AI Deepfakes of Celebrities Are Spreading on Instagram With Little to Stop Them
A fake Mia Goth image posted by a fan account with 1M+ followers and a paid blue check drew nearly 500,000 likes before Instagram removed the account. It also spread to gossip sites and fueled a false Google AI overview.
AI-generated images and videos of celebrities are proliferating across Instagram and other social media platforms in 2026, with fake content routinely going viral while platforms struggle to enforce their own policies against it.
The problem came into sharp focus when an AI-generated image of actress Mia Goth appeared on Instagram shortly after she filed for divorce from actor Shia LaBeouf. The fake image showed Goth wearing a shirt reading “MY EX IS MY BIGGEST FAN,” set to Beyoncé’s “Single Ladies.” The post, published by a fan account with over one million followers and a paid blue verification checkmark, collected nearly 500,000 likes — including from entertainment industry peers — before Instagram removed the account for impersonating a public figure.
The image spread beyond Instagram into celebrity gossip outlets and triggered an AI-generated Google search overview that described Goth as taking “a clear swipe at her ex.” Goth’s representative Christine Tripicchio called it “incomprehensible that Instagram evidently prioritizes the dissemination of forged and manipulated images over the protection and integrity of a public figure’s likeness.”
The Goth incident is far from isolated. Instagram hosts AI videos of Taylor Swift rapping Drake lyrics, AI depictions of Steve Jobs and Lionel Messi set to popular songs, and accounts that post sexualized AI images of actresses including Scarlett Johansson, Millie Bobby Brown, and Zendaya. After The Verge flagged eight such accounts to Meta, the company removed some of them, stating it does not permit accounts to use someone’s likeness to impersonate them, whether the content is authentic or AI-generated.
A key gap in Meta’s rules makes the problem harder to address: still images are exempt from the platform’s AI labeling requirement, meaning the fake Goth image would not have required any disclosure under existing policy. Videos of real people and AI-generated audio do require labels.
YouTube has developed a more robust response, offering a likeness detection program that scans for visual and audio matches and allows individuals to request takedowns. In the first half of 2026, YouTube removed over 95 percent of videos reported through the program. The US Take It Down Act, signed into law in 2025, criminalizes nonconsensual intimate AI imagery and requires platforms to remove it within 48 hours, but no broad federal right of publicity law currently exists.