YouTube asks creators to disclose when a video contains realistic content that has been meaningfully altered or synthetically generated, such as AI-generated footage of a real place or a cloned voice of a real person, in a way viewers could mistake for real. You do this with a simple yes/no setting in YouTube Studio during upload. Using AI for scripts, ideas, captions or obviously unrealistic animation does not normally need disclosing. This guide explains the rules as they stand at the time of writing and how to apply them in a real workflow.
Check the source. YouTube's policies change. The authoritative page is YouTube Help: >Disclosing use of altered or synthetic content. Treat this guide as a practical explainer, and check that page before relying on any detail.
What the disclosure is and when it arrived
YouTube announced its approach to labelling AI-generated and altered content in late 2023 and introduced the disclosure setting in YouTube Studio in 2024. When you tick "yes", YouTube adds a label telling viewers the content is altered or synthetic.
At the time of writing:
- For most videos, the label appears in the expanded description.
- For videos about sensitive topics, such as elections, ongoing conflicts, natural disasters, finance or health, YouTube may show a more prominent label on the video player itself.
- YouTube says it may add a label itself where a creator has not disclosed and the content could mislead viewers.
- YouTube also says that creators who consistently choose not to disclose when required may face penalties, which can include content removal or suspension from the YouTube Partner Program.
The disclosure is about transparency for viewers. It is separate from YouTube's other policies, which still apply in full whether or not you disclose: a misleading or harmful video is not made acceptable by a label.
What needs disclosing
YouTube's test is whether the content is realistic, meaning a viewer could easily mistake it for a real person, place or event, and whether it has been meaningfully altered or synthetically generated. The Help page gives examples along these lines:
- Making a real, identifiable person appear to say or do something they did not say or do, including a digitally generated or cloned voice.
- Altering footage of a real event or place, such as making a real building appear to be on fire or changing a real city scene.
- Generating a realistic-looking scene of an event that did not happen, such as a lifelike depiction of a tornado moving towards a real town.
For creators using AI tools, that translates to questions like these:
| What you did | Likely to need disclosure? |
|---|---|
| Cloned a real, identifiable person's voice for narration | Yes |
| Face-swapped or lip-synced a real person | Yes |
| Photorealistic AI footage of a real historical event presented as if it were real | Yes |
| Photorealistic AI "news footage" of an event that never happened | Yes, and it may break other policies too |
| Realistic AI footage of a real, recognisable place changed in a way viewers could believe | Yes |
What does not need disclosing
YouTube's guidance says disclosure is not required when the content is clearly unrealistic, or when AI or editing tools are used in ways that do not create a misleading realistic depiction. Examples of the kind YouTube lists include:
- Clearly unrealistic content such as animation, fantasy scenarios, or someone riding a unicorn.
- Colour adjustments, lighting filters, background blur and vintage-style effects.
- Beauty filters and other minor visual enhancements.
- Using generative AI to help with productivity: writing or outlining a script, generating ideas, or creating automatic captions.
- Special effects that are obviously effects.
So a stylised, cartoon-like AI animation, a Pixar-style character short or a clearly illustrated explainer will generally not need the label, even if every frame was generated.
The grey areas creators actually hit
The examples above are clear. Real channels tend to sit in between. These are the questions we see most, with how we would reason about them. None of this is official YouTube guidance.
Generic AI narration
A stock text-to-speech voice reading your script is not imitating a real, identifiable person, and the viewer is not being misled about a real event. YouTube's examples focus on realistic depictions of people, places and events, so a generic narrator voice is not the main target of the rule. If your voice sounds like, or is presented as, a specific real person, disclose. See our guide to writing for AI voiceover for the practical side of TTS narration.
Photorealistic AI images in a documentary
This is the area to be most careful with. If a history or true-crime video uses lifelike AI images of real people or real events, a viewer could reasonably believe they are archive photographs. That is exactly what the disclosure exists for. Either disclose, or make the style clearly non-photographic (illustrated, diorama, animated) so no one could mistake it.
Reconstructions and dramatisations
Broadcasters have long used reconstructions with an on-screen "reconstruction" caption. The same instinct works on YouTube: if you recreate a real event realistically, tick the disclosure box, and consider an on-screen note too.
AI-generated B-roll of generic scenes
A realistic AI clip of "a busy office" or "rain on a window" does not depict a specific real person or event, so it is less likely to mislead anyone about something that matters. Whether it needs the label depends on how it is used. If it is presented as footage of a particular real place or incident, disclose.
When in doubt
Disclosing when you did not strictly need to is a small cost. Failing to disclose when you should have risks viewer trust and, if it is a pattern, penalties. If you genuinely cannot tell, disclose.
How to disclose in YouTube Studio
The steps below reflect the Studio interface at the time of writing; labels and positions can move.
During upload
- Upload your video in YouTube Studio as normal.
- On the Details page, find the section asking about altered content. It asks, in plain terms, whether your video contains realistic altered or synthetic content.
- Select Yes if your video meets the criteria above, or No if it does not.
- Complete the rest of the upload and publish or schedule as usual.
For a video that is already published
- Open YouTube Studio and go to Content.
- Select the video to open its details.
- Find the altered content setting, change it, and save.
The YouTube mobile app offers the same question when you upload Shorts or other videos from a phone.
Content made with YouTube's own AI tools
At the time of writing, YouTube says content made with some of its own generative AI features is labelled automatically. You do not need to do anything extra for those, but you still need to disclose any other realistic synthetic content in the same video.
How this fits with other YouTube policies
The disclosure setting is one of several things to get right when AI is part of your production:
- Monetisation policies. The YouTube Partner Program has its own rules about reused and mass-produced content (at the time of writing, grouped under "inauthentic content"). A disclosure label does not make low-effort, repetitive content eligible for monetisation, and using AI does not automatically make content ineligible. Original scripting, editing and value to the viewer are what matter.
- Privacy requests. YouTube has a process for people to request removal of AI-generated or altered content that simulates their face or voice. Using a real person's likeness without consent is risky even with a label.
- Misinformation and harmful content. A label does not excuse content that breaks the Community Guidelines.
- Copyright. Training data questions aside, you still need rights to music, footage and voices you use. See our guide to copyright-safe music.
Our faceless YouTube channel workflow shows where disclosure decisions sit in a full production pipeline.
Quick checklist
Before you publish any video that used AI tools, ask:
- Does any part look or sound like a real, identifiable person saying or doing something they did not?
- Does any part show a real place or event altered in a way viewers could believe?
- Does any part show a realistic-looking event that did not happen?
- Could a reasonable viewer mistake AI imagery for real footage or photographs?
- If yes to any: set altered content to Yes in Studio.
- If AI was used only for scripts, ideas, captions, clear animation or cosmetic effects: No is usually correct.
- Re-check the official YouTube Help page if your use case is new to you.
Keeping a simple note in your project folder of which AI tools touched which shots makes this decision quick and consistent. Our guide to video project folder structure suggests where to keep it.
FAQ
Does ticking "yes" reduce my views or stop monetisation?
YouTube presents the disclosure as transparency information for viewers. It does not describe the label itself as a monetisation penalty. Monetisation depends on the YouTube Partner Program policies, which apply regardless of the label. Check YouTube Help for the current position.
Do I need to disclose AI-written scripts?
No. At the time of writing, YouTube's guidance treats using AI to help write scripts, outlines or ideas as productivity use that does not need disclosing.
My channel uses cartoon-style AI animation. Do I need the label?
Generally not, because clearly unrealistic or animated content is exempt. If a cartoon depicts a real person saying things they never said in a way that could be taken as genuine, think again.
Where can I read the official rules?
YouTube Help: Disclosing use of altered or synthetic content, at >support.google.com/youtube. It includes YouTube's own examples and any recent changes.