
The test
TikTok requires creators to disclose when a video is materially AI-generated or edited to look real. The rule traces to a newsroom post dated 19 September 2023, announcing a manual labeling tool and stating 'the policy requires people to label AI-generated content that contains realistic images, audio or video, in order to help viewers contextualize the video'. The current, more detailed version now lives in TikTok's Community Guidelines under Integrity and Authenticity, retrieved as it reads on 16 September 2026.
What the evidence says
The 2023 post is TikTok's own product announcement: it introduced the labeling tool and said the company was 'testing' automatic detection and renaming in-app AI effects to include 'AI' in their titles. The 2026 guidelines page defines terms the 2023 post did not: 'AI-Generated Content (AIGC)' is 'any image, video, or audio made or changed by AI', and 'Realistic-Appearing Scenes or People' means content that 'looks like it could be real'. The current page splits disclosure obligations by who appears in the content, defining 'Public Figures' as people 18 and older with 'a significant public role, such as a government official, politician, business leader, or celebrity', and stating TikTok does 'not identify people under 18 as public figures', treating all minors as private figures. Required disclosure applies when, for example, 'a face is replaced with someone else's' or 'AI-generated audio mimics the voice of a real person', but not for small edits like color correction or generic text-to-speech that does not mimic a known voice.
The sample and the variance
Both documents are TikTok's own policy statements, not an audit of how much AI content actually carries a label. The 2023 announcement describes a tool's launch date; the 2026 guidelines describe current scope and carry TikTok's own version tag for a 2025 second-half update, so the wording quoted may not match an earlier or later version. Neither document states a compliance rate or how the rule applies to AI content made elsewhere and reposted.
What to try next
As an editorial checklist, a creator altering a face, voice, or a scene's depicted reality should apply TikTok's AIGC label or an equivalent caption before posting, and check whether anyone shown is a minor, since the guidelines treat all under-18 individuals as private figures regardless of following. A publisher covering this policy should treat the 2023 launch date and the 2026 definitions as two layers of one record, not interchangeable citations.
- Does the content realistically depict a person or scene in a way a viewer could mistake for unaltered footage?
- Is anyone shown under 18, which under TikTok's current definitions removes them from the 'public figure' category regardless of fame?
- Has the labeling requirement or its examples changed since this record was retrieved on 16 September 2026?
Together, the two documents show a policy that began as a single disclosure tool in 2023 and has since grown into a more detailed rule with defined terms for what counts as realistic, whose likeness triggers stricter disclosure, and which edits fall outside the requirement altogether.
Sources & limits
- New labels for disclosing AI-generated content ↗
Announces the original AIGC labeling tool and states the policy requires labeling of realistic AI-generated images, audio or video.
Source · Source date: 2023-09-19 · Archive retrieval: 2026-09-16 - Community Guidelines: Integrity and Authenticity ↗
States the current, detailed labeling rule, including definitions of AIGC, realistic content, and the public/private figure distinction that governs required disclosure.
Source · Source publication date not stated · Archive retrieval: 2026-09-16