Celeb-bait is a form of scam advertising in which fraudsters use the names, photographs, voices or AI-generated likenesses of famous people to make a fraudulent offer look endorsed. The celebrity has agreed to nothing. The face is there to stop the scroll and lend borrowed trust to an investment scheme, a miracle health product or a fake giveaway, and the click leads to a site that asks for money or personal data. Meta, which made the term common from October 2024, describes it as the misuse of images of public figures "to lure people into ads that lead to scam sites". The tactic exists because a familiar face lowers suspicion faster than any copywriting.

How a celeb-bait campaign works

The mechanics follow a stable sequence, documented in court filings across several countries.

Creative. Operators pick a well-known person associated with money or health: a TV presenter, a founder, a physician. Older campaigns used doctored photographs and fake headlines. Newer ones use deepfakes, meaning synthetic video or cloned audio. In February 2026 Meta sued a Brazilian group, Brites Corp, alleging it used deepfakes of a prominent physician to sell unapproved health products.

Delivery. The ads are bought through ordinary self-serve tools such as Meta Ads Manager or Google Ads, usually from compromised, rented or freshly created accounts. Many use cloaking, which shows ad reviewers a harmless page and real users the scam. Meta's February 2026 filings also targeted former Business Partners accused of renting out "trusted" accounts.

Landing page. The click typically lands on a cloned news article carrying an invented interview with the celebrity.

Conversion. Victims are pushed off the platform into closed channels. In the Frankfurt Finanzfluss case, ads using co-founder Thomas Kehl's image routed users into WhatsApp investment groups. In Japan, plaintiffs said ads using the faces of entrepreneurs Yusaku Maezawa and Takafumi Horie led to LINE groups that collected money for forex and crypto "investments".

The facial recognition countermeasure

On October 21, 2024, Meta began testing facial recognition against celeb-bait ads, three years after it shut down Facebook's face recognition system and deleted templates for about one billion users. According to Monika Bickert, Meta's vice president of content policy, the system works as a backstop:

  1. Existing machine learning classifiers flag an ad as suspect.
  2. If the ad contains the image of an enrolled public figure, Meta compares the face in the ad with that person's Facebook and Instagram profile pictures.
  3. If the faces match and the ad is judged a scam, the ad is blocked.
  4. Facial data generated for the one-time comparison is deleted immediately, match or not.

The initial test enrolled about 50,000 public figures, who could opt out through Accounts Center. The UK, the European Union, South Korea and the US states of Texas and Illinois, all places with strict biometric rules, were excluded at launch. A UK version followed in March 2025 on an opt-in basis, and Meta extended the tools to more public figures in the EU, UK and South Korea in late September 2025, according to Engadget.

The roster has grown quickly. Meta put it at nearly 500,000 public figures in September 2025 and more than 500,000 by February 2026. It reported that user reports of celeb-bait ads fell 22% globally in the first half of 2025, and that facial recognition more than doubled the volume of celeb-bait ads detected and removed in testing.

Google takes a policy route rather than a biometric one. Its Misrepresentation policy has allowed immediate account suspension for public-figure impersonation since March 2024. According to its 2024 Ads Safety Report, that enforcement suspended more than 700,000 advertiser accounts and cut reports of the scam type by 90%.

Origin and evolution

The first landmark case predates the term and came from the UK. In April 2018 consumer journalist Martin Lewis sued Facebook for defamation over more than 1,000 scam ads using his name and face. He settled on January 23, 2019, when Facebook agreed to donate £3 million to Citizens Advice and build a UK scam ad reporting button.

On November 12, 2019, the Amsterdam District Court ordered Facebook to act against fake bitcoin ads using media figure John de Mol, under penalty of up to €1.1 million. More than 150 Dutch nationals had lost €1.7 million.

Australia followed. On March 18, 2022, the Australian Competition and Consumer Commission (ACCC) sued Meta in the Federal Court over crypto ads that falsely used Dick Smith, David Koch and former New South Wales premier Mike Baird. "The essence of our case is that Meta is responsible for these ads that it publishes on its platform," said Rod Sims, then ACCC chair. Mining billionaire Andrew Forrest pursued Meta in both countries. Australian prosecutors dropped his criminal case on April 12, 2024, citing insufficient evidence. In June 2024, Judge P. Casey Pitts of the Northern District of California refused to dismiss his civil suit on Section 230 grounds, finding a factual dispute over whether Meta's ad tools were "neutral tools". Japan's turn came in May 2024, when Maezawa filed his own suit in Tokyo.

Meta's October 2024 announcement gave the problem its industry name. By the time the company sued scam advertisers in Brazil, China and Vietnam on February 26, 2026, celeb-bait was a named policy category with its own litigation programme.

Why it matters for the marketing community

Celeb-bait sits uncomfortably close to legitimate advertising. It uses the same auctions, formats and targeting as real celebrity endorsement campaigns, so it competes for the same inventory and erodes trust in the format. A Reuters investigation covered by PPC Land reported internal Meta projections that about 10% of 2024 revenue, roughly $16 billion, came from scam and banned-goods ads. Meta called the 10% figure "rough and overly-inclusive".

The commercial consequences are visible. IAB Sweden expelled Meta in March 2026 over its handling of deceptive ads. Platforms have responded by tightening identity checks on buyers: Meta said in March 2026 that verified advertisers would drive 90% of its ad revenue by the end of 2026, up from 70%. Advertiser verification adds friction for legitimate small businesses as well as fraudsters. For brands that pay talent, celeb-bait dilutes genuine endorsements and raises right of publicity questions.

Limitations, criticisms and disputes

Coverage gaps. Facial matching only protects people on the roster, and only when an enrolled person has usable profile pictures. Lesser-known commentators, local doctors and executives are less likely to be enrolled. The match also runs only after classifiers flag an ad, so a scam that passes the first filter never reaches the face check.

Self-reported results. The 22% drop, the doubling of detections and the roster size are Meta's own figures and have not been independently audited. Meta's metrics also vary in kind. User reports of celeb-bait, user reports of all scam ads (down more than 50% over 15 months, per Meta in December 2025) and total removals (159 million scam ads in 2025) measure different things.

Biometrics. The tool revived a technology Meta abandoned in 2021 citing "growing societal concerns", and the company paid Texas $1.4 billion in 2024 to settle biometric data claims. The opt-out default outside Europe contrasts with opt-in enrolment in the UK and EU.

Liability. The central legal dispute is whether a platform that auctions and targets ads can claim to be a neutral host. Courts have split. Amsterdam (2019) and Frankfurt (2026) said no, at least in part. A Tokyo court went the other way in November 2026. The Forrest case is testing the same question under Section 230.

  • Celebrity endorsement is a paid, consented use of a famous person's image. Celeb-bait is its counterfeit: no contract, no consent, no product the celebrity has seen.
  • Deepfake describes the synthetic media technique. Celeb-bait describes the scam. Many celeb-bait ads use plain doctored photos, and many deepfakes are not ads.
  • Clickbait is sensational framing designed to win clicks, usually for legitimate if low-quality content. Celeb-bait is fraud.
  • Public figure impersonation is the broader policy category, covering fake accounts, pages and messages as well as ads. Meta treats account impersonation and ad impersonation as separate enforcement checks.

Recent developments

The pressure has shifted from platforms' own tools to courts and regulators, with mixed results. On August 10, 2026, Judge Pitts found Meta liable for "gross negligence" in failing to preserve evidence in the Forrest case. On September 16, 2026, the Frankfurt Regional Court ruled that Meta faces fines of up to €250,000 per violation if fake Finanzfluss ads reappear, rejecting its Digital Services Act hosting defence. That ruling is not final. On November 10, 2026, by contrast, the Tokyo District Court dismissed Maezawa's suit against Meta over the fake investment ads, according to Kyodo and The Japan Times.

In the UK, Ofcom proposed a fraudulent advertising code on July 10, 2026, listing impersonation of public figures and deepfakes as risk signals, with fines of up to 10% of global revenue. Consultation closed on October 2, and final codes had not been published as of November 2026. Denmark's bill L 49, tabled October 8, would require publishers to prove consent for realistic digital imitations of a person from January 1, 2027. On October 22, Rhode Island Attorney General Peter Neronha sued Meta in Providence Superior Court over scam advertising, according to Reuters.

Disclosure is no defence. Google has clarified that AI labels won't save public figure impersonators from Google Ads suspension: an AI-generated likeness of a public figure used to deceive remains banned even when the ad carries an AI disclosure. Disclosure is also rare. Resemble AI's scan of 37,768 Meta Ad Library files, reported by PPC Land on October 27, found only two of 3,379 AI-flagged files carried a disclosure. Some 71.3% of files from "bait pages" were flagged as AI-generated, against 6.9% for brand pages, though celebrity content made up only 1.6% of the labelled fakes.

Timeline

  • April 2018 - Martin Lewis files a defamation claim against Facebook over more than 1,000 scam ads using his name.
  • January 23, 2019 - Lewis settles; Facebook pledges £3 million to Citizens Advice and a UK scam ad reporting button.
  • November 12, 2019 - Amsterdam District Court orders Facebook to act against fake bitcoin ads using John de Mol.
  • March 18, 2022 - ACCC sues Meta over crypto scam ads using Dick Smith, David Koch and Mike Baird.
  • March 2024 - Google's strengthened Misrepresentation policy allows immediate suspension for public-figure impersonation.
  • April 12, 2024 - Australian prosecutors drop Andrew Forrest's criminal case against Meta.
  • May 2024 - Yusaku Maezawa sues Meta in Tokyo over investment scam ads using his image.
  • June 2024 - US District Judge P. Casey Pitts denies Meta's Section 230 motion to dismiss Forrest's civil suit.
  • October 21, 2024 - Meta begins testing facial recognition against celeb-bait ads with about 50,000 public figures.
  • October 29, 2024 - Thirty plaintiffs in Japan sue Meta for about 435 million yen over fake celebrity investment ads.
  • March 2025 - Meta brings opt-in celeb-bait protection to the UK.
  • September 29, 2025 - Meta expands facial recognition to more public figures in the EU, UK and South Korea; roster near 500,000.
  • February 26, 2026 - Meta sues celeb-bait advertisers in Brazil and China; roster above 500,000.
  • July 10, 2026 - Ofcom publishes draft fraudulent advertising codes.
  • August 10, 2026 - Judge Pitts finds Meta grossly negligent in preserving evidence in the Forrest case.
  • September 16, 2026 - Frankfurt Regional Court holds Meta liable for fake Finanzfluss ads.
  • October 2, 2026 - Ofcom's consultation closes.
  • October 8, 2026 - Denmark tables bill L 49 on realistic digital imitations.
  • October 22, 2026 - Rhode Island sues Meta over scam advertising.
  • November 10, 2026 - Tokyo District Court dismisses Maezawa's suit against Meta.

Summary

Who. Fraud operators, from small affiliate sellers to organised groups in China, Brazil and Southeast Asia, who exploit public figures such as Martin Lewis, John de Mol, Andrew Forrest, Yusaku Maezawa and Thomas Kehl. Meta and Google enforce against it, while courts and regulators from the ACCC to Ofcom police the platforms.

What. Scam advertising that uses a famous person's image, voice or deepfaked likeness without consent to make a fraudulent offer look endorsed, typically leading to fake news pages and off-platform investment or health scams.

When. Fake celebrity endorsements drew litigation from 2018. Meta named the problem celeb-bait and began facial recognition tests on October 21, 2024, expanded them through 2025 and sued operators in February 2026. Court outcomes in 2026 have gone both ways.

Where. Mainly on Facebook and Instagram, and also on Google, YouTube and other self-serve ad platforms, with major cases in the UK, the Netherlands, Australia, Japan, Germany, Brazil and the US.

Why. Recognition borrows trust. Self-serve ad systems make that trust cheap to rent, and generative AI makes the likeness cheap to fake, so the tactic persists despite facial matching, verification, disclosure rules and lawsuits.