YouTube's New Likeness Detection Tool: Should You Hand Over Your Face and ID?
YouTube's new Likeness Detection tool scans for AI deepfakes of your face, but enrolling means handing over your ID and a selfie video. What it covers.
The short answer
Yes, if you are a visible creator and the risk of a convincing deepfake of your face actually worries you. YouTube's Likeness Detection tool scans uploads across the platform for matches to your face and lets you request removal, but enrolling means handing over a government ID and a selfie video first. It only covers your face, not your voice, and it will not catch everything, so treat it as one layer of protection, not the whole plan.
What the tool actually does
Likeness Detection started inside the YouTube Partner Program and has since opened up to all eligible creators aged 18 and over, rolling out gradually rather than switching on everywhere at once. You set it up in YouTube Studio, under Content Detection, then Likeness.
Setup means a one-time identity check: a government-issued ID plus a short selfie video, so YouTube can build a reference of what you actually look like. Once that is done, the system scans video uploaded to YouTube for visual matches to your face, including in thumbnails, and surfaces anything it finds in a review queue inside Studio. You decide what happens next. A match is not an automatic takedown.
That last point matters more than it sounds. Parody, commentary, satire and other recognised fair-use style content stay protected even when they clearly use your likeness. So a reaction channel making fun of your take on something, or a commentary video with a thumbnail that includes your face, is not what this tool is for, and reporting it will probably go nowhere. It is built for the specific case of a video where someone has used AI to generate or substitute your face, usually to make it look like you said or did something you did not.
What it does not cover, yet
It's visual only for now. Your voice is not scanned, which matters a lot if your content leans on a distinctive voice more than a distinctive face, or if a convincing AI voice clone of you turns up narrating something you never said. YouTube has said audio is coming, but there is no committed date, so do not plan around it arriving by any particular point.
It also will not reliably catch content uploaded before you enrolled, and it only works within YouTube. A deepfake using your face that is posted to TikTok, Instagram, X or a site outside Google's reach is entirely outside this tool's scope. If most of your risk sits off-platform, which for a lot of creators it does, this solves a smaller slice of the problem than the announcement implies.
And the tool itself is still labelled experimental. YouTube's own help documentation says availability varies by country and that the feature is not guaranteed everywhere, so the honest move is to check Content Detection in your own Studio account rather than assume it is live for you because you read it was rolling out.
The actual trade-off: verification data for protection
The part worth sitting with before you enrol is what you are handing over to get this. A face template built from your selfie video, plus a copy of a government ID, sitting with YouTube. For most creators whose faces are already the product, on camera in every upload, this is a smaller incremental exposure than it would be for someone whose face is not already public. You are not giving up anonymity you did not already trade away by making video content under your own likeness in the first place.
It is still worth asking what happens to that data, how long it is retained, and what your options are if you change your mind. YouTube's documentation says creators can leave the programme and request deletion of the facial data later, which is the right question to confirm actually works before you rely on the programme rather than after.
If you are a creator who deliberately keeps your face out of your content, or who works under a persona precisely to keep a layer between your real identity and your public one, enrolling in a face-verification programme is a different calculation entirely, and for a lot of those creators the better answer is still no.
What to actually do about it
If your face is already your brand, set it up. The downside is a few minutes of admin and a verification step you have effectively already accepted by being a visible creator. The upside is a direct channel to flag deepfakes without having to find and report them manually, which is the slow part of dealing with impersonation today.
Keep your expectations narrow. This catches AI-generated or AI-substituted video of your face on YouTube. It does not catch your voice, it does not catch anything off-platform, and it will not help with a commentary channel you simply dislike. Pair it with the basics that still do most of the actual work: Google Alerts or a similar watch on your name, a standing process for reporting impersonation on each platform you actually use, and, if this is a recurring problem rather than a hypothetical one, a conversation with a solicitor about what UK law does and does not currently cover for non-consensual AI likeness use. That last point is general information, not legal advice, and the law here is still moving.
Where this fits
Likeness protection is one more thing added to the list of platform admin a working creator has to track alongside everything else: uploads, clipping, scheduling, disclosure labels, and now identity verification settings. CORE handles the repurposing and channel management side of that list, so creators can decide on things like this with their attention rather than losing the afternoon to Studio settings.
Get these as they go out
Two a week, on what actually changed and what to do about it. No pitch.
We only use it to send the posts. Unsubscribe any time.
Comments
No comments yet.