Yes. TikTok's Community Guidelines require that realistic content generated or heavily edited by AI be labeled; unlabeled content may be removed, have its reach limited, or get a label added by the platform. The Terms of Service also state that watermarks and content credential metadata must not be removed. When posting through official APIs, the Content Posting API has an is_aigc field and the Accounts API has an is_ai_generated field for declaring AI generation. When you're unsure whether something counts as realistic, labeling is safer than not.
What the official rules say
On November 20, 2026, we read the Integrity and Authenticity section of TikTok's Community Guidelines, the US Terms of Service, and the documentation for two official APIs. These are the parts directly related to AI labels:
| Source | Rule |
|---|---|
| Community Guidelines | Realistic content generated or heavily edited by AI must be labeled |
| Community Guidelines | Unlabeled content may be removed, have its reach limited, or get a label added by the platform |
| Community Guidelines, Terms of Service | Watermarks and content credential metadata must not be removed |
| Content Posting API (Direct Post) | Has an is_aigc field; when set to true, the video shows "Creator labeled as AI-generated" |
| Accounts API (publish video) | The publish parameters include is_ai_generated, used for AI-generated labeling |
| Symphony Creative Studio | TikTok's own AI video tool; videos it generates get an "AI-generated" label automatically |
Beyond labeling, the same guidelines say unoriginal content, or content that reuses material, isn't eligible for the For You feed. Labeling is about whether viewers know AI made it; originality is about whether it can be recommended. You need both. For the second, see Why AI videos all look the same.
What needs a label: judge by the visuals
The official key words are "realistic" and "heavily edited." The table sorts common cases according to the text we read; where the text doesn't cover a case, we say we couldn't find it. This table is our reading of the official text, not an official classification:
| Case | Label it? | Basis |
|---|---|---|
| AI-generated realistic people or lifelike scenes | Yes | Guidelines text: realistic AI-generated content |
| Real footage heavily edited with AI (face swaps, scene changes, altered movements) | Yes | Guidelines text: heavily edited content |
| AI visuals in an obviously cartoon or illustrated style | Not addressed case by case; we suggest labeling | No explicit text found |
| Real people on camera, with AI used only for the script or subtitle translation | Not addressed case by case | No explicit text found |
| Screen recordings with template subtitles and a few AI shots | If the AI shots are realistic, we suggest labeling | Judged by the realistic shots |
For teams promoting software or selling products, the most common AI footage is "generated use scenes": someone using the software in an office, a family using the product in a living room. These shots are usually realistic and need a label under the rules. Teams using AI for product promotion can read how software and brand teams use it and how cross-border sellers use it.
Watermarks and content credentials: leave them in
Many AI video tools leave a visible watermark in the finished video, or write metadata into the file describing where it came from. The Terms of Service state that watermarks and content credential metadata must not be removed. In practice, three steps are where they tend to get removed by accident:
- Re-editing and exporting. Some editing apps don't keep the original file's metadata by default on export, so check your export settings.
- Cropping out a corner watermark to make it look cleaner. That is removing the watermark. Don't do it.
- Compression and transcoding tools. Batch compression can wipe metadata along the way.
We couldn't open the official help article on whether the platform reads content credentials such as C2PA to detect AI content automatically, so this article draws no conclusion. Whether or not the platform can detect it, labeling it yourself is the safest approach.
Keep a record of where your material comes from, too. Which footage you shot yourself, which is licensed and which is AI-generated is best tracked separately in your asset library, so you can answer clearly if anyone asks later. TinyCoal stores assets and finished videos on your own computer. When cloud AI is needed to generate shots, the app tells you first, and every time content leaves your computer it's logged locally; see Local data and privacy.
How to label when posting through an API
If your team can build software and posts through official APIs, each of the two APIs has a field:
| API | Field | Notes |
|---|---|---|
| Content Posting API (developers.tiktok.com) | is_aigc | When set to true, the video shows "Creator labeled as AI-generated" |
| Accounts API (business-api.tiktok.com) | is_ai_generated | Used for AI-generated labeling when posting to owned accounts |
Both APIs have review requirements and use restrictions, so you can't just plug them in; details in TikTok Business Center vs. third-party tools. If you post manually through the website or app, tick the AI-generated content option the platform offers on the posting page.
Checks before posting
Building labels into your review process is more reliable than counting on people to remember. We suggest running through these steps, in order, before every video goes out:
- Mark where AI was used. On the review sheet, note which shots are AI-generated and which are real footage or screen recordings.
- Decide whether it's realistic. If there are realistic AI people or scenes, or real footage was heavily edited, it needs a label.
- Check watermarks and metadata. After export and compression, confirm the source information is still there and no visible watermark was cropped out.
- Tick the option or set the field when posting. On the web, tick AI-generated content; through an API, set the matching field.
- Spot-check after posting. Open the published video and see whether the label shows. If it doesn't, add it to your daily checklist and deal with it.
In TinyCoal, every video from AI video production must pass human review before it can enter the publishing schedule, and the AI label is a required item in that review. Once scheduled, each post still waits for a person to confirm by default, and posting happens inside each account's separate environment. Posting results are summarized in the daily checks and reports. For who reviews and who posts when several people work together, see How a team manages multiple TikTok accounts. The software only makes the check steps routine; responsibility for labeling stays with the person posting, and the requirement that users follow platform rules is written into our Terms of service. For TinyCoal's full flow from content to posting, see All features.
Check other platforms separately
This article verifies TikTok's rules only. If you post the same video on other platforms, check each platform's current rules separately; don't assume what works in one place applies everywhere. TinyCoal adds platforms one at a time, and the order is in the FAQ. When one video goes to accounts in several countries, every language version needs the label; for language and time zone setup, see Matching language and time zone across countries. Teams posting AI content for clients should agree with the client on who's responsible for labeling at the start of the engagement; see how agencies use it.
FAQ
Will an AI-generated label reduce my reach?
We found nothing on the official pages saying a label lowers recommendations. What we found points the other way: realistic AI content without a label may be removed, have its reach limited, or get a label added by the platform.
The footage is real people, and only the voice-over is AI. Do I need a label?
The guidelines text targets realistic content generated or heavily edited by AI. We found no specific statement on whether AI voice-over alone must be labeled. If the voice-over could make viewers think a specific real person is speaking, we suggest labeling.
If AI drafts my message replies, do I need to tell the customer?
This article covers labeling rules for video content. For messages, TinyCoal's Inbox only lets AI draft, and replies are sent after a person confirms; see How to reply to messages in several languages. For how data is handled, see our Privacy policy.
Sources
The official pages below were read on November 20, 2026. Platform rules change, so always check the latest version:
- TikTok Community Guidelines: Integrity and Authenticity
- TikTok US Terms of Service: Terms of Service (US)
- TikTok for Developers: Content Posting API reference: Direct Post
- TikTok API for Business: Publish a public video post to an owned account
- TikTok Ads Help Center: About Symphony Creative Studio
What we could not verify (and therefore did not write): whether TikTok uses content credentials such as C2PA to detect AI content automatically (we couldn't open the official help article); the specific standard for "realistic"; whether cartoon-style AI visuals, AI voice-over alone or AI-written scripts must be labeled; and whether labels affect recommendations. TikTok is a trademark of its owner. TinyCoal is not affiliated or partnered with TikTok.
