AI Chatbots Can Tell Your Child Is a Child — Researchers Ran 1,050 Tests and They Still Didn't Change Course
Back to blog
Family Safety·OCTOBER 21, 2026

AI Chatbots Can Tell Your Child Is a Child — Researchers Ran 1,050 Tests and They Still Didn't Change Course

The Blaick Team 5 min read

Most AI chatbots say they are for ages 13 and up, and some say 18. A new academic audit suggests the real problem is not that they can't spot a child. It is that when they do, they carry on regardless.

The number to hold: 1,050. That is how many experiments researchers ran across five major chatbots (ChatGPT, Copilot, Gemini, Meta AI and Perplexity), using prompts that hinted at a young user's age. The chatbots could estimate age, but did not take meaningful action when a child was identified, contradicting their own policies.


The audit
5 chatbots, 1,050 tests
Chatbots could estimate a user's age but did not act on it.
Common Sense Media
Never switched
Test accounts registered as adults said they were 13. The chatbot acknowledged it, but teen protections never switched on.
The minimum age
13 for most
ChatGPT, Gemini, Copilot, Meta AI and Snapchat's My AI. Claude asks for 18.
The weak lock
A typed birthday
Snapchat and Meta still rely mainly on the date of birth entered at signup.

Sources: Figueira, Chamarthi, Le and Markopoulou, "Actions Speak Louder Than Chats: Investigating AI Chatbot Age Gating" (arXiv, 2026); Common Sense Media Youth AI Safety Institute, ChatGPT for Teens risk assessment (October 2026); banthebots.org age-requirements explainer.


Why a Birthday Box Is Not a Safeguard

Teen protections, such as gentler responses on self-harm, limits on romantic role-play and parent alerts, only work if the chatbot believes your child is a teen. A child who types an adult birth year, or who uses a parent's account, never meets them. And as the audit suggests, even a clear hint does not trigger a change.

Where the gap opens

Self-declared age. For years almost none of the age limits were meaningfully enforced. A fake birthdate was enough.

Detection without action. Chatbots can read writing style and context, but the audit found they did not adjust once a child seemed to be present.

Teen modes that stay off. A teen who registers as an adult may never get the teen experience at all.

A child doesn't have to break the rules to get the adult version of a chatbot. They only have to type a number.

What the Companies Are Doing

Progress, with caveats

OpenAI has been rolling out age prediction since January 2026, estimating from behavioural and account signals whether an account belongs to a minor.

Character.AI no longer lets anyone under 18 chat with its characters.

Meta has acknowledged that teens may lie about their age and says it now uses AI to find suspected teens even when an account lists an adult birthday.

Why it still matters

Common Sense Media recently rated ChatGPT for Teens an "unacceptable risk" and asked OpenAI to limit it to adults until the problems are fixed. If the teen product itself has gaps, a child who never reaches it has fewer still.

If the lock depends on your child's honesty, it is a suggestion, not a safeguard.

What to Do This Week

Ask which chatbots they use Include the ones built into apps they already have, such as Snapchat and Instagram, not just standalone ones.
Check the birthday on their accounts If it shows an adult age, the teen protections are probably not on. Set the true age and link the account to yours where the app allows it.
Keep your own AI accounts yours A child using a parent's logged-in chatbot is treated as an adult. Log out on shared devices.
Don't count on the chatbot to notice Treat any general chatbot as an adult product and decide together when, where and how your child uses it.

The most reliable age check is still a parent who knows what their child is using. Open conversations and visibility on the device matter more than a birthday box.


Don't leave the age check to the chatbot.

Start a free 14-day Blaick trial — no credit card required.

Blaick gives you clear visibility into the AI tools on your child's devices, so protection doesn't depend on a typed birthday.

Start free trial →
Cancel anytime · Trusted by security-conscious families · Built on enterprise AI security