YouTube likeness detection scans newly uploaded videos for AI-generated or AI-altered versions of your face, then hands you the matches to review. It is a detection tool, not a removal tool: a match starts a review, and only a request that passes YouTube’s privacy test can take a video down. This guide explains who can enrol, how enrolment works, what each match can actually get you, and what the tool still cannot do.
Direct answer: YouTube likeness detection is a free YouTube Studio tool that watches new uploads for videos in which your face appears altered or generated by AI. You enrol with a government ID and a short selfie video, review each match YouTube finds, and choose to request removal, file a copyright claim, or archive it. Detection alone removes nothing.
What likeness detection is: a Content ID-style scan for a person’s face rather than for copyrighted audio or video. YouTube’s own help page describes it exactly that way, and the comparison is useful: Content ID finds copies of your work, likeness detection finds copies of your face. Both hand you a queue to review; neither deletes anything by itself. This guide sits inside our face search and photo privacy guide, which covers every route for finding and removing photos and videos of you.
If the fake is a still image rather than a video, this tool is not the right route at all. A still photo of you that has been AI-generated or altered is handled by the still-image reporting process, which our guide to an AI-generated image of you and how to report it covers step by step. Likeness detection watches videos on YouTube only. Keep that boundary clear and you will not waste a week waiting for a tool that was never built for your case.
What is YouTube likeness detection?
YouTube likeness detection is an experimental feature inside YouTube Studio that finds videos where your face may have been altered or generated by AI. Once you enrol and verify your identity, YouTube’s systems perform a one-time search of newly uploaded videos for your face, and any possible matches appear in a review list under Content detection, then Likeness, then For review. You open each one, watch it, and decide what to do.
Three facts from YouTube’s own help page shape everything else in this guide. First, the scan covers newly uploaded videos: it is a watch on what comes next, not a trawl through everything ever posted. Second, the system can surface genuine, unaltered footage of you as well as fakes, because it matches faces, not manipulation. Third, a match is a lead, not a verdict: YouTube still weighs parody, satire, public interest and consent before anything is removed.
That last point is the one most guides bury. A face match does not prove the video is fake, and a fake does not automatically prove the video must come down. The tool finds candidates; the privacy complaint process decides. Anyone who treats the review list as a takedown queue will file weak requests, and weak requests teach you nothing except that YouTube said no.
The tool also exists because the manual alternative has stopped working. A creator with a recognisable face cannot search YouTube for themselves every day and hope to catch every AI clone, false endorsement or fabricated news clip. Automated matching is the early-warning layer; your judgement is the decision layer. Both are needed, and neither replaces the other.
Who can enrol in YouTube likeness detection?
Any creator aged over 18 who owns or manages a channel can enrol, following the expansion YouTube announced in May 2026. YouTube’s help page states the requirements plainly: you must be over 18, you must be a Channel Owner or Manager to set up the feature, and you must complete identity verification with a government-issued ID and a brief selfie video. The selfie video doubles as a face reference for the detection system.
Enrolment is per person, not per channel. If two people appear on camera, each enrols separately. Channel Owners, Managers, Editors and Editors (limited) can review matches and report privacy violations on behalf of any enrolled Owner or Manager, so a team can share the review work, but the person whose face is being watched must enrol themselves.
Two cautions belong here, because they are the reasons most enrolment attempts fail. First, the rollout is phased and the feature is not available in every country, so the Likeness tab may simply not exist in your Studio yet. That is a rollout fact, not an error on your part, and the section near the end of this guide explains what to do when the tab is missing. Second, teams that share one Google account run into name mismatches: the account becomes tied to the first member who completed Payments or AdSense verification. YouTube’s own troubleshooting advice is to add each team member’s individual Google account as a Channel Manager, have each person log in with their own account, and make sure the name on the Google payments profile matches the government ID exactly.
The rollout reached its current shape in stages. The timeline matters because older guides still describe gates that no longer exist: most commonly the Partner Program requirement, which ended in May 2026.
| Date | What happened | Who could enrol |
|---|---|---|
| September 2024 | YouTube previews likeness management technology | Nobody yet, announcement only |
| December 2024 | Pilot begins | Actors and NBA and NFL athletes represented by Creative Artists Agency |
| April 2025 | Pilot widens to creators | Six creators, including MrBeast and Marques Brownlee |
| 21 October 2025 | Launch to monetised creators | YouTube Partner Program members |
| March 2026 | Expansion to high-risk public roles | Journalists, government officials and political candidates |
| 21 April 2026 | Expansion to entertainment | Talent agencies, management companies and the celebrities they represent |
| May 2026 | Opened to all adult creators | Any creator aged 18 or over; Partner Program membership no longer required |
| 23 September 2026 | Made on YouTube announcements | Voice detection to be folded in later in 2026; enrolment and match alerts coming to the mobile app |
One wording wrinkle is worth knowing if you are exactly 18. The announcement language says “18 and older” while the help page says “over 18”. If you are exactly 18 and the option does not appear, check eligibility in Studio rather than assuming you are excluded, and check again after the next rollout wave.
How do you enrol in YouTube likeness detection?
Enrolment takes about ten minutes of your time and up to five days of Google’s. You do it on a computer, in YouTube Studio, the mobile app does not yet offer enrolment, though YouTube said at Made on YouTube in September 2026 that mobile enrolment and match alerts are coming. Do not enrol from a shared team login; use your own Google account, because the verification ties to the person, not the channel.
- Open YouTube Studio on a computer and sign in with your own Google account. Confirm you hold the Owner or Manager role on the channel.
- Go to Content detection in the left menu, then Likeness, then Start now. If Likeness is not there, your account, country or rollout wave does not have it yet, see the later section on a missing tab.
- Read the consent screen and agree. You are agreeing to biometric processing of a face reference. YouTube also says it will process audio from your uploads and may ask for a voice reference later. This is the point to read, not to skim.
- Complete Google identity verification. Provide a clear photo of your government-issued ID and record a brief video of your face, the selfie video. A QR code may appear for you to scan with your phone to complete this step. Use a clear, well-lit photo of the ID; a blurry one is the commonest reason verification stalls.
- Decide on the optional model-improvement checkbox. YouTube offers a separate, optional consent to use your face and voice templates to develop and improve its detection models. It is not part of enrolment, it can be unticked later, and your data is not used to train Google’s generative AI models without your consent.
- Wait for the confirmation email. Verification can take up to five days. You will get a confirmation email when it completes, and only then does scanning begin.
- Repeat for each person on camera. Every person who appears regularly and wants protection enrols separately, with their own ID and selfie video.
Before you start, run through the enrolment checklist. Each item is a real reason enrolments fail, taken from YouTube’s own help page.
| Check | What to confirm before you start | If it is wrong |
|---|---|---|
| Age | You are over 18 | Under-18s cannot enrol; a parent or guardian cannot enrol on a child’s behalf for this tool |
| Role | You are Channel Owner or Manager | Ask the Owner to add you as Manager first |
| Account | You are signed in with your own Google account, not a shared team login | Shared accounts create name mismatches tied to the first member’s Payments or AdSense verification |
| Name match | The name on your Google payments profile matches your government ID exactly | Fix the payments profile name before verifying, or verification fails |
| ID photo | A clear, well-lit photo of a valid government-issued ID | Retake it, a blurred or cropped ID is the commonest stall |
| Selfie video | A short, front-facing video of your face in good light | Retake it; this video is also your face reference for detection |
| Tab present | Content detection, then Likeness, appears in Studio | Your country, role or rollout wave does not have it yet, do not assume an error |
| Model consent | You have decided whether to allow template use for model improvement | You can change this later under Manage likeness detection |
One honest note about this article: we did not perform a live enrolment while writing it. Enrolling requires signing in to a Google account and submitting a government ID, and our standing rule is never to log in with the owner’s Google ID. Every step above was verified against YouTube’s live help page on 12 November 2026, and the wording follows that page. If a step on your screen differs from what is written here, trust your screen, the feature is experimental and details move.
What happens when YouTube finds a match?
When likeness detection finds a possible match, it appears in YouTube Studio under Content detection, then Likeness, then For review. You can filter the list by a video’s total views or a channel’s subscriber count, which is how you triage: a fake with a million views and a link in the description matters more than a fake with eleven views. Select Review next to a video to watch it and choose an action.
Each match offers exactly three actions, and choosing the right one is the whole skill of using the tool. The decision table below is the ranking asset for this article, work through it for every match instead of defaulting to “request removal” every time.
| What the match is | What it gets you | Which action to take |
|---|---|---|
| A realistic AI-made or face-swapped video of you that you never approved, presented as if it were genuine | A privacy complaint; YouTube reviews it under its privacy guidelines and emails you the outcome. The uploader may first get time to trim or blur the video | Submit a likeness removal request |
| A video that reuses your own footage without permission: your clips, your recording, your edit | A copyright removal request, judged under copyright law, not privacy policy | Submit a copyright removal request, but first consider fair use, public domain or another copyright exception |
| Genuine, unaltered footage of you that another channel re-uploaded | Nothing under the privacy route: real footage cannot be removed under YouTube’s privacy policies just because it appears in the list | Copyright request only if the footage is yours and no exception applies; otherwise archive it |
| A parody, satire or clearly labelled joke using your face | Usually nothing, parody and satire are factors YouTube weighs, and the complaint is likely to be rejected | Move to archive, unless the video also makes a false claim that harms viewers |
| A video that is clearly about someone else, a lookalike, or a false positive | Nothing, and filing a complaint wastes your credibility with reviewers | Move to archive |
| A video you are unsure about and want to keep as evidence | The video stays live on YouTube, but it leaves your review list; you can return to it later from the Archived tab and start the privacy complaint process then | Move to archive |
Two practical habits make the review list useful instead of overwhelming. First, record context for every match you act on: the channel name, the video URL, the publication date, what the video claims, and any sign of deception or missing consent. A complaint that says “this video uses my face without permission, presents me endorsing a product I never endorsed, and links to a payment page” is reviewable; a complaint that says “this is me” is not. Second, do not publicly accuse an uploader before YouTube has decided. A match is a lead, and lookalikes, commentary and transformative uses all appear in the same list.
If you change your mind after submitting a likeness removal request, you can retract it through the link in the confirmation email YouTube sends you. That link opens a short form. Keep the email until the case is closed.
What can you actually get removed, and what stays up?
Removal is decided under YouTube’s privacy guidelines, not by the detection tool. YouTube’s privacy page lists the factors reviewers consider, and they are worth reading as a checklist before you file anything: whether the content is altered or synthetic, whether it is disclosed to viewers as altered or synthetic, whether the person is uniquely identifiable, whether the content is realistic, and whether it contains parody, satire or other public-interest value. You can read the full criteria on YouTube’s privacy guidelines, which also explain that a first name without additional context, or a fleeting image, would not likely count as uniquely identifiable.
The strongest removal cases are the obvious ones: a realistic AI video that puts your face on a false endorsement, a fabricated news clip, or a scam that uses your likeness to ask viewers for money. Those fail the consent test, pass the uniquely-identifiable test, and carry no parody or public-interest defence. The weakest cases are the ones creators file most often: a comedy sketch that plainly labels itself as AI, a news report that uses a short clip of you in a story about you, or a commentary video that criticises you. YouTube reserves the final decision, and it says so plainly.
Real footage deserves its own warning, because it surprises people. The detection system matches faces, not manipulation, so your review list can include genuine clips of you that another channel uploaded. That footage cannot be removed under the privacy route merely because it appears in the list. If the footage is your copyrighted content used without permission, the copyright route is the right one, provided you have considered fair use, public domain or a similar exception first. If neither route applies, archive the match and move on.
There is also a limit on who can file. Only you, a parent or guardian, or a legal representative can report on your behalf, and YouTube will not share your name or contact details with the person who posted the content. If the video is intimate in nature, an AI-generated intimate image or video of you, do not use this workflow at all. That case belongs to the specialist removal routes, which our guide to an AI-generated image of you and how to report it signposts in its first section, including StopNCII and the Take It Down route. File there, not here.
What likeness detection does not do
The gaps matter as much as the feature, because each one has a different correct route. Getting the route wrong costs days.
- It does not detect voices, yet. At the time of writing, likeness detection aims to find facial content that looks like you. YouTube says it is working to extend detection to audio in 2026, and some users now have the option to record a speaking-voice sample during setup. Until voice detection reaches your account, a fake of your voice must be reported through the privacy complaint process yourself. Never tell a reader that voice detection is live for everyone; it is not.
- It does not scan old uploads. The system performs a one-time search of newly uploaded videos. A fake posted before you enrolled will not surface in your review list. If you find it yourself, report it through the privacy complaint process directly, enrolment does not backfill the past.
- It does not watch other platforms. A fake of you on TikTok, Instagram or X is outside YouTube’s system entirely. Still images need the still-image route; videos elsewhere need each platform’s own reporting process. Our guide to what to do when your photo is used in a YouTube or TikTok video covers the video side across platforms.
- It does not identify anyone who has not enrolled. The scan detects faces in videos, including bystanders and children, but it immediately discards data for faces that do not match an enrolled creator, and it cannot identify anyone else. You cannot enrol a friend, a colleague or a child and watch for their face.
- It does not remove anything on its own. Every match waits for your decision, and every removal request waits for YouTube’s privacy review. There is no automatic takedown, and there should not be, a system that deleted videos without a human checking them would delete parody, news and commentary along with the fakes.
- It does not work inside the YouTube mobile app yet. Enrolment and match review are desktop Studio actions at the time of writing. YouTube announced mobile enrolment and alerts at Made on YouTube in September 2026, but until that ships, use a computer.
Your data is also worth a clear-eyed paragraph, because enrolment asks for a government ID and a face video. YouTube says your selfie video, legal name and likeness template are stored for up to three years from the last time you signed in to YouTube, unless you withdraw consent or delete your account. Turning the feature off deletes that feature data, and ID information stored in your Payments Profile can be accessed and deleted separately. The optional model-improvement consent can be revoked at any time under Content detection, then Likeness, then Manage likeness detection. That is a reasonable bargain for a creator whose face is their livelihood, and a bargain worth refusing if you only appear on camera once a year.
Worked example: the manual check before enrolment
To see what a creator faces before enrolment, we ran a real test on 12 November 2026 using the site’s own reverse image search tool. The test photo was a public portrait of Marques Brownlee, one of the six creators in YouTube’s likeness-detection pilot in April 2025, taken at SXSW on 13 March 2026 and uploaded to Wikimedia Commons on 21 March 2026 (photographer Jay Dixit, CC BY-SA 4.0, 2,073 ร 2,700 px). We pasted its direct image URL into the tool and ran each engine separately.
Google Lens failed on all three attempts, returning “No image at the URL” and zero results, the Google and Lens buttons open the same Lens page, so the failure counted twice. TinEye ran and searched its index of more than 86.7 billion images, but found 0 matches, saying it had “yet to crawl any pages where this image appears”. Bing did best: it identified the person by name: “Marques Brownlee”, and its first visual match was the exact same photo, on the English Wikipedia article. The Commons file page itself did not surface. Yandex also identified him by name and showed 16 similar thumbnails, all different photos of Brownlee, but the exact photo was not among them and no Commons or Wikipedia page appeared in its site list. Pinterest was not a visual search at all: its button runs a keyword search for pins from the image’s host domain and returned 28 unrelated pins.
The score, then: two engines named the person, one found the exact photo, none found the Commons source, and Google found nothing at all, on a photo of one of the most photographed creators on YouTube, taken eight months earlier. Two tool quirks are worth knowing: “Search All” opened only the Google tab, and the Pinterest button is a domain keyword search rather than a visual search.
The lesson of that test is the reason likeness detection exists at all. A manual reverse-image search can confirm that a photo of a creator is circulating, and it can sometimes name the pages using it, but it cannot watch YouTube for new videos, it cannot tell a genuine re-upload from an AI fake, and it cannot file anything. It is a spot check. Likeness detection is the standing watch that replaces the spot check, once you are enrolled, and only for videos uploaded after enrolment.
If you are not a creator and you are reading this because a video of you appeared somewhere, the same manual check still has value: run a still frame from the video through a reverse image search to see where else that frame appears. Our guide to how to tell if an image is AI-generated explains the checks that work on a single frame, and Google Lens inside YouTube Shorts can identify objects and places in a clip while you watch it. None of those replaces a privacy complaint, but they tell you whether you are dealing with one upload or a pattern.
What to do if the Likeness tab is missing or you cannot enrol
If Content detection, then Likeness, does not appear in your Studio, work through this list before assuming something is broken.
- Check your role. Only a Channel Owner or Manager can set up likeness detection. If you are an Editor, ask the Owner to change your role or to enrol the person themselves.
- Check your age on the account. You must be over 18. If the account’s date of birth is wrong, fix it first.
- Check your country. The feature is not available everywhere, and YouTube does not publish a country list. If you are outside the rollout, the tab simply will not appear.
- Check your own account, not the shared one. If your team shares a Google account, sign in with your own account and have the Owner add it as a Manager.
- Wait for the next rollout wave. Access has widened roughly every few months since late 2024. If you are eligible on paper and the tab is absent, it is a rollout fact, not a fault.
While you wait, the privacy complaint process is still open to you. You do not need to be enrolled to report a video that uses your face without permission: open the video, use the report route, and choose the privacy option, or start from YouTube’s privacy guidelines page. The complaint asks you to explain how you are uniquely identifiable and why the content violates the guidelines, and YouTube may give the uploader time to trim or blur the video before its own team reviews it. That route works for a voice-only fake today, for a video posted before you enrolled, and for a video the tool missed. Enrolment makes finding fakes easier; it was never the only way to report one.
Finally, keep your own evidence pack even after enrolment. Save the URLs of matches you act on, keep dated copies of your original recordings, and keep the confirmation emails from every complaint. Likeness detection is experimental, matches can be wrong, and a request can be rejected and need a second, better-documented attempt. The tool is a watchtower, not a vault: it shows you what is coming, and your records are what make a removal request stick.
Does YouTube likeness detection automatically remove AI videos of me?
No. Detection only finds possible matches and lists them for review in YouTube Studio. You must choose an action for each match, and a removal request is then judged under YouTube’s privacy guidelines. Parody, satire, public interest and consent can all keep a video up, and YouTube emails you the outcome.
Who can enrol in YouTube likeness detection?
Any creator over 18 who is a Channel Owner or Manager can enrol, following the May 2026 expansion to all adult creators. You verify yourself with a government-issued ID and a short selfie video, which also serves as your face reference. Enrolment is per person, not per channel, and the feature is not available in every country.
Does likeness detection find AI fakes of my voice?
Not yet for most users. The tool currently detects faces only. YouTube says it is extending detection to audio in 2026, and some users can already record a speaking-voice sample during setup. Until voice detection reaches your account, report a fake of your voice through YouTube’s privacy complaint process instead.
What should I do if a fake video of me does not appear in likeness detection?
Report it yourself through YouTube’s privacy complaint process, enrolment is not required to file a complaint. The tool only scans newly uploaded videos, so anything posted before you enrolled will never appear. Save the video URL, note what it claims, and explain in the complaint how you are uniquely identifiable and why you did not consent.







