An Instagram account can have an adult birthday on file and still be moved into Teen Account settings. On November 9, Meta announced that it was expanding worldwide the AI age estimation it uses to identify accounts it suspects belong to teens. The company expects millions more teens to get the protections. That is a projection, not a count of accounts already moved.
The expansion follows a U.S. test that began in April 2025. A guide linked from the new announcement, published in Meta’s Family Center, says the company can draw on clues in posts, comments and profiles, such as references to birthdays or school grades. The age estimate matters because Instagram uses it to decide which rules apply to the account.
The change affects who can reach a teen and what the teen can see. Instagram’s Teen Account settings limit incoming messages to people the teen follows or is already connected to. Meta’s June update says Teen Accounts default to a content filter inspired by 13+ movie ratings, and teens need parental permission to switch to the setting that shows more content.
In its earlier test announcement, Meta said people would have an option to change their settings if the company got their age wrong. In separate guidance published in May, Meta said users it suspects are misrepresenting their age to avoid protections must verify it in cases such as changing a birth date that marks them as under 18 to one that would make them an adult, using an ID or Yoti’s facial age estimation tool. That example describes an attempted birthday change; it does not establish what every user wrongly moved into a Teen Account will be asked to do.
A U.S. settlement approved in August sets a separate requirement for participating jurisdictions. The agreement’s appeal provision calls for a clear way to challenge a decision that misidentified someone as a minor and for a timely response explaining the decision. The provision gives no specific deadline for deciding an individual appeal.
The November announcement does not give an error rate or say how long a correction takes. Those are two different measures of how the system works: how often it gets a decision wrong, and how long a user has to live with the result.