A convincing image of Mia Goth appeared to show the actor wearing a pointed message about her ex, Shia LaBeouf. The post was fabricated, yet a verified Instagram account helped it gain attention, and coverage of the image carried the claim into search results. The episode reflects a wider problem: AI-made depictions of public figures can circulate widely before audiences understand what they are seeing.
A fabricated post gains credibility
The Instagram account @miagoth_ had more than a million followers and a blue checkmark. It shared an AI-generated image of Goth days after she filed for divorce from LaBeouf. The post added the song “Single Ladies (Put a Ring on It)” by Beyoncé and showed a T-shirt reading “MY EX IS MY BIGGEST FAN.”
The account was labeled as a fan page, but the verification mark and the timing gave the post a veneer of authority. Within days, it had nearly 500,000 likes, including from people in the entertainment industry. Goth’s representative, Christine Tripicchio, said the representative had previously complained to Instagram about “counterfeit, AI images” posted by the account.
That attention did not stop at Instagram. Gossip outlets picked up the post, and their coverage helped it appear in Google searches. An AI-generated overview described the image as a swipe at LaBeouf, while a prominent search result repeated the claim in its headline. Instagram eventually removed the account for impersonating a public figure.
Celebrity likenesses become social content
The fabricated Goth image is one example in a much larger stream of AI content featuring recognizable people. Some posts put celebrities into invented performances or situations; others depict them in sexualized ways. The article describes accounts using AI images of Scarlett Johansson, Millie Bobby Brown, Zendaya and Ariana Grande, including content that presents them partially undressed or in sexualized positions.
Some of this material is tied to efforts to make money. One profile focused on AI-generated images of Johansson and included contact information for advertisers, along with links to Patreon and Telegram channels selling subscriptions. The article also describes accounts devoted to narrow themes, such as “Stranger Things” actors getting married to each other.
Other videos follow repeatable formats. An AI-generated Taylor Swift video had 1.8 million views on Instagram; a similar clip featuring Shrek had 3.1 million. Videos depicting Steve Jobs and Lionel Messi in comparable scenes had 371,000 and 2.1 million views. A clip showing Formula 1 drivers Lewis Hamilton and Max Verstappen lip-syncing to “APT” by Bruno Mars and Rosé had 2.9 million views.
Platforms face a difficult enforcement task
Meta says its rules prohibit accounts from using someone’s likeness to pose as that person, whether the content is authentic or AI-generated. After being asked about eight accounts posting celebrity deepfakes, spokesperson Alex Dziedzan said Meta had removed some. The Goth and Johansson accounts appeared to be taken down.
That policy sits alongside a system in which users can purchase blue verification marks even if they are not prominent figures and their pages are not official. In Goth’s case, the checkmark did not establish that the page was authentic, but it appeared alongside a large follower count and a post that reached a substantial audience.
On YouTube, a likeness detection program scans for visual matches and, more recently, sound-alikes. People can request takedowns, but the company says it weighs public interest, newsworthiness and consent. The program, first available to public figures such as celebrities, creators, politicians and journalists, now covers every adult with a YouTube account. YouTube has said the number of takedown requests it receives is relatively small.
Detection and context matter
Other episodes cited in the article show how varied the disputes can be. An AI-generated Drake sound-alike song was removed from YouTube after copyright strikes. Scarlett Johansson hired a legal team after OpenAI released a ChatGPT voice that sounded like her without her consent. Taylor Swift publicly endorsed Kamala Harris for president in 2024 after Donald Trump posted fake images suggesting she endorsed him.
The examples raise practical questions for platforms and audiences. A blue checkmark, a high view count or a search result can make a post seem established, but none of those details alone confirms that an image or video is authentic. When fabricated content is repeated by other accounts or reported as gossip, its reach can outlast the original post’s context.
The article notes that, in the absence of national right of publicity laws, disputes over AI-generated depictions of real people are being addressed on platforms. It also describes YouTube’s takedown system and the US Take It Down Act, signed into law in 2025, which criminalizes distributing nonconsensual intimate imagery, including AI-generated content, and requires platforms to remove it within 48 hours. Those measures address parts of the problem, while the examples show how quickly fabricated likenesses can move through social feeds and search.