TikTok has started testing an opt-in tool that scans for AI-generated likenesses of creators, giving them a way to flag unauthorized deepfakes directly to the platform. The feature is currently limited to a select group of US creators, with identity verification handled by a company called Jumio.
How the tool works
Creators invited into the test will need to verify their identity before the scanning kicks in. That means a real-time selfie scan and a government ID check, both processed through Jumio’s verification infrastructure.
TikTok spokesperson Zachary Kizer told The Verge that TikTok does not retain ID documents, and facial information is not stored after the verification process.
Once verified, the tool monitors the platform for AI-generated content that mimics a creator’s likeness. If something suspicious surfaces, the creator can report it to TikTok for review and potential removal. Social media consultant Matt Navarra first spotted the feature being rolled out.
YouTube launched its own likeness detection system back in October 2025, initially available to YouTube Partner Program creators. By 2026, YouTube expanded the tool to include non-creator entities like journalists and civic leaders.
The deepfake problem by the numbers
As of July 10, 2026, TikTok reported having automatically labeled over 3 billion AI-generated videos. That number uses a combination of C2PA Content Credentials and TikTok’s proprietary watermarking and detection models.
The identity verification angle
Jumio, the company handling verification for TikTok’s new tool, already provides identity verification services across various platforms, including TikTok Shop.
The C2PA Content Credentials standard that TikTok uses for labeling AI content is backed by Adobe, Microsoft, and others, and addresses the same content provenance problem that several blockchain-based projects have been targeting.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.

1 hour ago
31









English (US) ·