Evaluate Any AI Tool in 10 Minutes

A new AI app for children launches roughly every week. Any list of recommendations — including a good one — is out of date within months.

What does not go out of date is the ability to open something unfamiliar and know, in ten minutes, whether it is worth your child's time. This is that method: six tests, a scoring card you can print, and the three signals that mean stop immediately.

Last verified: August 2026.

Why ten minutes is enough

Most parents evaluating a children's app do one of two things: read the marketing page, or hand it to the child and watch their face. Neither works.

The marketing page tells you what the company wants to be true. The child's face tells you whether it is novel, which every new thing is for about four minutes.

What actually predicts whether a tool is worth returning to is structural, and structure is visible fast — usually in the first three screens. Ten minutes is enough because you are not evaluating the content; you are evaluating the shape.

The three signals that mean stop

Before the tests, three things that end the evaluation immediately. If you see any of these, close it — nothing that follows will compensate.

1 · An open chat box aimed at children, with no adult in the loop. Not "a chatbot" — specifically an unbounded conversational surface marketed to kids, with no parent visibility. This is where nearly every genuinely bad outcome originates. Bounded tools (draw this, build that, practise this) are a different category entirely.

2 · It asks the child for personal information before an adult is involved. Full name, school, birthday, photo, location — collected in a child-facing flow. Whatever the privacy policy says, the design is telling you what the product is for.

3 · There is no way to see what happened. If a parent cannot review the session afterwards, you cannot supervise it, and "supervision" becomes a word rather than a practice.

Everything else is a judgement call. These three are not.