Starting August 13, 2025, YouTube will use AI to figure out which U.S. users are under 18—even if they lied about their birthdate when they signed up.
This change is meant to stop relying on self-reported age data and instead enforce safety settings automatically for minors.
The system analyzes three key signals:
- Search history (e.g., queries for teen-oriented content)
- Video categories watched frequently
- Account longevity (how long the account has existed) – Machine learning cross-references these patterns to flag potential minors. The technology has already been tried out in other places, and it will be slowly introduced in the US.
Teens now have more protections
Accounts that are flagged as underage will automatically receive:
- Non-personalized ads (replacing targeted ads)
- Content safeguards blocking repetitive recommendations on sensitive topics like body image
- Digital well-being tools: “Take a break” alerts, bedtime reminders, and upload/privacy nudges. These mirror existing protections for verified teen accounts but will now apply proactively.
What adults can do if they get flagged wrong
YouTube acknowledges the AI isn’t perfect. Users who have been misidentified can verify their age by:
- Uploading a government ID
- Submitting a credit card (no charges made)
- You can also take a real-time selfie. Without verification, access to age-restricted videos remains blocked.
Final words
The update comes in response to mounting pressure from U.S. states and international regulators (such as those in the U.K. and E.U.) for stricter measures to protect children online.
YouTube’s behavior-based system is meant to be stronger than porn sites’ age checks, which are easy to get around. Teen viewers may not click on as many ads, which could mean less money for creators, but YouTube says this will only affect a small number of channels.
YouTube will keep an eye on the trial in the U.S. before going global. It means a tighter safety net for parents and teens, one that puts real-time behavior ahead of self-reported data.
