YouTube used its Made on YouTube event on 23 September 2026 to pair new creation tools with a quieter security message: likeness detection, the company’s system for finding AI-generated videos that appear to use a person’s face, is being extended. Aparna Pappu, vice president at YouTube, wrote that billions of channels are eligible to sign up for automated deepfake detection and can control their visual likeness through the existing tool. Later this year, she said, YouTube will start integrating speaking-voice detection with facial detection to improve match accuracy, as groundwork for broader voice protections. The same post said likeness detection will be available in the YouTube mobile app so creators can enrol, review matches and act while away from a computer.
The announcement does not claim that every impersonation will be removed. Earlier YouTube posts on likeness detection, which this article treats as background rather than a new statistical claim, have said a match lets someone request removal when the content violates privacy guidelines, and that parody and public-interest exceptions still apply. Detection is not the same as a takedown. Voice matching, still described as forthcoming, is also not a guarantee that cloned audio used in scams on other apps will be caught.
The rest of Pappu’s post is about Studio features: insights, draft feedback, dynamic thumbnails, conversational editing with Gemini in Shorts and the YouTube Create app, and opt-in AI comment moderation that learns a channel’s style. YouTube said that in August 2026 alone, hundreds of thousands of channels used Gemini Omni in Shorts and Create each day, and it cited an internal GenAI trends report that 72 percent of U.S. video posters aged 14 to 44 had used AI to create or edit in the past year. Those product metrics are YouTube’s. They do, however, explain why impersonation tools and creator-defence tools are arriving together: the same generation of models that helps with editing can fabricate a face and, soon, a voice.
For journalists, civic figures and small businesses, the practical question is enrolment. Likeness detection has required identity verification, including government ID and a face video, in YouTube’s help material. That biometric enrolment is a privacy trade of its own. YouTube has said in prior posts that setup data is used to verify identity and power the feature, not to train Google’s generative models. Anyone considering enrolment should read the current Studio flow rather than assume those limits still hold.
What to do
• If you run a public-facing channel, look in YouTube Studio for likeness or content-detection settings and decide whether identity verification is acceptable for you.
• Once enrolled, review match alerts on a schedule. Request removal only for content that actually misuses your image or voice, and do not expect parody to disappear.
• Warn your audience on a channel you control that you will not ask for money, crypto or passwords by video.
• Keep channel recovery email, 2-Step Verification and backup codes current so an impersonator cannot also steal the real account.
• Small organisations should assign one person to monitor impersonation and to file privacy complaints, rather than leaving it to whoever notices a fake clip first.
Primary source: YouTube Blog, “New tools to power your creation journey from start to finish”, 23 September 2026.
Sources
Cyber71 links the primary material behind each report so readers can check it directly. External pages are not under our control and may change.
