YouTube AI disclosure rules 2026 are simple at heart: you must label realistic synthetic content, and you do not need to label AI help behind the scenes. If a viewer could mistake your video for something real that never happened, label it. If AI only helped you write the script, no label is needed.
YouTube AI disclosure rules 2026: what you must label
YouTube asks creators to disclose altered or synthetic content when it looks real. The test is realism, not whether you used AI at all. YouTube Help names three cases that need a label: a real person shown saying or doing something they never did, footage of a real event or place that has been changed, and a realistic scene that never happened.
That covers the obvious cases. A deepfake of a real person needs a label. A fake news clip needs a label. A photorealistic scene that you made with AI needs a label, because a viewer could take it for a real recording. The same goes for AI music that sounds like a real performance.
Where the label shows depends on the video. Since 27 May 2026, the photorealistic label sits directly below the video player on long-form videos, and as an overlay on Shorts. For animated or lightly edited content, the label stays in the expanded description.
What you do not need to label
Most everyday AI help needs no label at all. You do not need to label an AI-written script, an AI-written title or thumbnail, AI captions, or upscaling and filters. You also do not need to label cloning your own voice for a voiceover or a dub, because the voice is yours.
Animation and obvious fantasy are also exempt. An AI-generated missile inside a fully animated video needs no label, because no viewer thinks it is real footage. YouTube cares about what a viewer could mistake for real, not about policing your production tools.
A useful rule of thumb: ask whether the AI changed what the viewer believes happened. If the script is AI-assisted but the facts are yours and the footage is real, the viewer is not misled. If the footage shows something that never happened, you must say so.
Since May 2026, YouTube labels some videos itself
The biggest change in 2026 is that YouTube now applies labels on its own. Since 27 May 2026, its systems add an AI label when they detect significant photorealistic AI that you did not declare. The triggers are content made with YouTube's own AI tools such as Veo or Dream Screen, content that carries C2PA metadata, and content its internal systems detect as AI made or changed.
You get a notice in YouTube Studio when a label goes on. If the label came from a wrong detection, you can answer No in the disclosure survey and the label can come off in most cases. But a label from YouTube's own tools, from C2PA metadata, or from a manual review cannot be changed. That matters most for licensed clips, because a clip that carries fully AI C2PA metadata brings a permanent label with it.
C2PA is provenance data that some AI tools write into their files, a kind of signed record of how the file was made. Google writes it into its own AI output. So if you use a fully AI-made clip from a tool that stamps it, YouTube will know, and the label will stay.
Disclosure is not monetization, and creators keep mixing them up
This is the confusion that costs channels the most time. Disclosure and monetization are two separate systems. Disclosure asks whether the viewer knows the content is synthetic. Monetization asks whether the content is original and made with real effort, or mass-produced and templated.
An AI label does not cost you money. YouTube says the label does not change how a video is recommended and does not affect whether it can earn. A video can carry an AI label and still be fully monetized, as long as it is original work with your own voice and commentary. The label is just a note to the viewer, not a strike.
What does cost you money is the inauthentic content policy, which YouTube tightened in July 2025 and clarified again in July 2026. That policy targets mass-produced, repetitive content: the same music, repetitive AI images and an AI script with no human layer on top. A channel that posts fifty near-identical videos a day can lose monetization even if every video is labelled perfectly.
Skipping disclosure is a different risk. Creators who keep posting realistic synthetic content without labelling it can face removal of videos or suspension from the Partner Program. So the two rules point in the same direction: label the realistic stuff, and put your own human work into every video. For the money side in full, read can you monetize AI videos on YouTube and our guide to YouTube automation after the new rules.
June 2026: the spam policy rewrite names mass production
In June 2026 YouTube rewrote its spam policy to name mass production directly: the same music, repetitive AI images and an AI script, repeated across video after video. For faceless channels this is the line that matters most. A faceless video with your own script, your own voice, real research and real editing is a normal video. A faceless video made by feeding the same prompt into the same tools, with only the topic changed, is exactly what the policy describes. The fix is not to hide the AI. The fix is to add the human layer: your take, your structure, your examples, your voice.
Watch your own channel like a reviewer would. If you muted the audio and looked at ten of your videos side by side, would they look like ten different videos or one template? If the answer is one template, change the format before YouTube notices.
How to label a video in YouTube Studio
Labelling takes less than a minute, and it happens while you upload. In YouTube Studio, open the video details, find the altered or synthetic content question, and answer Yes if your video has realistic synthetic content. If you are unsure, answer Yes. Over-labelling costs you nothing, and under-labelling is the risk.
For Shorts, the label sits as an overlay on the video itself, so it is visible the moment someone watches. That makes the check even more important: a Short with a realistic fake scene and no label is the easiest kind of video for a viewer to report.
A quick checklist before you hit publish
Run this check on every upload. Does the video show a real person doing something they never did, or a realistic scene that never happened? Does it use a synthetic voice to sound like a real person? If any answer is yes, label it. If the AI only helped with the script, titles, captions or your own voice clone, no label is needed.
YouTube AI disclosure rules 2026: the bottom line
YouTube AI disclosure rules 2026 ask one thing: tell viewers when something realistic on screen is synthetic. Label deepfakes, altered real footage and photorealistic scenes that never happened, and let YouTube label the rest itself if it spots it. You do not need to label AI scripts, titles, captions or your own cloned voice. And remember the split that trips creators up: a label never costs you monetization, but mass-produced templated videos can. Label honestly, add your own work, and you are on the safe side of both rules.
Frequently asked questions
Do I need to disclose AI on YouTube in 2026?
Yes, when the content looks real. You must label realistic synthetic content, such as a real person shown saying or doing something they never did, altered footage of a real event or place, or a realistic scene that never happened. You do not need to label AI scripts, titles, thumbnails, captions, filters, animation or cloning your own voice.
Does the AI label hurt monetization?
No, YouTube says the AI label does not change how a video is recommended and does not affect whether it can earn money. Monetization is decided by a separate policy on inauthentic content, which targets mass-produced and templated videos. A labelled video with your own voice, research and commentary can still be fully monetized.
Can YouTube label my video without asking me?
Yes, since 27 May 2026 YouTube adds labels itself when its systems detect significant photorealistic AI you did not declare. The triggers are content made with YouTube AI tools, content carrying C2PA metadata, and content its own detection flags. A wrong detection can usually be fixed in Studio, but labels from YouTube tools, C2PA or manual review cannot be removed.
What is C2PA, and why does it make a label permanent?
C2PA is provenance data that some AI tools write into a file, a signed record of how it was made. When a file carries C2PA metadata saying it was fully AI-generated, YouTube applies a label that cannot be taken off. Check licensed clips before you buy or use them, because the metadata travels with the file.
What happens if I never disclose AI content?
Creators who keep posting realistic synthetic content without labelling it can face removal of videos or suspension from the YouTube Partner Program. A single missed label is unlikely to end a channel, but a pattern of undisclosed realistic AI content is treated as a real problem, so label whenever you are unsure.