Slop TVNewsLatest
Explainers

TikTok launches an AI content checker that needs no account

The Transparency Center tool reads C2PA credentials and TikTok's VeraDura watermark, while Likeness Detection rolls out to verified creators and public figures.

Illustration: TikTok launches an AI content checker that needs no account
Illustration: SLOP TV News. Logos are trademarks of their owners.

Key takeaways

  • TikTok published a public AI content checker on November 19, 2026 that reads C2PA Content Credentials and its own VeraDura invisible watermark without requiring an account.
  • TikTok's Transparency Center page lists one-file uploads of images up to 20 MB (JPEG, PNG, JPG, WebP) and videos up to 100 MB (MP4, MOV, WebM, AVI).
  • A clip made with another company's AI model can come back unmarked, because the checker reads only signals added by TikTok and its parent company ByteDance.
  • TikTok says it has labeled more than 5 billion AI-generated videos to date, and that it removes more than 99% of rule-breaking content before anyone reports it.

TikTok now runs a public tool that checks whether a file was made or edited with the company's own AI features, and it has started handing verified public figures a way to find videos that may use their likeness without consent. TikTok announced both at its 2026 Safety and Innovation Forum in London, in a newsroom post published on November 19, 2026.

The checker is the broader of the two. It sits on TikTok's Transparency Center, needs no account, and exists because the labels TikTok puts on AI content tend to fall off the moment a clip is downloaded and reposted somewhere else. Likeness Detection is narrower and aimed at the people most often targeted by deepfakes: politicians, government accounts and creators with a large following. Here is how each one works, and where the checker stops.

What TikTok launched

TikTok's post describes two additions.

The first is a public AI content checker, filed under the Transparency Center page Identifying content made with TikTok AI. Anyone can upload one file at a time and get a result without a TikTok account. TikTok's page says the tool covers images, videos and audio, though the format list it publishes covers images and videos only.

The second is Likeness Detection, which TikTok says it is starting to roll out globally. It is an opt-in tool for eligible creators and public figures, including accounts the platform classifies as Government, Politician, or Political Party Accounts, or GPPPAs.

How the AI content checker works

TikTok says its creation tools build three layers of AI transparency into content: visible labels, C2PA Content Credentials and VeraDura, which TikTok calls its proprietary invisible watermark. The public checker reads the second and third.

Content Credentials are an industry metadata standard that travels inside a file and describes how it was made. TikTok adopted them in 2024 and says it was the first video platform to do so. Its checker reads only the Content Credentials that TikTok itself added; for credentials from other companies, the page points users to the Content Authenticity Initiative's verifier at verify.contentauthenticity.org (https://verify.contentauthenticity.org/).

VeraDura is the fallback for when metadata is gone. TikTok says the watermark is built directly into the content, which makes it more durable and harder to remove. When a signal survives intact, the tool also reports which specific TikTok tool made the file.

The upload limits come from TikTok's own page: 20 MB for an image (JPEG, PNG, JPG or WebP) and 100 MB for a video (MP4, MOV, WebM or AVI). Euronews reports that the same 100 MB limit applies to audio, and that the checker covers parent company ByteDance's CapCut, Dreamina, Hypic and Pippit apps, not only TikTok. TikTok says it deletes an uploaded file as soon as it shows the result and never uses the file to train its models.

What the checker cannot catch

A clean result does not prove a file is real.

The checker reads TikTok's signals and nothing else. A clip made with another company's model can come back unmarked because there is nothing for the tool to read, and TikTok's own page says a missing signal could mean any of three things: the file was not made with TikTok's AI tools, it was made with another platform's AI tools, or it was not AI-generated at all.

TikTok also lists the ways its own signals can fail. The C2PA metadata can be stripped from a file, the invisible watermark can degrade, and a file can predate the signals entirely. The page warns that "no AI detection tool is 100% foolproof," and that the checker "cannot evaluate the origin, nature, or ownership of content beyond those signals." The London Daily put the consequence bluntly: a negative result does not establish that a file is genuine.

How Likeness Detection works

For creators who qualify, TikTok describes three steps.

Eligible creators sign up by verifying their ID with a third-party partner. TikTok then notifies them when content and accounts on its platform may be using their likeness, and they can review those matches inside the tool. From there they can report content straight to TikTok's safety teams. TikTok says it removes content that violates its community guidelines against harmful, misleading or nonconsensual depiction of private individuals, and that it does not keep ID documents.

RTÉ reports that the tool lets politicians and accounts with a large number of followers find and report AI-generated content that uses their image without consent, and that it is being rolled out gradually. The feature builds on a reporting route that already exists: anyone, whether or not they use TikTok, can flag nonconsensual AI-generated content through TikTok's privacy reporting form, and TikTok says it removes more than 99% of content that breaks its rules before anyone reports it.

What TikTok does not say

The post leaves the operational details out.

It does not say what makes a creator eligible, beyond naming verified creators, public figures and government accounts. It gives no timeline for when the rollout reaches every eligible account. It publishes no accuracy figure for the likeness matching, and it does not say how long a report takes to review once it reaches TikTok's safety teams. Ground.news, aggregating the day's coverage, notes only that the feature is being rolled out gradually.

What TikTok does state are its operating figures, and they are the company's own: more than 5 billion AI-generated videos labeled to date, the 99% proactive-removal rate, and more than 4 million people helped by the AI literacy hub it launched in July. None of those numbers has been checked by an outside party.

What it means if you make AI video yourself

If you make AI video, the checker shifts how exports behave. Anything you create with TikTok's AI features or ByteDance apps such as CapCut and Dreamina now carries signals that anyone, on or off TikTok, can read, including which tool produced it. Stripping those signals is possible, but TikTok built VeraDura to survive editing and resharing, so the marks are meant to outlast the first copy.

That cuts both ways for people who sell AI video. A provenance trail the file carries is something a client can check, which is useful when you are the one who made the clip. It is a smaller comfort when the clip was not made by any tool in TikTok's family, because then the checker has nothing to report and the answer looks the same as a genuine file.

To test a file yourself, open TikTok's Transparency Center page "Identifying content made with TikTok AI" and upload one image or video. Eligible creators and public figures can join Likeness Detection from the app, after the third-party ID check.

Sources

  1. newsroom.tiktok.com
  2. tiktok.com
  3. newsroom.tiktok.com
  4. euronews.com
  5. rte.ie
  6. montegre.com
  7. ground.news
  8. londondaily.com