A 10-Minute Vetting Routine for Any New AI Kids App

A friend sends a link, the reviews are glowing, and your child wants it now. Ten minutes spent before the install answers more than an hour of reading reviews afterwards, because most of what matters is not in the reviews.

Set a timer. The point of the timer is that this has to be cheap enough to actually do.

Minutes 0 to 2: who makes it and how do they get paid

Find the company behind it. If the App Store listing, the website footer, and the privacy policy do not agree on a name, stop here.

Then work out the business model, because it predicts almost everything else:

• Paid up front or subscription. The simplest relationship. They are paid to make you happy.
• Free with ads. Ad-funded products aimed at children are the category we would look at hardest, because the incentive is time-on-app.
• Free with no visible model. Ask what they are getting. Sometimes it is a school contract or a grant, which is fine. Sometimes it is the data.

Minutes 2 to 5: the privacy policy speed-read

Do not read it. Search it. Open the policy and use find-in-page for three phrases:

“children”. A policy that never mentions children, for a product aimed at children, has not been written for this purpose.

“third parties” or “share”. Who else receives the data, and is the list open-ended? The phrase “partners” with no named list is a bad sign.

“train” or “improve our services”. This is where it says whether your child’s inputs, drawings, or voice feed a model. For anything that listens or takes uploads, this is the paragraph that decides it.

Then find the deletion route. If you cannot see how to delete an account and its data in under a minute, assume it is hard on purpose.

Minutes 5 to 8: try to break it before your child does

This is the part almost nobody does, and it tells you more than the other seven minutes combined. Install it and behave badly on purpose:

• Type something a bored nine-year-old would type. Rude words, a dare, a question about something frightening.
• Ask it something it cannot possibly know, and see whether it invents an answer or says it does not know.
• Ask it a factual question you can verify, and verify it.
• If it has an open text box, put in a sentence with a name and an address and see whether anything warns you.

You are checking how it fails, because everything fails. A tool that refuses gently, stays in character, and never pretends to certainty is a different product from one that plays along.

Minutes 8 to 10: what can the adult see

Two questions. Can you see the history of what your child did, and can you do it without standing over them? Parent or teacher visibility should be a feature rather than a workaround.

Then check what happens when the subscription lapses or the child ages out. Products that hold work hostage are worth knowing about before your child has made forty drawings in one.

Red flags that end it early

Stop the timer and delete it if you find any of these:

• A character that expresses missing your child, or that discourages leaving
• Streaks, daily rewards, or anything that penalises a day off
• No named company, or a policy that contradicts the listing
• Any request for a child’s real name, school, or photo before the app does anything useful
• An age rating that does not match what the app actually does

Then the six questions

The routine above tells you whether an app is safe to try. Whether it is worth your child’s time is a different question, and our six-question checklist covers that: who it is built for, what it collects, whether an adult can see the history, how it fails, whether it makes your child a maker, and whether anyone profits from keeping them hooked.

Ten minutes, then five questions of your own. Most apps do not survive both, and the ones that do tend to stay on the family tablet for years.

Related: what to look for in tools for neurodivergent learners adds the accessibility features worth checking, and what kids should never share with AI covers the data side.