Skip to content
FastClip Logo
fastclips.app
Learn
Algorithm updatesAI disclosure labels

Algorithm updates

AI edits in clips: when to tick "AI use" on YouTube, what auto-labels do, and whether labels cost reach

Since May 27, 2026, YouTube shows AI labels on the player and can apply them itself. Which AI edits in clips need disclosure, and what a label does to reach.

Updated October 4, 20269 min read8 sources

In short

Only if the AI makes realistic content that could mislead: a real person saying or doing something they did not, altered footage of a real event or place, a realistic scene that never happened, or AI music that is the main focus. AI captions, upscaling, sharpening, audio repair, scripts, thumbnails and cloning your own voice for voiceovers do not need disclosure. Since May 27, 2026, YouTube shows the label on the player for photorealistic AI content and can apply it automatically. YouTube says a label alone does not change how a video is recommended or whether it can earn money.

On this page
  1. What changed on May 27, 2026
  2. How the rules developed from 2024 to 2026
  3. What you must disclose, in YouTube's words
  4. Common AI edits in clips: disclose or not?
  5. Auto-labels: when you can remove one and when you cannot
  6. Does an AI label cost you reach or money?
  7. What happens if you do not disclose
  8. What YouTube has not said about AI labels
  9. What to do now: an AI checklist for each clip
  10. Questions
  11. Sources

What changed on May 27, 2026

On May 27, 2026, YouTube announced two changes to its AI labels. A label is a short notice that tells viewers a video contains AI-generated or AI-altered content.

  • More visible labels. For photorealistic and meaningfully AI-altered or AI-generated content, the label moved. On long-form videos it now sits directly below the player, above the description. On Shorts it appears as an overlay on the video itself. For content that is unrealistic, animated or slightly altered, the label stays in the expanded description.
  • Automatic labels. Starting in May 2026, if a creator does not say whether they used AI, and YouTube's systems "detect significant photorealistic AI use", YouTube applies a label itself. YouTube's help page adds that you get a notification in YouTube Studio when this happens.

YouTube still requires creators to disclose manually. The automatic system is a backstop, not a replacement for the "AI use" setting.

How the rules developed from 2024 to 2026

The disclosure duty is not new. What changed is where the label appears and who can apply it.

DateChange
March 18, 2024YouTube introduced a disclosure tool in Studio for realistic content "a viewer could easily mistake for a real person", place, scene or event. Most labels appeared in the expanded description. Videos on sensitive topics such as health, news, elections or finance also got a label on the video
March 2024YouTube began letting people in the US and India request removal of AI content that looks or sounds like them, through the privacy request process
May 27, 2026Labels for photorealistic AI content moved onto the player (long-form) and onto the video (Shorts). Automatic labelling started
September 23, 2026YouTube said likeness detection will add speaking voice detection to face detection "later this year". See likeness detection

What you must disclose, in YouTube's words

YouTube's disclosure help page lists four cases. You must disclose AI content that:

  • Makes a real person appear to say or do something they did not do.
  • Alters footage of a real event or place.
  • Generates a realistic scene that did not actually occur.
  • Creates music that is the main focus of the video.

This covers content that is "fully or partially" altered or created with AI. YouTube's examples include AI-generated extra footage of a real place, a realistic AI match between two real tennis players, and making it appear that someone gave advice they did not give.

The page also lists what you do not need to disclose: clearly unrealistic content; minor, mainly aesthetic edits; beauty filters; colour and lighting filters; background blur; production help such as AI scripts, outlines, thumbnails, titles and infographics; caption creation; "video sharpening, upscaling or repair"; voice or audio repair; idea generation; and "Cloning one's own voice to create voice overs or dubs".

Common AI edits in clips: disclose or not?

The rows marked "YouTube's list" come straight from the help page. The other rows are our reading of the four tests above; YouTube has not ruled on them by name.

AI edit in a clipDisclose?Basis
AI-generated captions or subtitlesNoYouTube's list: caption creation
Upscaling, sharpening or repairing a low-quality stream recordingNoYouTube's list
Removing noise or fixing a speaker's audioNoYouTube's list: voice or audio repair
AI-written hook text, title or thumbnailNoYouTube's list: production assistance
Your own cloned voice as narratorNoYouTube's list
AI extending the background to fill a vertical frameUsually noOur reading: close to YouTube's example of extending a backdrop; it does not change what happened
A generic synthetic narrator that is not a real personUsually noOur reading: no real person is shown saying something. Other policies still apply (see below)
The streamer's or guest's cloned voice saying words they never saidYesMakes a real person appear to say something they did not
Face swap, or a real person inserted into a sceneYesMakes a real person appear to do something they did not
Realistic AI B-roll of a real place or eventYesYouTube's examples: extra footage of a real place; altered real events
AI-generated music as the main focusYesYouTube's list
A Short made with Reimagine, Veo or other YouTube AI toolsLabelled for youYouTube applies a label it calls permanent

Translation is a grey area. Dubbing a guest into another language with a clone of the guest's own voice shows them saying words in a language they did not speak. YouTube's list exempts cloning your own voice for dubs; it says nothing about cloning someone else's. Our advice: disclose in that case, and get the speaker's consent.

Auto-labels: when you can remove one and when you cannot

If YouTube labels your video and you think the system made a mistake, you can usually change it. In YouTube Studio, open the video's details, find "AI use" under Attributes, and select No. The Studio notification stays until you update that setting.

Some labels cannot be changed. YouTube's help page lists three cases:

  • Content made with YouTube's own AI tools, such as Veo or Dream Screen.
  • Content that contains C2PA metadata, a provenance record that some cameras and AI tools attach to a file. YouTube's blog post describes this case as metadata showing the content was fully generated by AI.
  • Content labelled after manual review.

This matters if you edit clips in tools that write C2PA metadata into exports. This is our inference: an AI tool that marks its whole output as generated could trigger a permanent label on a clip that only used AI for one effect. Check whether your editor writes C2PA data, and test with an unlisted upload before a campaign deadline.

Does an AI label cost you reach or money?

YouTube's answer is no, for the label itself. The May 27, 2026 post says "a disclosure label alone does not change how a video is recommended" or whether it can earn money. The help page says disclosure "won't limit a video's audience" or affect eligibility to earn.

The word that matters is "alone". The same AI use can break other rules that do affect reach and money:

  • Monetization. Since July 2026, YouTube's Generic or Repetitive section lists AI content made with generic templates "without adding the creator's original, authentic insights". The Unsatisfying or Off-putting section lists unrelated AI clips stitched together to shock. See YouTube's monetization policies explained.
  • Spam. The Spam Policy bans using AI to "churn out high volumes of similar content with minimal changes". See the Spam Policy and detection evasion.
  • Recommendations. In YouTube's October 1, 2026 Shorts update, as reported by Search Engine Journal, creator liaison Rene Ritchie said "VO descriptions of what's happening on screen" do not add enough value. An AI narrator that only describes the clip is the cheapest version of that.
  • Viewers. YouTube does not publish data on how viewers react to the new overlay on Shorts. Whether fewer people watch a labelled Short to the end is unknown; nobody outside YouTube has shown it with data.

What happens if you do not disclose

YouTube says creators who "consistently choose not to disclose" may get a label applied manually, or penalties including "removal of content or suspension from the YouTube Partner Program." A label applied after manual review cannot be removed by you.

Non-disclosure also weakens your position if the person in the clip complains. When YouTube evaluates a privacy complaint about AI content that looks or sounds like someone, its privacy guidelines list factors including whether the content is synthetic, "whether the content is disclosed to viewers as altered or synthetic", whether the person can be identified, and whether it is parody or satire. Disclosure does not make a deepfake of a real person acceptable; it is one factor among several.

What YouTube has not said about AI labels

Several details are still open. Plan around them instead of guessing.

  • The threshold. YouTube says auto-labels apply when its systems detect "significant photorealistic AI use". It has not defined "significant", so a clip with one short AI shot may or may not be labelled.
  • The detection method. The May 2026 post speaks of "new internal signals" and does not say what they are.
  • Old uploads. YouTube has not said whether the systems scan videos uploaded before May 2026, or only new ones.
  • Viewer effect. YouTube says the label does not change recommendations. It has not shared whether the Shorts overlay changes how long viewers watch.

If any of these matter for a campaign, upload a test Short as unlisted, check YouTube Studio for a notification, and only then publish the real one.

What to do now: an AI checklist for each clip

  1. List every AI step in your edit. Captions, upscaling, reframing, voice, music, B-roll, thumbnails.
  2. Apply the four tests. Does any step show a real person saying or doing something new, alter a real event or place, create a realistic scene, or add AI music as the focus? If yes, set "AI use" to Yes in the Attributes section at upload.
  3. Never clone the creator's voice or face to make new statements without their written consent, even for a campaign. Beyond disclosure, it invites a privacy complaint and, once likeness detection covers voice, a detection match.
  4. Keep AI narration out of the main value of the clip. If the only thing you add is a synthetic voice describing the footage, YouTube's monetization and recommendation rules treat that as little or no added value.
  5. Check Studio after upload. Look for an auto-label notification. If the label is wrong and the case is changeable, select No; if it is permanent, find out which tool wrote C2PA metadata.
  6. Compare platforms. TikTok and Instagram label AI content under their own rules. See TikTok AI labels and Instagram AI labels.

Questions

Do AI-generated captions need an AI label on YouTube?

No. YouTube's help page lists caption creation among uses that do not need disclosure.

I upscaled a blurry stream recording with AI. Do I need to disclose?

No. YouTube lists video sharpening, upscaling or repair, and voice or audio repair, as edits that do not need disclosure.

Where does the AI label appear on a Short?

Since May 27, 2026, the label for photorealistic or meaningfully AI-altered content appears as an overlay on the Short. For unrealistic, animated or slightly altered content, it appears in the expanded description.

Can I remove an AI label that YouTube added automatically?

In most cases, yes: select No under "AI use" in the Attributes section in YouTube Studio. You cannot change labels on content made with YouTube's AI tools, content with C2PA metadata, or content labelled after manual review.

Will ticking "AI use" hurt my campaign views?

YouTube says a disclosure label alone does not change how a video is recommended. It has not published data on how viewers react to the label. The bigger risks come from other rules, such as templated AI content under the monetization policies.

Does YouTube's auto-dubbing need disclosure?

YouTube's disclosure page does not address its own auto-dubbing directly. It does say that content made with YouTube's AI tools may be labelled automatically and that such labels cannot be adjusted.

Sources

We checked every claim on this page against these sources on the date above. Platform rules and rates change often. Found an error? Tell us.

  1. 1YouTube Blog: Improving AI labels for viewers and creators (May 27, 2026)
  2. 2YouTube Help: Disclosing use of GenAI content
  3. 3YouTube Blog: How we're helping creators disclose altered or synthetic content (March 18, 2024)
  4. 4YouTube Help: Protecting your identity (privacy guidelines)
  5. 5YouTube Help: YouTube channel monetization policies
  6. 6YouTube Help: Updates to YouTube's policies (changelog)
  7. 7YouTube Blog: New tools to power your creation journey (September 23, 2026)
  8. 8Search Engine Journal: YouTube to cut reach of channels reposting barely changed clips (October 1, 2026)

Related guides

Was this page helpful?