Short answer: On YouTube, disclose content that is meaningfully altered or synthetically generated when it looks realistic—for example, making a real person appear to say something they did not or generating a realistic event that never occurred. On TikTok, creators must label realistic AI-generated content and are encouraged to label content that is completely generated or significantly edited by AI. Minor production help does not automatically require a YouTube label. A disclosure also does not make prohibited content acceptable.
Verified July 2026. This is educational platform guidance, not legal advice. Rules and labels change; review the linked official platform pages for the content, country, and publishing date.
Ask what the viewer could believe, not whether AI touched the file
AI can help outline a script, remove background noise, translate captions, generate a photorealistic person, or fabricate footage of a real disaster. Treating all of that as one category produces either useless overlabeling or dangerous underlabeling.
Both YouTube and TikTok focus heavily on realistic content that could mislead a viewer about what is real. The practical test is:
Could a reasonable viewer believe a real person said or did this, a real event happened, or a real scene was recorded when it was actually created or materially changed?
If yes, use the platform disclosure. If the subject is sensitive or the risk of confusion is high, add plain-language context inside the content or caption as well. The platform switch should not carry the entire explanation.
YouTube vs. TikTok at a glance
| Question | YouTube | TikTok |
|---|---|---|
| Core required category | Meaningfully altered or synthetic content that seems realistic | Realistic images, audio, or video generated or significantly modified by AI |
| Broader voluntary category | Creators can disclose other synthetic content | TikTok encourages labeling content completely generated or significantly edited by AI |
| Common tool | AI use setting under Attributes during upload or editing | Creator AI-generated label setting, clear caption/text/hashtag, and automatic labels |
| Automatic detection | YouTube may apply a label even if the creator does not | TikTok may auto-label content, including through Content Credentials metadata |
| Does the label itself reduce reach? | YouTube says disclosure does not limit audience or monetization | TikTok says the setting does not affect distribution when content follows guidelines |
| Does labeling make prohibited content allowed? | No | No |
| Ad disclosure covered? | No | No |
The platforms use different wording and interfaces. Do not copy a TikTok caption to YouTube and assume both systems are satisfied.
What YouTube requires creators to disclose
Want help landing brand deals? Creators Agency represents finance YouTubers and handles everything from negotiation to payment. See if you qualify to join our roster.
YouTube says creators must disclose content that is meaningfully altered or synthetically generated when it seems realistic.
Its examples include content that:
- makes a real person appear to say or do something they did not;
- alters footage of a real event or place;
- generates a realistic-looking scene that did not occur;
- uses another person's cloned voice for voiceover or dubbing;
- synthetically creates realistic music or other media in covered circumstances.
YouTube distinguishes those cases from minor edits and production assistance. Examples it says generally do not require creator disclosure include:
- color adjustment or lighting filters;
- beauty or special-effects filters that do not create a misleading realistic event;
- background blur and visual enhancement;
- idea generation, outlines, scripts, thumbnails, titles, or infographics made with production assistance;
- caption creation or repair;
- audio or video cleanup;
- clearly unrealistic fantasy content;
- cloning your own voice for voiceovers or dubs, under YouTube's listed examples.
Do not stretch an example beyond its context. If a minor-looking edit changes the meaning of a real event, it is no longer merely cosmetic.
How to disclose on YouTube
During upload in YouTube Studio, open Attributes, find AI use, and select the appropriate answer. In most cases, you can also update the disclosure after publishing.
YouTube says photorealistic AI content may receive a label in the video player, while non-photorealistic or animated AI content can receive a label in the expanded description. It may also label content made with YouTube's own GenAI tools, content carrying C2PA metadata, or content detected by its systems.
YouTube can also apply a label when the creator does not, and repeated failures to disclose may lead to content removal or suspension from the YouTube Partner Program. That is a strong reason to make the decision during production rather than minutes before publication.
What TikTok requires creators to label
TikTok defines AI-generated content broadly as images, video, or audio generated or modified by AI. Its mandatory focus is realistic AIGC: content showing people or scenes that appear real.
TikTok says creators must label AI-generated content that contains realistic image, audio, or video. It also encourages creators to label any content completely generated or significantly edited by AI.
Creators can disclose by:
- turning on TikTok's AI-generated content setting before publishing;
- adding clear text, a sticker, or caption context;
- using an unambiguous label such as “AI-generated”; or
- relying on an automatic label only when TikTok has actually applied it.
TikTok can automatically label content it identifies as AI-generated. It also supports Content Credentials, a metadata standard that can carry information about how media was made. Do not strip provenance metadata to avoid a label.
Some AI-generated content is prohibited even when labeled
TikTok's Community Guidelines prohibit certain synthetic or manipulated media. Its help page gives examples involving fake authoritative sources or crisis events, and some uses of public or private figures. The exact prohibited categories and exceptions belong to the current official policy.
The operational point is simple: labeling is not permission. First determine whether the content is allowed. Then determine how it must be labeled.
Example classifier
| Content example | YouTube | TikTok | Why |
|---|---|---|---|
| AI helped brainstorm the script; creator filmed normal footage | Generally no altered-content setting required | Usually no AIGC label for the finished footage solely because of brainstorming | Production assistance did not create misleading realistic media |
| Noise removal and automatic captions | Generally no | Generally no if the media's meaning is unchanged | Cleanup and accessibility support |
| Creator uses a beauty filter that smooths skin | Generally no under YouTube's minor-edit examples | Consider degree; label if the edit becomes significant or realistically changes identity/content | Materiality matters |
| Fictional dragon in an obviously fantasy animation | Generally no required realistic-content disclosure | Not required solely because it uses an artistic or unrealistic style under current Community Guidelines; the creator can still use the AI label for broader transparency | Platform standards differ |
| Photorealistic executive says words they never said | Required | Required and may be prohibited depending on context/person | False realistic speech |
| Real protest footage altered to add violence that did not occur | Required | Required; could also violate misinformation/manipulated-media rules | Changes a real event's meaning |
| AI avatar demonstrates a fictional app feature as though it exists | Disclose realistic synthetic content; also correct the false product claim | Disclose; policy and advertising rules still apply | Label does not cure deception |
| Creator-approved clone of their voice translates an ad | Evaluate YouTube's self-voice example plus the video's other alterations; disclose when realistic material change could confuse | Label realistic or significantly AI-modified audio | Consent and platform disclosure are separate |
When the classification is close, look at the complete post, not the isolated asset. A clearly absurd synthetic image can become misleading when placed under a caption claiming it is breaking news.
AI disclosure and sponsorship disclosure are separate
Suppose a brand pays a creator to publish a photorealistic AI video.
The audience may need to know at least two things:
- the media is altered or synthetic;
- the creator has a commercial relationship with the brand.
The AI label answers the first. The paid-partnership or branded-content disclosure answers the second. Use both when both apply.
Do not hide sponsorship disclosure in a long caption after an AI label, and do not assume #ad explains that a realistic person is synthetic. For US endorsements, review the FTC's official disclosure guidance. Brands and creators should get legal advice for their specific claims and markets when the risk is meaningful.
What brands should put in the brief
AI disclosure often fails because nobody owns the source information. A reviewer sees a finished video but does not know which voice, face, scene, or event was generated.
Add an asset record to the brief:
| Field | What to record |
|---|---|
| Asset | File or scene using AI |
| Tool and operator | What made it and who controlled the output |
| Real subject | Person, voice, place, product, or event depicted |
| Material change | What was generated or altered |
| Consent or rights | Permission for likeness, voice, source media, and commercial use |
| Viewer risk | What someone could mistakenly believe |
| Platform label | YouTube, TikTok, both, or neither with rationale |
| In-content context | Caption, voiceover, graphic, or disclosure language |
| Approver | Named brand and creator owner |
This does not need to become a college thesis. It needs to let the person publishing answer the platform's question accurately.
Creator approval still matters
A brand may have the technical ability to generate a creator's face or voice. That does not mean the use was agreed.
Before any synthetic extension of a real creator, settle:
- the exact likeness, voice, or performance being used;
- platforms, territories, formats, and term;
- whether the model can train on creator material;
- what the creator approves before publication;
- whether edits, translations, or new statements are allowed;
- paid-media scope and spend cap;
- deletion, archival, and model-retention duties at expiry.
Those are contract and consent issues, not solved by a platform label. Our AI likeness and creator contract guide covers that boundary. Brands deciding whether synthetic talent belongs in the strategy can use our AI influencers vs. human creators guide.
Pre-publish decision tree
1. Is any final image, audio, or video generated or materially altered?
If no, the AI-label question ends. Normal sponsorship, copyright, and advertising rules still apply.
If yes, continue.
2. Could it look or sound real?
If it realistically depicts a person, statement, event, place, product behavior, or scene, default to the platform disclosure.
If it is clearly unrealistic, YouTube may not require the setting, while TikTok still encourages labeling fully or significantly AI-created content.
3. Does the change affect meaning?
Noise cleanup is different from changing what someone said. Color correction is different from adding an object that changes the event. If meaning changes, disclose.
4. Is the content allowed?
Check current Community Guidelines, especially for public figures, private individuals, crises, news, elections, health, and financial subjects. Stop if prohibited; a label does not fix it.
5. Is a commercial relationship involved?
Apply the platform's paid-partnership or branded-content tool and any clear endorsement disclosure required for the market. Do this in addition to the AI label.
6. Would a short explanation help?
When stakes or ambiguity are high, say what was generated: “AI-generated reenactment,” “voice translated with AI with the creator's approval,” or “synthetic product visualization.” Specific context is more useful than a mysterious badge.
Pre-publish checklist
☐ Every generated or altered final asset is identified.
☐ A real person's face, voice, action, or statement is not fabricated without documented permission and review.
☐ Real events and places are not materially changed without disclosure.
☐ YouTube's AI use setting is completed accurately.
☐ TikTok's AI-generated-content label is enabled when required.
☐ Any automatic label is confirmed rather than assumed.
☐ Current prohibited-content rules were checked.
☐ AI and sponsorship disclosures are handled separately.
☐ Product claims remain accurate and substantiated.
☐ Creator and brand approvers understand exactly what was generated.
☐ Source files, consent, and approval records are retained.
Add the viewer-belief check
Before the next upload, add one line to the review process: “What could a viewer think is real that was actually generated or materially changed?” Answer it in plain English. Then set the correct platform label, add context where useful, and complete sponsorship disclosure separately.
Official sources
Frequently Asked Questions
YouTube lists production assistance such as generating a script, outline, title, thumbnail, or infographic among examples that generally do not require altered-content disclosure. If the final realistic audio or video is also synthetically generated or materially changed, evaluate that separately.
YouTube lists beauty filters among minor edits that generally do not require disclosure. Degree and context matter. If a filter or edit significantly changes a real person's identity, body, action, or a claim being demonstrated, re-evaluate under both platforms' current rules.
TikTok's mandatory rule focuses on realistic AIGC, and it encourages creators to label content completely generated or significantly edited by AI more broadly. Using the creator label on fully AI-generated fictional media is the clearer choice.
YouTube says disclosure itself does not limit audience or monetization eligibility. The underlying content still must meet Community Guidelines, monetization rules, and advertiser-friendly standards.
TikTok may apply automatic labels based on its detection or Content Credentials. Creators should not try to evade provenance. Check the current help page for how a specific label was applied and any available appeal process.
No. It explains current YouTube and TikTok platform tools and practical publishing steps. Advertising, privacy, publicity, election, and consumer-protection laws can impose separate obligations. Use qualified legal counsel for a specific high-risk campaign.
Stop leaving money on the table.
We represent finance and business YouTubers and handle brand deals from pitch to payment. Apply to join the roster and let us do the heavy lifting.
Apply to Join Our Roster →Also building on YouTube? Check out Money Matchup for creator resources.