TikTok requires you to label AI-generated or significantly edited content that shows realistic-looking people or scenes, either by turning on the AI-generated content setting before you post or with your own clear caption, sticker or watermark. Small edits such as colour correction or cropping, obviously artistic styles like anime, and generic text-to-speech narration do not need a label. Some AI content is banned even when labelled: TikTok's AI-generated content help page rules out the likeness of adult private figures used without their permission and the likeness of anyone under 18.
What TikTok's updated guidelines say
The rule sits in the Edited Media and AI-Generated Content section of TikTok's Integrity and Authenticity guidelines, in the version released on 25 August 2026 and effective from 24 September 2026. It requires creators to label AI-generated or significantly edited content that shows realistic-looking scenes or people, and says unlabelled content may be removed, restricted or labelled by TikTok's team depending on the harm it could cause. It also says that some edited or AI content stays harmful even with a label, so content that is misleading about matters of public importance or harmful to individuals is not allowed at all.
The expandable definitions under that section do most of the work:
- AI-generated content is any image, video or audio made or changed by AI, in realistic or artistic styles.
- Significantly edited content makes it seem someone did or said something they did not, or changes their appearance until they are unrecognisable. TikTok lists cutting phrases to change meaning, rearranging or combining clips, changing speed and adding or removing audio or video as examples, so the category reaches edits made without AI.
- Likeness means a recognisable image, video or audio representation of a person, including face, body, voice and gestures.
- Private figures are all minors plus every adult who is not a public figure, and TikTok never treats a minor as a public figure.
The help page frames the duty in two tiers: TikTok encourages labels on anything completely generated or significantly edited by AI, and requires them on AI content with realistic images, audio or video. Labelling every substantial AI edit keeps you inside both readings.
Label decision table
| Edit or content | Label? | Notes from TikTok's pages |
|---|---|---|
| A face replaced with someone else's | Required | You also need that person's permission if they are a private figure |
| AI that makes someone appear to say words they never said | Required | This applies to your own face and voice as well as other people's |
| A background, object or person added or removed in a misleading way | Required | The misleading effect is what triggers the duty |
| AI audio that mimics a real person's voice | Required | A recognisable voice counts as part of someone's likeness |
| Realistic clips cut, sped up or rearranged so a statement changes meaning | Required | Counts as significantly edited content even without AI |
| Colour correction, reframing or cropping | Not needed | Listed as small edits that need no disclosure |
| Anime, cartoon or painterly AI styles | Not needed | Artistic styles fall outside the realistic-content duty |
| Generic text-to-speech narration | Not needed | Unless the voice is recognisably a known individual |
| Video made only with TikTok's own AI effects | Usually labelled for you | If you also used an outside AI tool, follow the labelling rule yourself |
| An upload carrying Content Credentials that show AI | May be labelled automatically | An automatic label cannot be removed from the post |
| Real footage with no AI or misleading edits | Do not label | Misleadingly labelling unaltered content breaches the Terms of Service |
A worked example shows how the rows combine. Say you film a talking-head video in English, then use an AI dubbing tool to produce a Spanish version with your cloned voice and adjusted lip movements. The result shows a real person, you, saying words you never actually said, in a realistic voice. That meets TikTok's description of significantly edited content, so the Spanish version needs the label even though the original did not. A caption line such as “Spanish version dubbed with AI” alongside the setting makes the disclosure clear to viewers who never saw the original.
Banned even with a label
The guidelines list AI and edited content that is not allowed whatever the disclosure. In plain terms, TikTok does not allow:
- using the likeness of private figures without their consent;
- sexualised, fetishised or victimising depictions;
- AI-created likenesses made to bully or harass someone;
- accounts centred on AI images of young people styled in adult clothing, poses or expressions;
- AI or significantly edited content that misleads on matters of public importance, such as material made to look like a real news source, invented crisis events, public figures degraded or linked to crimes, and public figures shown backing products, taking political stances or making endorsements they never made;
- anything else that breaks the Community Guidelines, including the rules on impersonation, misinformation and hate speech.
Two further notes matter for creators. Realistic content that has not yet been confirmed as AI-generated, but presents matters of public importance in a misleading way or could harm private figures, is ineligible for the For You feed. And humour or art such as a spoof, meme or dance trend is allowed, which does not remove the label duty if the spoof looks real. The practical line for a business account: never generate a public figure appearing to promote your page, product or offer.
How to turn on the AI-generated content setting
The help page gives separate steps for photos and videos in the TikTok app. For a photo:
- Tap the Add post button and capture or choose your photo.
- Open the post settings at the bottom of the screen.
- Turn on the AI-generated content setting.
For a video:
- Tap the Add post button, then record or upload your video and continue.
- Tap Next.
- Open More options and turn on the AI-generated content setting.
TikTok says the setting does not affect a video's distribution as long as the content follows the Community Guidelines. The wording viewers see is not consistent across TikTok's own pages: the help article says the label reads “creator labeled as AI-generated”, while the TikTok Shop US policy page shows “Disclosed by creator as AI-generated”. Either way, it tells viewers the disclosure came from you.
When TikTok applies the AI label itself
TikTok may apply its own AI-generated label to content it identifies as completely generated or significantly edited with AI. The help page names two triggers: using TikTok's AI effects, and uploading content that carries Content Credentials, the C2PA metadata standard that lets platforms recognise AI-generated files. Once an automatic label is on a post, you cannot remove it. If your editing app attaches credentials for an AI step, expect the label and plan your caption around it. Do not strip the metadata to avoid the label; where the content is AI-made, that leaves you on the wrong side of the labelling rule.
Pre-post checklist
- Would a viewer take any part of this video for real footage or a real recording? Turn the setting on.
- Did you use an outside AI tool on top of TikTok effects? Label it yourself rather than relying on the automatic label.
- Whose face or voice is in it? A private adult needs to have given permission, and a public figure must not appear to endorse or say anything they have not.
- Could any AI-generated or altered likeness be read as a minor? Do not post it.
- Is a product being promoted? Check the TikTok Shop rules below and switch on the paid partnership setting where it applies.
- Does the caption call the footage real, raw or unedited? Rewrite it if AI shaped what viewers see.
- Have you saved the project file and source clips in case you need to explain an edit?
TikTok Shop adds stricter rules for product videos
Creators selling through TikTok Shop in the United States have an extra policy, AI-Generated Content Restrictions and Requirements, published in Seller University. It requires disclosure of content fully or significantly created or altered by AI, bans using AI to imitate someone's identity, likeness or voice for false endorsements, and bans AI that changes how a product looks or performs, from size, colour and material to quantity, bundle contents, compatibility, safety attributes and results. It does permit AI for visual styles, quality improvements, translation and subtitles, script support and accurate product demonstrations. Violations can lead to removal, restricted visibility, warnings, restrictions or permanent bans for severe cases. Our guide to TikTok Shop creator content guidelines covers the wider Shop rules, and the TikTok paid partnership guide covers ad disclosure, which the AI label does not replace.
If a video is removed or restricted
Start by checking whether the issue was a missing label or a prohibited use, because only the first is fixed by labelling. If you believe the decision is wrong, use the in-app appeal; our guide to appealing a TikTok ban or restriction walks through it. Creators who post professionally to EU audiences can have a legal disclosure duty on top of TikTok's rule, explained in the EU AI Act guide for creators, and anyone cross-posting should compare Meta's system in our Instagram AI label guide.
Limitations of this guide
This summary is based on TikTok's global English Community Guidelines, its AI-generated content help page and the TikTok Shop US policy as read on 1 October 2026. The help page and the guidelines word the requirement slightly differently, rules can vary by region, and TikTok's US operation has its own terms, so check the version shown in your app. Enforcement decisions rest with TikTok, and labelling does not make any content eligible for recommendation. For the broader legal picture behind AI content, including likeness consent and advertising rules, see the AI creator compliance checklist.