Guide
AI Content Rules on TikTok and TikTok Shop
TikTok and the FTC enforce AI-content rules independently. TikTok wants the AIGC label on anything realistic. The FTC, under 16 CFR Part 465, doesn’t care about the toggle — it cares whether your AI-voiced review is honest about being AI. Here’s what triggers each system, and how to stay clean on both.
TL;DR — quick rules
- •Flip the in-app AI-generated content toggle whenever your video contains realistic AI-made images, audio, or video — TikTok requires this, not just “encourages” it.
- •Files carrying C2PA Content Credentials auto-label on upload — the platform recognizes the metadata even if you didn’t flip the toggle.
- •Add a spoken or on-screen disclosure when the AI use isn’t obvious from context — especially for affiliate or sponsored content where viewers might mistake it for a real review.
- •The label does not rescue deepfakes of private people, fabricated public-figure statements, or AI-altered demos that misrepresent what a product actually does.
- •For TikTok Shop, never use AI to fake an expert endorsement, a customer testimonial, or a before-and-after — the FTC’s fake-review rule (16 CFR Part 465) and TikTok Shop’s AIGC policy both ban this independently.
- •AI voiceover on an affiliate review must make it clear the “reviewer” is synthetic — otherwise it reads as a fake testimonial under FTC rules.
- •Mislabeling real footage as AI to look on-trend is itself a Terms of Service violation. The label is not decoration.
What TikTok counts as AI-generated content
TikTok’s policy covers content that has been “either completely generated or significantly edited by AI” and contains realistic images, audio, or video — the rule is in TikTok’s help-center article on AI-generated content. TikTok Shop’s seller policy widens the definition: AIGC includes “videos, images, audio, text, or live streams that are fully or partially generated, edited, or synthesized using artificial intelligence technologies,” per the TikTok Shop Academy “AI-Generated Content Restrictions and Requirements” page.
Examples that trigger the rule
- •A photoreal AI image of a product on a model who doesn’t exist
- •An ElevenLabs-style voiceover narrating a review in a human-sounding voice
- •A digital human or synthetic avatar “hosting” a live or video
- •AI-generated B-roll dropped into an otherwise real demo
- •Heavy AI editing of a real face or voice in a way that changes meaning
Examples that usually don’t
- •Auto-captions or AI-generated subtitles
- •AI-assisted background music with no synthetic voice
- •Color grading, beauty filters, or speed ramps without identity manipulation
- •ChatGPT-drafted captions or scripts you then film yourself
The line is “realistic and meaning-changing.” If a reasonable viewer could mistake the AI element for real footage of a real person, label it.
The AIGC label toggle — when, how, what shows
TikTok launched the creator-side toggle on September 19, 2023 and made labeling a requirement for “realistic images, audio or video.” The help-center article confirms the workflow: after recording, tap Next, then More options, then enable AI-generated content — see the help-center walkthrough.
For short-form video
- 1Record or upload the video as usual
- 2Tap Next on the post screen
- 3Open More options → AI-generated content
- 4Toggle on. TikTok adds an "AI-generated" caption tag visible to viewers
- 5Optional: add a brief written note ("AI voiceover") in the caption for extra clarity
The toggle is required, not optional. Per the help-center article: TikTok “requires creators to label all AI-generated content that contains realistic images, audio, and video.” Skipping the toggle on qualifying content is a policy violation even if no one complains.
Auto-labeling via C2PA Content Credentials
On May 9, 2024 TikTok became the first video platform to read Content Credentials from the Coalition for Content Provenance and Authenticity (C2PA). When you upload a file that already carries C2PA metadata — for example, an image exported from Adobe Firefly or a clip generated in OpenAI’s tools — TikTok reads that metadata and applies the AI label automatically, no toggle required.
What this means in practice
- •If you use a major generative-AI tool, the file likely arrives pre-tagged. The label will appear on your video whether you wanted it or not.
- •Strip-the-metadata workflows (re-export through tools that drop C2PA) won’t hide the AI use under TikTok’s policy — the obligation to label is still on you.
- •TikTok also attaches its own Content Credentials to AI content created with its in-app tools, so the provenance follows the file when downloaded and re-uploaded elsewhere.
When the label is not enough — verbal or on-screen disclosure
The platform label proves you flipped the switch. It doesn’t prove the viewer noticed. For commercial content the safer rule is layered disclosure: toggle on, and add a spoken or on-screen line that names the AI use. This matters most when viewers might form an impression that depends on the speaker being real.
Add explicit disclosure when
- •An AI voice is reviewing a product you’re promoting on TikTok Shop
- •A synthetic avatar is presenting the content as if it were a real person
- •AI imagery shows the product in use (demos, before/after, lifestyle shots that look real)
- •The content makes any factual claim a viewer might rely on for a purchase decision
Plain language works: “AI voiceover,” “generated with AI,” “this image was AI-made.” Caption text and on-screen text are both fine — pick whichever a scrolling viewer would actually catch.
What TikTok bans even with a label
The label is not a permission slip. TikTok’s Community Guidelines, Integrity and Authenticity section prohibits several AI-content categories outright, label or not. The TikTok Shop AIGC policy reinforces these and adds its own.
Realistic AI depictions of private individuals without consent
If the person in the video is recognizable and isn’t a public figure, you need their consent. “It was AI” is not a defense.
Fabricated statements or actions attributed to public figures
AI-generated clips of a real politician, executive, or celebrity saying things they didn’t say are removed even with a label.
Misleading information about real events
Synthetic media that misrepresents what happened in news, public emergencies, or other verifiable events.
Manipulated election content
AI-altered content of candidates, election officials, or processes designed to mislead voters.
AI-fabricated product results the product can’t deliver
Per the TikTok Shop AIGC policy: showing “results that the product cannot realistically achieve” is prohibited.
Synthetic experts, doctors, or research institutions
TikTok Shop explicitly bans fabricating “professional identities of experts, doctors, scientific research institutions” to create false endorsements.
AI of yourself vs. AI of other people
The rules differ depending on whose likeness, voice, or identity the AI is producing. Both paths require labeling. Only one is fully under your control.
AI of you
- •AI voiceover narrating your own script: allowed with the label and a clear disclosure
- •AI avatar trained on your face presenting your content: allowed with the label
- •Heavy AI editing of your real footage: label if it’s realistic; skip if the AI is just stylized
AI of someone else
- •Private person, recognizable: needs explicit consent
- •Public figure, fabricated statements: prohibited regardless of label
- •“Generic” synthetic person who isn’t supposed to be anyone real: allowed with the label, but watch the FTC rules below if they’re reviewing a product
TikTok Shop’s AI rules go further
The TikTok Shop AI-Generated Content Restrictions and Requirements policy applies on top of the platform-wide rules and adds Shop-specific prohibitions. The enforcement language is explicit: violations trigger a tiered response from warnings to “account restrictions, including limiting posting capabilities, permanently disabling commission withdrawal privileges” up to permanent bans.
Shop-specific bans
- •AI must not “replace or fabricate product displays in a way that misleads users about the actual product”
- •Fabricating expert, medical, or research-institution identities to drive endorsement is banned
- •Using anyone’s likeness, voice, name, or trademarks without permission is banned
- •Showing AI-generated results the product cannot realistically deliver is banned
AI voiceover on affiliate and review content
This is the single most common AI-content trap for TikTok Shop creators. An AI voiceover reviewing a product reads, structurally, as a customer testimonial. If the voice isn’t a real person who actually used the product, it’s a fabricated testimonial — and that’s prohibited under 16 CFR § 465.2 whether or not the AIGC label is on.
Allowed
- ✓AI narrating a product explainer in your own script, with the label on and “AI voiceover” in the caption
- ✓AI clone of your own voice reading your own opinion, clearly labeled
- ✓AI voice reading neutral product facts on the product page
Not allowed
- ✗AI voice presented as a customer who “used this for 3 months”
- ✗AI voice cloning a real reviewer or influencer without their permission
- ✗AI voice making medical, weight-loss, or income claims even with the label on
Synthetic creator avatars and AI endorsers
A synthetic avatar — a digital human, a generated face, an “AI influencer” — can host content if the AIGC label is on and the avatar isn’t presented as a real person who actually used the product. The line is identity: a generic digital spokesperson is fine; a digital human pretending to be a verified user is not.
The hardest case: AI avatars styled to look like real customers giving testimonials. Even with the AIGC label, the testimonial is fabricated. FTC 16 CFR § 465.2 and the TikTok Shop AIGC policy both ban this. The label tells viewers the speaker isn’t real, but the implicit claim — “real customers love this” — is what the rules target.
AI-generated product imagery
Product images carry an implicit claim: this is what you will receive. AI imagery that materially diverges from the real product runs straight into the TikTok Shop AIGC rule against fabricating product displays.
Clean lifestyle render of a real product, AIGC label on
An AI-generated room background with the real product composited in, color and proportions accurate. Label it AIGC, and you’re fine.
AI before-and-after of a beauty result
Synthesized “30 days later” imagery showing results the product can’t reliably produce. Banned even with the label.
AI “demo” of a feature the product doesn’t have
Generating a video of capability the actual SKU doesn’t ship with. Fabricated product display, prohibited.
FTC 16 CFR Part 465 — the fake-reviews rule
The FTC finalized its Trade Regulation Rule on the Use of Consumer Reviews and Testimonials (16 CFR Part 465) in 2024. The rule explicitly covers AI-generated reviews and testimonials and applies whether the offender is a brand, a seller, or an individual creator. Civil penalties are available per violation.
What the rule prohibits
- •§ 465.2: writing, creating, selling, buying, or disseminating fake or false reviews and testimonials — including AI-generated ones from reviewers who don’t exist or didn’t use the product
- •§ 465.4: paying for positive or negative reviews conditioned on the sentiment expressed
- •§ 465.5: insider reviews — soliciting reviews from employees, managers, or their relatives without clear and conspicuous disclosure of the connection
- •§ 465.7: review suppression — using unfounded legal threats to take down or hide honest negative reviews
- •§ 465.8: misuse of fake social-media indicators such as bot followers and engagement
In plain language: every AI-voiced review of a product on TikTok Shop has to be honest about two things — that the speaker is not real, and that the speaker didn’t use the product. Otherwise it’s a fabricated testimonial under § 465.2, regardless of whether the AIGC toggle is on.
FTC Endorsement Guides and AI-assisted content
The FTC Endorsement Guides (16 CFR Part 255) require disclosure of material connections — paid partnerships, affiliate commissions, gifted products — “clearly and conspicuously,” per § 255.5. AI doesn’t change that. An AI voiceover doing an affiliate review still has to disclose the affiliate relationship, exactly the same way a human voiceover would.
Bottom line on the two FTC rules: Part 255handles the “is this a paid promotion” disclosure. Part 465 handles the “is this a real customer” question. AI-voiced affiliate content has to satisfy both. Toggling AIGC on TikTok handles neither in the FTC’s eyes.
See the companion TikTok Shop Disclosure Guide for the exact disclosure mechanics inside the app.
If a video gets auto-labeled and you disagree
Auto-labeling can fire on assets that carried C2PA metadata from upstream tools even when the AI use was incidental. Before you appeal, audit the file: AI B-roll, an exported still from a generative-AI tool, or any element processed through an AI-enabled editor can trip the auto-label legitimately.
Order of operations
- 1Identify every source file in the project — original footage, B-roll, stills, audio. Any AI provenance anywhere usually justifies the label.
- 2If the AI use is real but you think it shouldn’t require labeling (e.g. auto-captions), ask whether the asset is realistic and meaning-changing — that’s the threshold, not whether AI “touched” the file.
- 3If you’re confident the file has no qualifying AI, submit feedback through the in-video options. Don’t delete and re-upload as a workaround — that pattern triggers further review.
- 4If the AIGC label persists incorrectly, run the project file through a tool that strips C2PA metadata and re-export. Note: this is only legitimate if the underlying AI use truly doesn’t qualify.
Don’t fake the label off. Per the help-center article, “misleadingly labeling unaltered content” — and by symmetry, stripping the label off qualifying content — “is a violation of our Terms of Service and may result in the removal of content.”
Enforcement patterns
TikTok’s public language describes a tiered escalation rather than a single strike count. The TikTok Shop AIGC policy describes “warnings,” account restrictions including “limiting posting capabilities” and “permanently disabling commission withdrawal privileges,” and permanent bans for severe or repeated violations.
Lower severity
Unlabeled realistic AI on isolated videos. Typical pattern: video removal, warning, possible reach suppression on that content.
Mid severity
Repeated unlabeled AIGC, AI-fabricated product displays, or synthetic testimonials. Posting limits and CHR impact begin to compound.
High severity
Deepfakes of real people without consent, fabricated experts, AI election content. Commission withdrawal can be disabled; permanent bans are explicitly within scope.
See also the Violation Appeal Guide if you’ve already been flagged.
Pre-post checklist
Need this checked against your specific script? Run a Pre-Check — Polici grades AI voiceover scripts against both the TikTok AIGC policy and 16 CFR Part 465 before you post.