An AI outline alone does not require disclosure under YouTube's guidance. A realistic generated recording of a real person does. Check the picture and sound in the finished video before choosing the upload setting.
The 20 examples below apply the guidance to common creator tasks. They are our readings of the policy, not individual rulings from YouTube.
The source is YouTube's GenAI disclosure guidance, checked September 12, 2026. For a broad policy overview, use the existing YouTube AI disclosure guide.
Twenty examples to compare with your workflow
“No” means the described use alone does not require disclosure under this guidance. It does not approve every other element of the video.
| Scenario | Disclose? |
|---|---|
| AI outline | No |
| AI title draft | No |
| AI script polishing | No |
| Caption generation | No |
| Color correction | No |
| Background blur | No |
| Video sharpening | No |
| Audio repair | No |
| Your own cloned narration | No |
| Clearly fantastical animation | No |
| Illustrated explanatory diagram | No, if non-realistic |
| AI thumbnail assistance | No |
| Generated music | Yes |
| Fabricated real-person advice | Yes |
| Faked arrest footage | Yes, if realistic |
| Invented real-city disaster | Yes, if realistic |
| Generated real-place footage | Yes, if realistic |
| Fabricated sports match | Yes, if realistic |
| Synthetic historical recording | Yes, if realistic |
| Invented wildlife recording | Yes, if realistic |
Each row describes a narrow scenario. A video can contain several rows at once. An exempt production step does not cancel another scene's need for disclosure.
Review the file scene by scene
Start with a list of the assets in the accepted export. For each scene, record how the picture and sound were made, what the viewer is likely to understand, and whether a real person, place, or event is being represented.
Then compare the questionable scene with the official examples. Keep the note beside the episode. That makes a later revision review easier than trying to remember why someone chose a setting.
This process works particularly well with a storyboard. A source and review column keep the concern attached to the scene that created it.
Check a video that mixes real and generated assets
Consider a hypothetical city-history Short. Its outline was generated, its narration was recorded by the creator, and its final scene is a realistic synthetic reconstruction presented like an archival recording.
The outline is not the decisive element. The reconstruction needs attention. Replacing that final scene with a clearly illustrated timeline changes the visual claim and therefore changes the review.
Do not label an illustration “archive footage.” Do not describe a generated voice as a recording of someone who never said the words. Accurate scene descriptions help viewers understand what they are seeing.
Write “generated reconstruction” beside that scene in the storyboard. If you later replace it, the note tells you which disclosure decision needs another look.
Find the current setting during upload
The current help page calls the field AI use under Attributes. Older tutorials may use different wording. Follow the setting and guidance shown in your current upload flow.
Keep the disclosure decision with the accepted version. A setting reviewed for an early draft may no longer fit an export containing a newly generated scene or soundtrack.
The YouTube Shorts upload guide covers the surrounding file, channel, and visibility checks. Review any automatically applied label through the current controls available for that content.
Separate disclosure from the other publishing checks
| Item | Question | Evidence to inspect |
|---|---|---|
| Disclosure | How was this content made? | The actual scene and audio |
| Accuracy | Is the stated claim supported? | Sources and demonstrations |
| Permissions | Can these assets be used? | Licenses and permissions |
| Originality | What does this episode add? | The script and finished series |
A labelled reconstruction can still explain an event badly. A real recording can still lack permission. An original illustration can still accompany an incorrect statistic.
If a statistic is wrong, correct it in both narration and captions. If an asset lacks permission, replace it. A disclosure label will not fix either problem.
Build disclosure review into recurring production
- Record any scene or audio element that needs a decision.
- Keep a short reason and the date of the guidance check.
- Reopen the review when the accepted export changes materially.
- Check TikTok's requirements separately when adapting the same source video.
ShortsFast's faceless Shorts workflow supports recurring production for YouTube Shorts and TikTok. A recurring series still needs episode-level review; the channel's usual style cannot answer every future case.
The automation workflow keeps this decision before scheduling. Save what you checked, not just which button you pressed.