TikTok's current AI rules, built on its Content Credentials rollout and its synthetic media guidelines, come down to three things: how a clip gets an automatic AI label, when you have to disclose one yourself, and what happens if you do not.
The net effect for AI creators is simple: flagging yourself early is better than getting caught late. Here's what the policy actually says, in plain English, and what a compliant publishing routine looks like.
What triggers the auto label
TikTok applies an AI generated label automatically from a few provenance signals:
- Content Credentials (C2PA) in the file: TikTok reads this metadata and applies the AI generated label automatically.
- An invisible watermark: TikTok adds a watermark only it can read to content made with TikTok's own AI tools, and it reads uploads that already carry Content Credentials, so the label travels with the clip.
- Proactive detection models: TikTok also runs detection models to catch AI generated content that arrives with no provenance signal at all.
What you should self label
TikTok gives you one switch, the AI generated content toggle. Turn it on for anything completely generated or significantly edited by AI. TikTok requires it for realistic images, audio, or video, and a label reading Creator labeled as AI generated then appears on the clip.
Self labeling AI generated content as AI generated has no ranking penalty. Turning the toggle on up front is also what keeps you out of the enforcement path below, which is reserved for realistic AI that was published without a label.
What actually downranks
Two things, and neither of them is "being AI":
- The auto label firing on a clip you didn't self label. TikTok reads this as an attempt to pass synthetic content off as real, and the clip can be removed or made ineligible for the For You feed. Disclose up front and this never applies to you.
- Reports from users. A high report rate relative to views triggers a manual review queue. AI clips that look uncanny get reported more often, labeled or not.
Topic categories with stricter rules
TikTok holds realistic AI to a higher bar in a few categories:
- Public figures. Any AI clip showing a recognizable public person (politician, celebrity, executive) needs the AI generated label even if the content is positive.
- News and events. AI clips that depict real world events (war footage, natural disasters, crime scenes) need the label even if the events themselves are real and the AI is purely illustrative.
- Medical, legal, financial claims. These are sensitive everywhere. As a ViralGen best practice, label AI clips that make claims in these areas and add a standard "not professional advice" disclaimer.
This is also where likeness rights matter most. Publish with a creator you have the rights to use (yourself or an AI persona), never a real person who hasn't signed off.
Account level consequences
TikTok runs an account wide strike system rather than an AI only one. A violation earns a strike, strikes expire after 90 days, and TikTok does not publish the exact thresholds. Enough strikes across policies can lead to a permanent ban, and individual clips can be removed or made ineligible for the For You feed, with a right to appeal. Disclosed AI content does not enter this system, which is the whole point of labeling up front.
A note on Instagram and Meta
Meta's parallel policy is looser. They label with an AI info tag and generally leave distribution alone, though misleading or fact checked AI media can still be demoted.
The right routine is identical on both platforms: disclose up front, leave provenance signals intact, and compete on quality. If the tag bothers you, the answer is better output, not hidden provenance. Tampering with the signals a platform reads is a terms violation that trades a cosmetic tag for real account risk.