FTC Notice of 1 October 2026 Asks Whether Ad Platforms Such as Google, Meta and Amazon Should Bear the Cost of Impersonation Scam Ads Their AI Tools Optimize
In 1920 the economist Arthur Pigou used a small example that lawyers and economists still argue about. Sparks from a steam locomotive set fire to woods and fields along the line, and the railway did not pay for the damage. Pigou treated the burned crops as a cost of running the trains that someone other than the railway bore. Forty years later Ronald Coase took the same sparks and asked a harder question in The Problem of Social Cost: who is best placed to prevent the fire, and what does it cost to make them?
The Federal Trade Commission cites Coase in a footnote on the first page of the advance notice of proposed rulemaking it published on 1 October 2026. In the FTC's version of the story the locomotive is the advertising machinery of large online platforms, and the sparks are impersonation scams. The ad is the fake Medicare offer shown to a retiree, the lookalike bank result at the top of a search page, the bogus brand giveaway in a social feed. The notice asks whether the platforms that tune and deliver those ads should carry part of the cost of the fire.
Practical takeaway. The FTC is testing a theory that a platform's own AI ad tools, which write, assemble, target and price ads, are the platform's conduct and can be regulated under Section 5 of the FTC Act. Any firm that sells optimization to third-party advertisers, or buys it, should read the 30 November comment window as the first formal step toward pre-screening and takedown duties for impersonation ads.
What the FTC published on 1 October 2026: an advance notice under 16 CFR part 461
The document runs from page 62347 to page 62357 of volume 91 of the Federal Register, under RIN 3084-AB90. It is an advance notice, the earliest formal stage of an FTC trade regulation rulemaking, and it proposes no rule text. It asks whether the Commission should amend the existing Rule on Impersonation of Government and Businesses at 16 CFR part 461, or write a new rule, to address what it calls "Platform UDAPs," meaning unfair or deceptive acts or practices by platforms. The notice defines platforms broadly as public-facing websites or apps on which third parties offer goods or services through paid ads or other listings, and lists Google.com, Facebook.com, Amazon.com, the Apple App Store and LinkedIn.com as examples. Comments close on 30 November 2026. OMB has determined that the notice is a significant regulatory action.
The scale figures come from the FTC's own data. In its June 2026 release, consumers reported losing about $16 billion to fraud in 2025, 25 percent more than in 2024. Impersonation was the most reported fraud, with more than one million reports and nearly $3.5 billion in losses. Almost 30 percent of people who reported losing money said the contact began on social media, with reported losses of $2.1 billion. The Commission's older-consumers report estimated that the true cost of fraud in 2024, adjusted for underreporting, could reach $195.9 billion.
Why the FTC dropped a broader platform provision in December 2024 and returned with a narrower one
This is the Commission's second attempt. When it finalized the Impersonation Rule in March 2024, it had also proposed a "means and instrumentalities" provision that would have made it unlawful to provide goods or services with knowledge or reason to know they would be used to impersonate. Commenters called that language overbroad and objected that it arrived through a supplemental notice instead of an advance notice. On 26 December 2024 the Commission said it would not proceed with the provision "at this time." A separate proposal to cover the impersonation of individuals went to an informal hearing in January 2025 and is still under evaluation.
The 2026 notice answers both objections. Any new provision would cover only ad-optimization practices by digital marketplace platforms, so it is necessarily narrower than the 2024 version. The Commission is also starting with an advance notice this time, which the 2024 commenters asked for.
How AI ad tools from Meta and Google enter the FTC's theory of platform conduct
The central move in the notice is to treat optimization as something the platform does. The tools it describes can write ad copy, generate images and video, enhance product listings, and use a person's demographic profile, browsing and search history to decide who sees what. It quotes Meta's description of Advantage+ as a suite that uses AI "to optimize campaigns in real-time and match ads to the people most likely to take action," and Google's statement that, once an advertiser uploads assets, "Google's AI will then find the best performing ad combinations." It cites a federal court in FTC v. Doxo, where an expert explained that search ads are assembled dynamically by the platform from up to 15 headlines supplied by the advertiser. The same machinery that serves a regional bakery serves a scammer who copies a bank's logo.
The incentive argument rests on a Reuters investigation by Jeff Horwitz. Reuters reported that internal Meta documents estimated about 10 percent of Meta's 2024 revenue, roughly $16 billion, came from ads for scams and banned goods, a figure Meta disputed and the FTC summarizes as scam-ad revenue. According to the same documents, users saw roughly 15 billion "higher-risk" scam ads a day. Advertisers were typically banned only when automated review reached 95 percent certainty of fraud. Lower levels of suspicion triggered higher ad prices. These are findings of a press investigation, which the Commission cites as reported. A watchdog study the notice also cites found that the 30 most active Medicare scam advertisers on Meta drew about 215 million impressions in 2025, 73 percent of them among people over 65.
The Commission frames the harm in market terms as well. It cites George Akerlof's 1970 "market for lemons" paper. When buyers cannot tell real sellers from fake ones, trust falls and honest firms lose sales. A pharmacy chain, a toll authority or a state benefits office pays twice when someone impersonates it: in lost business and in the cost of warning the public.
Why Section 230 may not cover ad optimization, according to the FTC's reading of Roommates, Anderson v. TikTok and Moody
Every platform lawyer will turn first to the notice's treatment of Section 230 of the Communications Decency Act. The Commission calls the statute an affirmative defense that protects platforms from liability for merely hosting third-party content. It then lists the limits. Under the Ninth Circuit's Roommates.com test, a platform that materially contributes to what makes content unlawful is treated as developing that content. In Anderson v. TikTok (2024), the Third Circuit held that algorithmic recommendation is the platform's own expressive activity and may fall outside Section 230. That court relied on the Supreme Court's reasoning in Moody v. NetChoice (2024) that curating others' speech is the platform's own expression. The FTC's conclusion is that any service that enhances an ad's effectiveness may sit outside the "publisher or speaker" protection.
The Commission is careful to call this a fact-intensive question, and it asks commenters for material on it. Other circuits have held the opposite on recommendation, and the Supreme Court has not decided how Moody affects Section 230. That disagreement is likely to decide whether a final rule survives in court.
Which obligations the FTC floats: pre-screening, a reporting tool, takedown and a possible safe harbor
Section E of the notice sketches the possible content of a rule. Platforms could be required to evaluate ads before they run and to stop providing optimization services to advertisers engaged in impersonation. They could be required to run a detection program, give consumers a clear tool to report suspected impersonation ads, investigate those reports, remove confirmed ads, end optimization for them and discipline the advertiser. The Commission suggests that these steps might instead form a safe harbor, a defense for platforms that follow them, with recordkeeping to support enforcement. It notes that platforms already run takedown processes under Section 512 of the Digital Millennium Copyright Act, the TAKE IT DOWN Act and the INFORM Consumers Act. Section 512 covers copyright but not the trademarks that impersonators usually copy, and the INFORM Act reaches only high-volume sellers on online marketplaces.
One question goes straight to AI policy. Question E.1.i asks whether any requirement "would discourage Platforms from innovating their services or tools, particularly with respect to services or tools incorporating the use of artificial intelligence." Question E.1.n asks whether liability should depend on some level of knowledge of, or participation in, the scam. The answers to those two questions will largely set how far a rule reaches.
The money matters too. The Supreme Court's 2021 decision in AMG Capital Management v. FTC foreclosed monetary relief under Section 13(b) of the FTC Act. The notice points to Section 19 as the Commission's route to consumer redress: a rule violation lets it seek redress in a single federal court action, without first running an administrative proceeding and then a second suit. A rule would change the economics of enforcement against platforms far more than one more lawsuit would.
How Quentir Reads It
In our reading, the notice extends to platforms a line of thinking the Commission has applied to advertisers: a firm answers under Section 5 for its own commercial conduct, including what its AI tools say and do. On the advertiser side, the FTC's proposed policy statement on deceptive accuracy claims treated an AI accuracy claim as part of the product being sold. Here the notice proposes a similar view of the platform: if its model wrote the headline, picked the audience and set the price, the platform would have difficulty describing itself as a passive host. No court or final rule has yet established that position. That reasoning will not stay inside consumer-protection law. It bears on how product-liability claims, the EU's Digital Services Act duties for very large platforms, and contract allocations between advertisers and ad networks treat automated creative tools.
The open legal risk is Section 230, and the open policy risk is the one the FTC names itself in Question E.1.i: a pre-screening duty written too broadly could slow the same AI review systems that platforms now use to catch scams. The most likely path is a safe-harbor design, with detection, reporting and removal duties that platforms can document. For the people the notice is about, mostly older adults who trust a government seal or a bank's name on a screen, the practical change would be fewer impersonation ads in their feeds, delivered by the companies best placed to stop them. Readers who follow this file from the Impersonation Rule through Section 230 case law and AI disclosure rules will find our earlier analyses collected in one place with an All-access membership, which opens Quentir's publications and archive in one subscription. The question for 30 November is whether Google, Meta and Amazon will argue that their optimization tools are protected speech, or argue for the safe harbor.
Sources: Federal Trade Commission, Rule on Impersonation of Government and Businesses, advance notice of proposed rulemaking, 91 FR 62347 (Federal Register, 1 October 2026; RIN 3084-AB90; comments due 30 November 2026); 16 CFR part 461, Rule on Impersonation of Government and Businesses (eCFR; final rule 89 FR 15017, 1 March 2024); Federal Trade Commission, "FTC Data Show People Reported Losing $3.5 Billion to Imposter Scams in 2025" (15 June 2026); Jeff Horwitz, Reuters, "Meta is earning a fortune on a deluge of fraudulent ads, documents show" (2025, as cited in the notice); Third Circuit, Anderson v. TikTok, Inc., 116 F.4th 180 (2024); Supreme Court of the United States, Moody v. NetChoice, LLC, 603 U.S. 707 (2024); Fair Housing Council of San Fernando Valley v. Roommates.com, LLC, 521 F.3d 1157 (9th Cir. 2008); AMG Capital Management, LLC v. FTC, 593 U.S. 67 (2021); A.C. Pigou, The Economics of Welfare (1920); R.H. Coase, "The Problem of Social Cost," 3 J.L. & Econ. 1 (1960); G.A. Akerlof, "The Market for 'Lemons'," 84 Q.J. Econ. 488 (1970). Quentir, When an AI accuracy claim becomes the product.
Published intelligence, built to inform your own decisions. Published: October 1, 2026.