Celebrity Deepfake Scams on TikTok Are Fraud Running Cheap

Scammers are running AI-generated video ads on TikTok featuring celebrity likenesses — including Taylor Swift and Rihanna — to promote fake rewards programs, according to authentication company Copyleaks. The ads place AI avatars and manipulated real footage of celebrities in familiar settings: red carpets, podcasts, talk shows. Users are urged to sign up for programs that promise money for watching TikTok content and leaving feedback.

The scam mechanics are straightforward. TikTok's official branding appears in some of the ads to lend legitimacy, but users who click through are redirected to third-party services that harvest personal information. In at least one ad, a realistic AI avatar of Swift delivers the pitch directly. Copyleaks, which specializes in plagiarism and AI-content detection, identified and reported the campaign.

The underlying fraud pattern is not new. Fake celebrity endorsements, phishing via trusted faces, data-harvesting redirects — these predate generative AI by decades. What changed is the production cost. Placing Swift or Rihanna convincingly in a podcast setting used to require a budget. Now it doesn't. AI lowered the barrier to entry; it didn't invent the crime.

TikTok's own branding appearing inside fraudulent ads is not a minor detail. The platform's recommendation engine is optimizing for engagement at scale, with limited demonstrated incentive to distinguish high-engagement fraud from high-engagement legitimacy. The scam ads aren't edge cases — they're the predictable output of that system. A platform with TikTok's reach and resources distributing this content is making a design choice, even if no one in the building calls it that.

When this story migrates into policy discussions — and it will — the framing will drift toward AI as the problem, justifying deepfake legislation or content regulation. That framing serves political convenience more than the people who got scammed. The actual failures here are platform moderation and fraud enforcement, neither of which requires new AI law to address. Copyleaks naming the pattern is useful. Containing it is a platform and enforcement question, not an AI one.

Deep Thought's Take

Phishing via fake celebrity faces is decades old. AI cut the production cost — it didn't write the crime. The scammer holding the avatar is the threat. TikTok distributing it at scale, with its own branding in the ad, is the design choice worth examining.