TikTok announced two tools on November 20, 2026, to identify content made with artificial intelligence and help protect creators' likenesses. The first checks provenance signals in images and videos generated or edited with the platform's AI tools and can be used without a TikTok account. The second, Likeness Detection, is beginning a global rollout to eligible creators who choose to enrol.

The features have different purposes and access requirements. The checker examines a file uploaded by a user. Likeness Detection searches TikTok for possible appearances of an enrolled person's likeness in content labelled or identified as AI-generated. The person can review matches and submit reports to the platform's teams. Detecting a match and removing a post are separate stages.

A phone is being secured to a tripod to record a video
George Milton

A checker for signals added by TikTok

The new checker looks for two types of information: C2PA Content Credentials metadata and the invisible VeraDura watermark. Both are used to identify material generated or edited with TikTok's AI tools. Content Credentials carry information about how content was made, while VeraDura is embedded in the material itself. TikTok describes its watermark as more durable and harder to remove.

The tool covers Content Credentials added by TikTok. A positive result indicates that the file was likely generated or edited with the company's AI tools. When the signal remains intact, the result can also identify the tool used. If the material was subsequently modified on other platforms, the checker does not identify those changes. Its information concerns the signals it can read from the file being examined.

A missing signal can have several explanations: the material was produced with another platform's tools, metadata was removed, the watermark was degraded, or the file predates the introduction of these signals. A “no signal” result does not certify that material is authentic. Checking TikTok provenance and establishing the complete production history of a video are operations with different coverage.

According to the transparency page, one file can be uploaded at a time. Accepted image formats are JPEG, PNG, JPG and WebP, with a 20 MB limit. Accepted video formats are MP4, MOV, WebM and AVI, up to 100 MB. Access without an account is confirmed in both the announcement and the tool's description; users do not need to be creators enrolled in Likeness Detection to perform this check.

The two features announced on November 20
FeatureWhat it examinesAccess and result
Provenance checkingTikTok signals in an uploaded fileNo account required; identifies supported signals
Likeness DetectionPossible appearances of an enrolled person in AI content on TikTokGradual rollout, eligibility and enrolment; supports reviewing and reporting matches

Likeness Detection requires voluntary enrolment and identity verification

TikTok states that the new likeness feature is available to certain verified accounts and creators aged at least 18. Accounts meeting the requirements see the option in the app's settings. The announcement also includes eligible public figures, including government, politician and political party accounts. The global launch describes the feature's rollout; actual access depends on account eligibility.

In its Help Centre, TikTok places the feature within TikTok Studio, in the additional tools area. Enrolment includes consent to biometric data processing and identity confirmation through the provider Jumio. A government-issued document and a facial scan are requested. TikTok says it receives a photo from which it creates a mathematical representation of the face, used to search for matches.

Scanning starts after enrolment and does not examine content posted before activation. Matches are sought only in material labelled or detected as AI-generated. The feature identifies the likenesses of people who have enrolled; it is not a service for identifying anyone who appears on the platform. These limits are explicit in the help documentation available when the rollout was announced.

From enrolment to a report review
  1. Enrolment: eligibility, consent and identity confirmation.
  2. Detection: searching for matches in AI content posted after activation.
  3. Review: the creator examines the indicated posts or accounts.
  4. Reporting: TikTok's teams assess reported content against the platform's rules.

A match is presented to the creator for review

After activation, users can see posts or accounts within the tool that might be using their likeness. They open the results and decide whether to submit a report about misuse of their image. According to TikTok, material that violates its rules on harmful, misleading or nonconsensual depictions of private individuals is removed following the review process.

Reporting remains separate from automated detection. An alert indicates a possible match, and the creator's review precedes a report to the safety teams. TikTok also retains its privacy reporting form, through which nonconsensual AI depictions can be reported, including by people who do not use the platform. The new tool adds a way for enrolled people to find and examine material.

Hands holding a phone with the TikTok app open
cottonbro studio

Checked files and enrolment data are handled differently

For the provenance checker, TikTok states that a file is processed solely to search for supported signals, then deleted immediately after the result is displayed. The company says uploaded material is not used to train models. This description applies to checking the file and does not replace the separate conditions for enrolling in the likeness detection feature.

In Likeness Detection, the facial representation is retained while enrolment remains active. TikTok says it does not keep identity documents, while Jumio may use them only for verification and may not retain them longer than necessary for that process. After the tool is disabled or the account is closed, the facial representation and related data are deleted within 30 days, according to the Help Centre.

The features concern two separate activities encountered in social media promotion: checking visual material and tracking use of a person's likeness. In TikTok's description, visible labels for realistic AI content, Content Credentials and invisible watermarking are complementary transparency layers. The file checker is available, while likeness detection is rolling out gradually, with enrolment and eligibility requirements.