TikTok is testing a new tool that scans for AI-generated videos and images that copy a creator's face, the company confirmed on July 17, 2026. TikTok US spokesperson Zachary Kizer confirmed the test to The Verge. The tool was first spotted by social media consultant Matt Navarra.

The test is initially limited to a small group of US creators. To use the tool, a creator must first prove who they are through a company called Jumio, which handles identity verification. The process involves taking a live selfie and showing an ID document. According to Kizer, TikTok does not keep the ID documents after verification. The facial information collected is used only to match the creator's face and to find potential unauthorized uses of their likeness.

Once verification is done, TikTok's system searches for AI-generated content that may be using the creator's face. Creators can review what the system finds and report unauthorized posts and accounts directly from those results. The tool is opt-in, meaning creators must actively choose to enroll rather than being enrolled automatically.

YouTube has a similar feature: the platform recently made its own AI likeness detection tool available to all adult users. The fact that both platforms are building these tools points to a shared problem that is now being solved at the platform level, rather than leaving individual creators to chase takedowns or take legal action on their own.

The way TikTok has designed this tool is worth a closer look. By requiring identity verification before any scanning happens, TikTok is making sure each enrolled creator is a confirmed real person. The facial scan taken during verification becomes the reference point that the system uses to find fake content. This approach is different from relying on watermarks or hidden labels in a file, because those can be removed or faked. Tying detection to a verified real person makes it harder to trick. The trade-off is that the system only works for creators willing to share their ID and facial data, which not everyone will want to do.

TikTok's stated rules around data handling address some of the obvious worries. No keeping of ID documents. Facial data used only for likeness matching. These are meaningful limits if they hold up in practice. The verification step does, however, send biometric data to a third party (Jumio), whose own data retention policies are outside TikTok's direct control.

The opt-in design also matters. A system that automatically scanned every creator's face against AI-generated content would raise very different consent questions. By making enrollment voluntary, TikTok is putting the decision about sharing facial data in the creator's hands. That is a reasonable approach given that laws around facial recognition technology in the US vary significantly from state to state.

For creators who do enroll, the steps are simple: verify your identity, let the system scan, review what it finds, and report anything unauthorized. What we do not yet know is how well the detection actually works. AI-generated likenesses range from clumsy face-swaps to highly realistic fake videos that are hard to tell from the real thing. The accuracy of TikTok's matching system will determine whether the tool is genuinely useful or just a convenient way to file reports after finding problems manually.

This is a limited test, and TikTok has not said when or if it will roll out more broadly. What it signals is that platforms are starting to build defenses against fake AI media directly into their products, rather than relying only on after-the-fact takedown requests. Creators are being asked to share some biometric data in exchange for the ability to find and flag unauthorized uses of their own faces.