Osegi
Mobile·2 min read

TikTok AI video checker: upload any file, but it has a catch

TikTok's new site lets anyone check whether a file came from ByteDance's AI tools. It misses Sora and Veo, though, and Likeness Detection goes global for creators.

A smartphone showing the TikTok logo next to a face scan outline, symbolising AI video and deepfake detection.
Image: Frandroid

TikTok now has a public AI video checker, and you don't need an account to use it. Upload an image, a video or an audio file, and the site tells you whether it was made or edited with TikTok's AI tools or those of its parent company ByteDance, including CapCut, Dreamina, Pippit and Hypic.

The timing makes sense. TikTok says it has labelled more than 5 billion AI-generated videos, up from 3 billion on 10 July. That is about 2 billion more in four months. But the label often vanishes once someone downloads a clip and reposts it elsewhere.

How does TikTok's AI checker work?

Since 2024, TikTok has added Content Credentials to content made with its tools. It is a standard backed by the C2PA coalition that stores, in the file's metadata, how the file was made. That data gets lost when a video is downloaded, edited and shared again. So the checker also looks for VeraDura, an invisible watermark developed by TikTok and hidden in the image, which is harder to erase.

The limits are 20 MB per image and 100 MB per video or audio file. TikTok says uploaded files are not stored.

Can it spot videos from Sora or Veo?

No, and TikTok admits it. The checker only detects signals from TikTok and ByteDance tools. A clip made with a rival generator such as OpenAI's Sora or Google's Veo falls outside its reach. A negative result therefore does not prove a video is real.

What is Likeness Detection?

The second tool warns creators when their face shows up in a deepfake. It is opt-in: the creator first verifies their identity with an outside provider, then TikTok flags videos or accounts that seem to use their face. The creator can review them and report them to moderators, who remove content that breaks the rules. TikTok says it does not keep ID documents.

TikTok has tested it since July with a small group of US creators. The global rollout starts now, but only for "eligible creators" and public figures, including government, politician and party accounts. TikTok has not published the eligibility criteria. YouTube got there first: its tool launched in late 2025 for monetised creators and has been open to all major creators since May 2026.

Apps worth a look

SUMO: the number puzzle

Tap the right tiles, add them up, hit the target. Easy to learn, hard to master. Free, for all ages.

Tags:tiktokai detectiondeepfakelikeness detectionbytedancecontent credentials

More from this day

See all →

More in: Mobile

See all →