Meta's Age Classifier

Back in 2022, Meta began rolling out one of its most important but understated safety tools: an AI-powered “adult classifier”, trained not on government IDs but on behavioral signals and usage patterns.

In 2024, Instagram adopted it at scale. In early 2025, Meta made a bolder move actively detecting teens pretending to be adults, and automatically shifting them to Teen Accounts with strict default protections.

The Original Challenge: Age Verification Is Broken

Verifying a user’s age online has long been one of the internet’s unsolved problems:

• Teens often fake their age to access restricted features

• Government ID checks introduce friction and raise privacy concerns

• Age misrepresentation poses serious safety and legal risks

Rather than rely on intrusive checks, Meta developed an AI-driven classifier trained on subtle, contextual signals:

• Birthday-related posts (e.g., “Happy 15th!”)

• Account creation history and friend demographics

• Interaction behavior with related posts, stories, and messages ( people in the same age group often interact with content they see in similar ways)

• Time-of-day usage, profile metadata, and more

These models were carefully anonymized and reviewed by humans, producing one of the most robust behavior-based age classifiers ever deployed at scale.

 Instagram’s AI Shift in 2024

In 2024, Instagram quietly began rolling out this AI to classify users under 18 based on signals like:

• Message content (e.g., birthday greetings)

• Who users follow and interact with

• Shared content patterns among age clusters

As part of this rollout, all teen accounts were upgraded automatically with enhanced safety features, including:

• Private by default 

•  Strangers blocked from sending DMs 

• Limited exposure to sensitive content 

• In-app nudges to “Block & Report” suspicious messages 

• Geo-specific AI models to increase relevance and accuracy

The 2025 Push: Auto-Detection of Teens Pretending to Be Adults

In early 2025, Meta escalated its enforcement strategy. The company publicly stated it would now proactively search for teens who misreport their age.

Here’s how this is unfolding:

AI “Adult Classifier” Deployment at Scale Meta’s tool now automatically separates users into two primary buckets: under 18 and 18+.

Signals Used:

• Profile details

• Follower networks

• Engagement patterns

• Usage history

• Matching emails used across age-misrepresenting accounts

• Phone identifiers and device signals

 If a teen is detected using an adult profile:

• They will be migrated to a Teen Account, regardless of the listed age

• Users under 16 will need parental consent to change privacy settings

• Users 16–17 will retain more control, but still start with protected defaults

Age Updates Now Require Real Proof Teens wishing to update their age will need to either:

1. Submit a government-issued ID (like a driver’s license), or

2. Use a video selfie analyzed by Yoti, a third-party AI verification provider. Meta ensures these images are not stored and are deleted post-verification.

The Strategic Shift: From Verification to Prevention

Meta isn’t just verifying age, it’s reengineering platform safety.

By using AI to identify risks before harm occurs, and automatically enforcing age-based settings, Meta has shifted from:

❌ Reactive moderation ✅ Proactive protection

This is critical because most abuse and exposure risks happen before a violation is flagged. By ensuring safety directly into onboarding, identity, and messaging behavior, platforms can dramatically reduce harm.