TikTok's AI disclosure policy is not a vibe check — it is a specific rule with a specific label. Any video, image, or audio made or edited with AI that shows a realistic-looking person, scene, or voice has to be disclosed, and TikTok gives you one switch to do it: the "AI-generated content" toggle in the post composer. Turn it on and the platform stamps the clip with a visible label under your handle; leave it off on content that needed it and TikTok can add the label itself, cut the post's reach, or remove it. The policy has two tracks that people constantly conflate. Organic posts run on the Community Guidelines and the creator-facing toggle. Paid ads run on TikTok's advertising policies and a separate control inside Ads Manager — "This ad contains AI-generated content" — with its own enforcement, its own lock-on-submit behavior, and a Spark Ads wrinkle where the label has to live on the original organic post before you boost it. On top of both sits auto-labeling: since May 2024 TikTok reads C2PA Content Credentials and tags AI files made elsewhere the moment they're uploaded, and once an auto label lands you can't take it off. This guide walks the whole policy — the label, the realistic-content line, the organic and ad tracks, the auto-label layer, and the penalties — so you know exactly what trips the rule and what doesn't before you ship AI content at any volume.
Strip away the think-pieces and the policy is narrow and concrete: if you post AI-generated or AI-edited content that is realistic — content a viewer could reasonably mistake for a real person, place, or event — you have to tell them it is AI. TikTok does not ask you to flag every file an algorithm touched. It asks you to flag the realistic ones, because those are the ones that can mislead. That single distinction — realism, not AI-involvement — is the thing most creators get wrong, and it is what the rest of this guide turns on.
The policy lives in two documents that enforce it differently. Organic posts are covered by TikTok's Community Guidelines on synthetic and edited media; paid ads are covered by TikTok's advertising policies. They share the same underlying obligation but give you different controls, different review, and different penalties, and conflating them is how compliant creators still get caught. We'll take the organic track first, because it is where most content lives, then the ad track, then the auto-labeling layer that sits over both.
For organic content, disclosure is a single switch. In the post composer — after you tap Next on a finished video or slideshow — open More options and turn on the control labeled "AI-generated content." When you publish, TikTok attaches a visible label to the post, shown near your username and caption, that reads along the lines of "Creator labeled as AI-generated." Viewers see it the way they see any other post metadata; it is not buried in a submenu. TikTok also accepts a clear in-content disclosure — a caption line, sticker, or on-screen watermark stating the content is AI — but the toggle is the native, machine-readable path and the one that keeps you cleanly inside the policy.
The label is additive and one-directional in an important way: you can always add it, but once it is on an automatically labeled post you cannot take it off. There is no penalty for over-labeling a borderline piece, which matters for the gray-area edits we cover below — when you are unsure whether something crosses the realism line, the asymmetry of the rule rewards labeling it.
Label it when the content is AI-made or AI-altered and realistic. TikTok's own framing covers: synthetic or significantly edited video, images, or audio depicting a realistic-looking person or scene; AI voice cloning of a real person; AI changes to what someone actually said in a clip; face swaps; and adding, removing, or altering people, objects, or backgrounds in a way that could mislead a viewer. Significantly edited media that makes someone appear to say or do something they didn't — or that makes a real person unrecognizable — falls under the same rule whether or not AI did the editing.
The exemptions are just as specific, and they are where creators over-correct and label things that don't need it. You do not need the AI label for color correction, cropping, or reframing; for everyday filters such as beauty and makeup effects; for generic text-to-speech voiceovers that aren't cloning a specific real person; or for AI-written captions and scripts. The logic is consistent: these don't fabricate a realistic depiction of something that didn't happen. A fully AI-generated spokesperson needs the label. An AI tool that wrote your caption does not.
Two things are off-limits even with the label on. AI content that could seriously mislead people about matters of public importance, and AI content that harms a real person, are prohibited outright — the toggle is not a permission slip for them. Disclosure handles the "is this real?" problem; it does nothing for the "should this exist?" problem, and TikTok treats those as separate questions.
Self-disclosure is only half the system. TikTok labels a lot of AI content without asking you, and understanding when it does keeps you from assuming an unlabeled upload is safe. First, anything made with TikTok's in-app AI effects is labeled automatically. Second, and more consequentially, TikTok reads Content Credentials — the C2PA metadata standard that records how a file was made and travels with it. On May 9, 2024, TikTok became the first video-sharing platform to implement Content Credentials, and it now applies an "AI-generated" label the moment you upload a file that carries them, even if that file was created in a completely different tool.
Because those credentials persist through downloading and re-uploading, you cannot launder an AI file's provenance by moving it between apps — if the metadata is attached, TikTok sees it on arrival. And the auto label, once applied, is locked: the creator can't remove it. The practical takeaway is that self-disclosure and auto-labeling aren't alternatives. The auto layer is the floor; your toggle is what covers the realistic AI content that happens not to carry readable credentials. For the broader context on how much of this TikTok is now doing — it has labeled billions of clips this way — see TikTok AI labeling at scale, and for how labeling works across every major platform, AI content labeling in 2026.
Paid content does not use the organic toggle. Inside TikTok Ads Manager, the AI disclosure control is a distinct switch worded "This ad contains AI-generated content," set during ad creation, and governed by TikTok's advertising policies rather than the Community Guidelines: an ad whose image, video, or audio is fully AI-generated or significantly modified by AI has to carry the label before it runs.
The ad control has behaviors the organic toggle doesn't, and each one is a live trap for performance teams. Once you submit an ad, the toggle locks — the only way to turn it off afterward is to swap out the video or image. Duplicating a campaign resets the toggle to off, so a cloned ad that was previously labeled can ship unlabeled unless you re-enable it. And Spark Ads — which boost an existing organic post rather than uploading fresh creative — follow the organic disclosure process, which means the AI label has to be on the underlying organic post before you authorize it for promotion; a label missed on that original post follows the ad into paid distribution. The deeper advertiser playbook for synthetic ad creative is in AI UGC ads for TikTok, and the broader question of what AI can and can't do for TikTok ad creative is in AI TikTok ads.
The enforcement is distribution-based, not legal. For organic posts, TikTok can add its own label to content it detects as AI, reduce the post's reach in the For You feed, or remove it. For ads, it can reject the creative or restrict the campaign so it never serves. There is no public fine; the cost is the reach and the spend you lose when a post or ad gets throttled or pulled — which, for content built specifically to perform, is the whole point of making it.
This is also why the realism exemptions matter in the other direction. If you reflexively label low-risk content — an AI-written caption, a beauty filter — you gain nothing and some creators worry about an unnecessary "AI" stamp on content that didn't need one. The policy is specific precisely so you can label accurately: flag the realistic, potentially-misleading AI, leave the cosmetic and textual AI alone, and you are both compliant and not over-signaling. Note that this is platform policy, distinct from the government rules now emerging — for how statutory AI-ad disclosure requirements layer on top, see how to comply with state AI-ad disclosure laws.
The policy is manageable for one video a week. It becomes a process problem the moment you are producing realistic AI content at volume — which is the entire premise of generating content with AI. If an engine is spinning up avatar videos, synthetic spokespeople, and face-locked persona images across a weekly calendar, "did this one need the label?" has to be answered for every piece, every time, or the misses accumulate into throttled posts you only notice in the analytics weeks later.
Kompozy is a full AI content generation and multi-platform publishing engine — not a repurposing tool — and some of what it produces is squarely the realistic AI the policy targets: Persona Shorts and Persona HeyGen avatar video put a synthetic person on screen; Persona Photos are face-locked AI images of a consistent character. Other Kompozy outputs sit clean outside the rule — a text post, a quote graphic, an AI-written blog. The policy doesn't care that AI made them; it cares whether they look like a real person or scene. So the compliance job is not "label everything," it's "know which formats trip the realistic-AI line and label those on the way out."
Kompozy's structural advantage here is its per-post review pipeline. Nothing publishes to TikTok straight from the model — every piece lands in a review queue first, which is the natural checkpoint to confirm the disclosure decision for that specific format before it ships, rather than discovering a missing label after a reach drop. You still flip TikTok's own toggle and own the final call; what the pipeline buys you is that the call gets made deliberately, per post, instead of being forgotten at scale. Because the same source can fan into many formats across the eight social platforms plus blog and email, that one review gate is also where you account for the fact that each destination has its own labeling rules — TikTok's toggle, and the parallel requirements covered in YouTube's AI disclosure and likeness rules and Instagram AI-content detection and labeling — so a single AI asset gets disclosed correctly everywhere it lands. Treat disclosure as a line item in production, and the policy stops being a risk and becomes a habit.
TikTok requires creators and advertisers to disclose AI-generated content that shows realistic-looking people, scenes, or voices — anything a viewer could mistake for real. You disclose organic posts with the "AI-generated content" toggle in the post composer, which adds a visible label under your username. Ads use a separate "This ad contains AI-generated content" switch in Ads Manager. The rule comes from TikTok's Community Guidelines for organic content and its advertising policies for paid content, and it applies to AI media made anywhere, not just with TikTok's own tools.
When the content is AI-made or AI-edited and realistic: synthetic or heavily altered video, images, or audio that show a real-looking person or scene, AI voice clones of a real person, face swaps, changing what someone says, or adding and removing people or objects in a way that could mislead. You do not need the label for minor edits like color correction, cropping, everyday beauty filters, generic text-to-speech voiceovers, or AI-written captions and scripts. The test is realism plus the potential to mislead, not whether AI touched the file at all.
Two ways. Content made with TikTok's own AI effects is labeled automatically. And since May 2024 — when TikTok became the first video platform to implement the C2PA Content Credentials standard — TikTok reads the Content Credentials metadata attached to files uploaded from other AI tools and labels them on upload. Those credentials stay attached even if a file is downloaded and re-uploaded. Once TikTok applies an auto label, the creator cannot remove it, so disclosing yourself is only ever an addition to, never a way around, the automatic layer.
Yes — same principle, different control and stricter enforcement. Organic posts use the composer's "AI-generated content" toggle and are governed by the Community Guidelines. Ads use a separate "This ad contains AI-generated content" switch in Ads Manager, governed by TikTok's advertising policies. Once an ad is submitted the toggle locks and can only be turned off by swapping the creative, and duplicating a campaign resets it — so a copied ad can ship unlabeled by default. Spark Ads are boosted organic posts, so the label must already be on the underlying organic post before you promote it.
TikTok can add its own label to the post, reduce its reach in the feed, or remove it outright; for ads, it can reject or restrict the campaign so it never runs. Some AI content is prohibited even with a label — anything that seriously misleads people about public issues or that harms a real person — so disclosure is not a universal permission slip. The practical risk is not a fine but lost distribution: an undisclosed realistic AI post that gets flagged loses the reach it was made to earn.
Kompozy is a full AI content generation and multi-platform publishing engine, and some of what it makes — HeyGen avatar video, face-locked persona images — is exactly the realistic-AI content TikTok's policy targets. The leverage is Kompozy's per-post review pipeline: before anything publishes to TikTok it passes a human gate, which is where you confirm the disclosure decision for each piece instead of discovering it after a flag. You still own the actual TikTok toggle; Kompozy makes the labeling call a routine step in production rather than an afterthought.
TikTok's AI disclosure policy requires you to label AI-generated content that shows realistic people, scenes, or voices — anything a viewer could take as real. For organic posts you flip the "AI-generated content" toggle in the composer, which stamps a visible label under your handle; TikTok also auto-labels files carrying C2PA Content Credentials. Ads use a separate "This ad contains AI-generated content" switch in Ads Manager. Minor edits, filters, and AI captions are exempt; undisclosed realistic AI risks reduced reach, removal, or a rejected ad.
Get started → · ← All guides · Compare Kompozy vs other tools