Does TikTok auto-label AI video? Only its own AI effects

On TikTok Shop's policy page, only one case is labeled for you: content made with official TikTok AI effects "is labeled automatically". A clip you generate with Sume is made outside the app, so it is not that case. If it has synthetic faces, voices or digital humans, or looks real, the page puts disclosure on you.

The page does not say whether TikTok also detects third-party AI content on its own, so do not plan around detection either way.

Who labels what

Where a Sume clip lands

Sume's video catalog exposes a generate_audio field per model, which shows whether the model can generate an audio track. If you pick a model that does, the voice in the clip is synthetic, and the policy names synthetic voices as a disclosure trigger. If you add an avatar presenter, the face is synthetic too.

Run through the three triggers for each clip and keep the answer next to the file name. Disclosure then becomes one of two actions at posting time: flip the toggle, or burn a note into the video.

• Synthetic face or digital human on screen: disclose.
• Generated voice: disclose.
• Realistic scene that a viewer could take for footage: disclose.
• Obvious cartoon with accurate product info: the page lists it as permitted.

A per-clip log

A short log removes the guesswork when an account is reviewed. For each uploaded clip, record the date, the source (a TikTok effect or an outside tool), the three trigger answers and the disclosure method used. If a clip mixes sources, for example a Sume-generated scene with a TikTok effect on top, the outside-generated parts still count under the three triggers.

The log also helps with the permitted list. Lighting, color and noise improvements, AI-assisted translation and dubbing, and copywriting support are named as permitted, so a clip that only uses those has a short entry.

What the label does not fix

Labeling is separate from the content rules. The same page prohibits AI that misleads about products, people or facts, and it lists penalties of removal or restricted visibility, warnings, permanent bans and a Creator Health Rating deduction. A correct label on a misleading clip does not make the clip allowed.