 ("'Stop the slop' efforts gain traction as platforms draw lines around AI content"), Wired on August 10 ("The AI Slop Backlash Is Actually Having an Impact") and The New York Times on August 17 ("Sick of A.I. Slop? So Are Tech Giants").

PPC Land has documented how the problem works in practice. In April it reported on DoubleVerify's discovery of AutoBait, a network of more than 200 domains producing AI slideshow articles for under $2.25 each, with up to eight ad banners per slide. On September 30, Integral Ad Science extended its low-quality generative AI avoidance tool to YouTube, without publishing accuracy figures for that platform.

Products and services: the FTC's scam data

Where scams start

Four of the 12 slides come from a single source: the FTC's Consumer Protection Data Spotlight "Reported Losses to Scams on Social Media Eight Times Higher Than in 2020," published on April 27, 2026, and covered by PPC Land at the time.

According to the FTC data, websites or apps were the contact method in 31% of 2025 fraud reports that indicated a loss, social media in 28%, phone calls in 11%, email in 10% and text messages in 7%. VAB adds the first two together for its headline: "Nearly 60% of reported U.S. consumer fraud originates from websites, apps and social media platforms." The FTC calculation excludes reports that gave no contact method and those classified as "other," according to the deck's footnote.

Meta's share, and a dollar figure that leaves out Instagram

Measured in dollars, social media ranks first. Losses to scams that began on social media reached $2.1 billion in 2025, split between Facebook ($794 million), WhatsApp ($425 million), Instagram ($234 million) and other platforms ($599 million). Websites or apps accounted for $1.1 billion, phone calls $1.1 billion, texts $639 million and email $569 million. VAB sums the five methods to $5.5 billion, giving social media a 38% share, websites and phone calls 20% each, texts 12% and email 10%.

The slide's headline reads: "Over $2 billion of consumer fraud originated from social media in 2025, with Meta platforms accounting for over 70% of that amount ($1.2B)." The percentage holds. By PPC Land's calculation, Facebook, WhatsApp and Instagram together account for $1.453 billion, about 71% of the $2.052 billion sum of the four social media segments. The dollar figure does not. $1.2 billion equals Facebook plus WhatsApp alone ($1.219 billion), or about 59% of the total, and leaves Instagram's $234 million out. The deck's one-page summary repeats the claim as "Meta accounts for 70% of the $2B in consumer fraud from social media."

Meta's exposure to scam advertising has been documented elsewhere. In November 2025, Reuters reported that Meta had projected about 10% of its 2024 revenue, roughly $16 billion, would come from scam and banned-goods ads. On April 21, 2026, the Consumer Federation of America sued Meta in the Superior Court of the District of Columbia over those ads.

Eight times 2020

The trend slide charts reported social media losses year by year: $261 million in 2020, $789 million in 2021, $1.2 billion in 2022, $1.5 billion in 2023, $1.9 billion in 2024 and $2.1 billion in 2025. That is the eightfold rise in the FTC's title. The FTC's separate report to Congress on older adults, released on October 19, found that people aged 60 and over lost about $750 million to scams starting on social media in 2025, roughly twelve times the 2020 figure for that group.

Ages, and a label that changes the measure

The final FTC slide breaks the data down by age: 40% for people aged 18 to 29, 32% each for the 30-39, 40-49 and 50-59 groups, 29% for 60-69, 23% for 70-79 and 14% for those 80 and over. VAB labels the chart "% Share of Total Reported U.S. Consumer Losses from Scams Originating on Social Media," and its summary page states that "social media accounts for 40% of total fraud losses for adults 18-29."

The FTC describes those percentages differently. According to the Data Spotlight, they are shares of loss reports - the proportion of each age group's reports of money lost in which the scam started on social media - not shares of dollars lost. The distinction matters because a small number of large losses can dominate dollar totals. In dollar terms, according to the FTC, social media was the costliest contact method for every age group under 80; for people 80 and over, phone calls caused the most losses.

Content: feared scams and AI-written websites

What consumers fear

The content section opens with a survey by Talker Research on behalf of Ping Identity, an identity security company. According to the deck, Talker surveyed 10,500 adults between July 28 and August 11, 2025: 2,000 in the US, 2,000 in the UK, 1,000 each in France, Germany, Australia and Singapore, and 500 each in India, Indonesia, the Netherlands, Sweden and the UAE. Ping released the results on October 1, 2025.

Asked which "modern" scams they fear most, 39% chose AI-driven phishing, defined in the deck as "personalized AI messages written using data from social media or web history that are meant to trick individuals into revealing sensitive information." Fake apps followed at 38%, deepfake attacks at 32% and voice cloning scams at 31%. Synthetic identity fraud and AI-generated influencers and personalities each drew 29%, and augmented reality scams 21%. Ping's own release added that only 23% of respondents felt very confident they could tell a legitimate offer from a scam.

The survey measures fear, not incidence, and its fieldwork took place more than a year before the supplement appeared. It sits in the deck under the question "How much online content is fake?", which it does not attempt to answer.

A third of new websites, and a finding the deck omits

The last data slide cites "The Impact of AI-Generated Text on the Internet," which VAB credits to Imperial College London and Stanford University via Shelly Palmer's April 28, 2026 post "A Third of New Websites Are AI-Generated." According to the study's project page, its authors are Jonas Dolezal of Imperial College London, Sawood Alam and Mark Graham of the Internet Archive, and Maty Bohacek of Stanford. The researchers sampled English-language pages from the Internet Archive's Wayback Machine over 33 monthly intervals from August 2022 to May 2025, targeting 10,000 URLs a month, and classified them with the Pangram v3 detector.

VAB's chart shows the share of newly published websites classed as AI-generated or AI-assisted at 17.6% in January 2025, 17.6% in February, 23.5% in March, 23.5% in April and 35.3% in May. The headline reads: "Due to tools like ChatGPT, over one-third of newly published websites are now AI-generated or AI-assisted which creates trust issues with consumers."

Beside it, VAB shows a survey result. Asked to rate the statement "As AI content becomes more common on the internet, I am encountering factually incorrect information more frequently," 4% strongly disagreed, 9% disagreed, 7% somewhat disagreed, 14% were neutral, 24% somewhat agreed, 23% agreed and 18% strongly agreed - 65% in agreement.

Two details from the paper change how that slide reads. First, the survey statement in the paper includes the words "and hallucinations," and the question was put to the part of the sample with 303 respondents, out of 853 unique US participants in total. The paper's text reports 75.1% leaning toward agreement, 14.3% neutral and 20.6% leaning toward disagreement - figures that sum to 110%. VAB's 65% matches the neutral and disagree numbers (100 minus 14.3 minus 20.6 is 65.1), so the paper's 75.1% appears to be the error.

Second, the study tested exactly the belief the survey captures. Its "truth decay" hypothesis - that more AI content means more factual errors - was not confirmed. According to the paper, the factual error rate showed no statistically significant correlation with AI prevalence (a coefficient of minus 0.19, with a p-value of 0.27). Of six hypotheses tested, two held: AI-written pages were about 33% more semantically similar to one another than human-written pages, and scored about 107% higher on positive sentiment. The researchers found that most survey respondents believed in harms their data did not support. VAB's slide presents the perception without the test.

VAB's conclusions

The supplement closes with three "Key Marketer Takeaways." The first: "Online fraud costs advertisers and consumers billions of dollars each year, with social media platforms being the major culprit for consumer scams which affects their own user base." The second: "AI has created a surge in 'bad' bot traffic, ad fraud, consumer scams, fake content and 'made-for-advertising' (MFA) websites; affecting all four main components of the internet - audience, advertising, products and content." The third: "Whether it's coordinated bot attacks or ad fraud, the United States is by far the primary target for malicious acts by bad actors and global criminals, so advertisers must remain vigilant and understand and demand full campaign transparency from their media partners."

Some of those claims reach beyond the data in the deck. None of the 12 slides measures made-for-advertising sites; that topic belonged to the February guide. And the bad bot share in the Thales series climbed steadily from 2018, years before ChatGPT's release in late 2022. The supplement's only direct link between bots and AI is the Thales count of AI-driven attacks.

A page of related reading follows. It points to VAB's analyses of Google's 2025 advertising revenue, YouTube content removals ("Deleted Scenes"), Facebook's fake account bans ("Friend or Frenemy?") and Google's ad enforcement ("Bad Actors"), along with a "Transparency & Brand Safety Hub." The second half of the page promotes VAB's "Premium Video" material, including "The Fandom Flywheel," "Best in Show" and "Expanding the Creator-Verse," and its "Upfront Planning Resources Hub."

Why this matters for marketers

The deck is advocacy: a TV industry body assembling evidence about the weaknesses of digital platforms, closing with links to its premium video research and upfront planning resources. That does not make the underlying figures wrong. The Thales bot data, the FTC loss data and the ANA slop share are all published by their original sources and broadly consistent with other measurements PPC Land has reported this year, including the FBI's count of $893 million in 2025 losses linked to complaints that mention AI tools.

What the supplement adds is framing and arithmetic, and the arithmetic is where it slips. Of the 12 slides, three carry figures that do not reconcile with the data behind them: the $1.2 billion Meta figure that drops Instagram, the $4.6 billion fake-follower cost whose stated method yields $8.9 billion, and the FTC age shares relabeled as shares of dollars. The 37.2% fake-follower rate, meanwhile, measures influencer accounts rather than followers, a wording problem that begins in SociaVault's own report. And the AI-written websites slide presents a perception that the cited study tested and did not confirm.

Several of the sources also have a commercial stake in the problems they measure. Thales sells bot protection. Ping Identity sells identity security. SociaVault drew its sample through its own data API. VAB's members sell the television inventory the deck's closing page promotes. For advertisers weighing where fraud and low-quality content sit in their media plans, the primary studies - with their methods, sample sizes and confidence intervals - carry information the slides leave out.

Timeline

Summary

Who: The Video Advertising Bureau, the marketing body of the US television industry, drawing on data from Thales, SociaVault, Juniper Research, the ANA (with eMarketer), the Federal Trade Commission, Talker Research for Ping Identity, and researchers from Imperial College London, the Internet Archive and Stanford University.

What: A 25-page supplement, "More Illusions of the Internet," adding 12 statistics on bad bots, fake followers, ad fraud, AI slop, consumer scams and AI-written websites to VAB's February guide. Its own estimate puts US ad spending on AI slop at $530.1 million to $978.7 million. Three figures do not reconcile with their sources: a $1.2 billion Meta total that omits Instagram, a $4.6 billion fake-follower cost whose stated method gives $8.9 billion, and FTC age shares relabeled as dollar shares. The 37.2% fake-follower rate counts influencer accounts, not followers.

When: The supplement is undated. VAB says its original report was released on February 16, 2026; the most recent item the supplement cites is dated August 17, 2026.

Where: Distributed by VAB to its members and marketers it describes as qualified, with data covering the United States and, for bots, ad fraud and scam fears, a range of other countries.

Why: VAB argues that "the internet continues to face a crisis of reality which is only getting worse" and urges advertisers to "demand full campaign transparency from their media partners," while pointing marketers toward its premium video research and upfront planning resources.