Instagram expands AI age checks: what a teen label changes

Who can message you on Instagram—and which posts you can open—can depend on how old Meta thinks you are, even when your account lists an adult birthday. On November 9, the company announced a global expansion of AI technology that places suspected teenagers into Teen Account protections. It expects to reach millions more teens, although the announcement supplies no numerical detection-accuracy results.

Teen Accounts themselves were already available worldwide. This expansion concerns finding people who should receive those settings despite the birthday they supplied. Facebook’s rollout of the detection technology continues; a similar global launch on Threads is planned for later this year, Meta says.

The consequences are more tangible than an age label. Under Instagram’s published Teen Account defaults, messages can come only from people a teen follows or already has a connection with. Tags and mentions are restricted to followed accounts. Private accounts are the default for existing and new users under 16, and for under-18s signing up.

A reminder appears after 60 minutes of daily use. That reminder is distinct from a parent-set usage limit that blocks access. Sleep mode, scheduled from 10 p.m. to 7 a.m., silences notifications and sends automatic message replies. Under-16s need parental permission to loosen the protections or use Instagram Live.

Content restrictions have also evolved. Meta’s April announcement says under-18s receive its updated “13+” setting and need parental permission to opt out. Restrictions extend beyond recommended videos to followed content and links received in messages. For example, a post sent by a friend may still be unavailable to open. The movie-rating comparison describes Meta’s policy approach: the Motion Picture Association does not rate Instagram’s content or endorse the settings. Meta acknowledges that inappropriate material can still appear.

Automatic teen protections also differ from parental supervision. Meta says supervision requires both parties’ consent; even with supervision, parents cannot read message contents.

Behind these changes are several ways of assessing age. They provide different evidence:

Apple’s distinction matters because Meta says it has begun using that signal in Australia. Apple’s declaration categories distinguish information supplied by the person, information supplied by a parent or guardian, and information confirmed using a credit card or government ID. A guardian’s declaration and an ID-supported confirmation therefore arrive with different evidence behind them. Meta’s announcement does not specify which declaration types it accepts in Australia, or announce worldwide use of Apple’s signal.

Nor does the global Instagram announcement establish that every detection technique is operating everywhere. In May, Meta described photo-and-video analysis as available in selected countries. A July correction said that analysis does not use biometric data, create facial fingerprints or identify the person pictured; it recognizes general visual patterns suggesting someone is underage. These are Meta’s descriptions of its system.

Being classified as a teenager is also different from being classified as too young to hold an account. Meta says suspected under-13 accounts can be deactivated, with proof of age required to prevent deletion. Its earlier guidance notes that Instagram’s minimum age is higher in some countries.

Independent research helps explain why estimates need scrutiny, without supplying an error rate for Instagram. A 2024 NIST evaluation of six facial-age algorithms found performance varied with image quality, age, gender and region of birth. That historical benchmark does not measure Meta’s expanded profile-based classifier or validate its effectiveness.

For someone assigned teen settings by mistake, Meta’s age-assurance guidance says age verification can restore the correct settings. Its May explanation describes ID or Yoti checks for suspected age misrepresentation, including attempts to change a birthday from under 18 to over 18. Those statements establish a correction route, but do not promise that every method is available to everyone, how quickly a case will be resolved, or how reliably mistakes will be corrected.