AI tools for kids: how to tell a real one from a repackaged chatbot

Why we're not giving you a ranked list

Ranked lists of AI tools for kids go stale within weeks — products change their age policies, add restricted modes, get acquired, or shut down. A list you can apply yourself keeps working. So this is a test rather than a table, and you can run it on any product, including the ones that launch after this was written.

It's also worth saying plainly: we build one of these tools. Treat everything below as a framework you can use against us as easily as against anyone else. If our answers to these questions were bad, you'd want to know.

The four-question test

• 1. Does it state a specific age range? — "For kids" is marketing. "For 7–12" is a design decision — it means someone chose a reading level, a topic space, and a supervision model. Vagueness here usually means an adult product with a kid-facing landing page.
• 2. Are topics restricted, or just filtered? — A filter blocks bad output after the model generates it. A restricted design limits what the system will engage with at all. The first fails occasionally by nature; the second fails much less often, and differently.
• 3. Can a parent see what happened? — Not a raw transcript dump — a summary you'd actually read. If there's no parent view at all, the product isn't expecting you in the loop, which tells you who it was designed for.
• 4. Does the child finish something? — Products that end in a made thing have a natural stopping point. Products that end when the child gets bored are optimising for time-on-app. This is the signal that predicts the most about a year of use.

The categories, and what each is good for

General-purpose assistants (ChatGPT, Claude, Gemini and similar) are the most capable and the least designed for children — most set a minimum age of 13 in their own terms. Useful with an adult for a specific task; a poor default for a nine-year-old alone.

Homework helpers and tutoring apps sit in the middle. The good ones refuse to hand over the answer and make the child work through it; the weak ones are answer machines with a study-themed skin. The four-question test sorts them quickly — especially question four.

Creation tools — image, music, story, and game builders — are where AI is most obviously worth a child's time, because the child has to decide what they want and judge what came back. Check where the output goes and whether anything is public by default.

Guided AI learning studios, the category Luchi is in, wrap creation in structure: missions with a defined goal, an AI the child directs rather than converses with, and a parent view. The trade-off is honest — less open-ended freedom than a blank chat box, in exchange for a much better default for this age.

Before you let your child use anything new

1. 1. Read the age line in the terms, not the homepage The homepage says who they'd like to sell to; the terms say who they built for. When those disagree, believe the terms.
2. 2. Find out what happens to the data Is chat history stored? Can you delete it? Is it used for training? US services aimed at under-13s must obtain verifiable parental consent before collecting personal information — a signup flow that ignores this is telling you something.
3. 3. Try it yourself for ten minutes first Ask it something a curious child would ask, including something slightly cheeky. You'll learn more in ten minutes than from any review, including this one.
4. 4. Check for a stopping point Use it and see whether the product ever suggests you're finished. If it never does, you'll be the stopping point, every single time.
5. 5. Set the two rules before the first session No personal details, and show me what you made at the end. Agreed up front, they're a boundary; introduced later, they're a punishment.

Questions parents ask

What's the best AI tool for a 7-year-old?

One built specifically for that age, where the child makes something and you can see what happened. At seven, general-purpose assistants are the wrong shape regardless of brand — the issue isn't quality, it's that they were designed for adults.

Are free AI tools for kids safe?

Free isn't the risk factor; the business model is. Ask what the product is optimising for — a subscription tends to align with the child finishing and being happy, whereas engagement-funded products are rewarded for keeping them there.

How many AI tools should my child use?

One, used well, beats five sampled. The habits — direct it, check it, explain it — transfer across tools. Collecting apps doesn't build anything.

Should I trust the 'kids mode' on an adult AI app?

It's better than nothing and worth turning on. But it addresses content, which is the risk you were already going to manage, and not dependence or unverified answers, which are the ones that actually shape a child.

Red flags worth walking away from

• A companion framing — "Your child's new best friend" is a marketing promise about attachment, not learning. A tool that positions itself as a relationship is optimising for the thing you least want a nine-year-old forming with software.
• Streaks, daily rewards, and pressure to return — Borrowed from games and social apps, and they work — that's the problem. They're engineered to make stopping feel like a loss, which is a strange property for an educational product to want.
• No age gate and no consent flow — If a nine-year-old can sign up alone with an email address and nobody asks for a parent, the product has not thought about children as a regulated category — whatever the homepage says.
• Public sharing on by default — Children's creations posted to a feed, with usernames, visible to strangers. Sharing can be lovely; on by default, with no parent in the loop, is a different product than it appears.

One tool used well beats five sampled

Parents who take this seriously often end up with a folder of six AI apps their child has tried twice. It feels like diligence and it usually isn't — because the thing that matters transfers between tools, and collecting apps doesn't build it.

What transfers is the habit: decide what you want, say it precisely, check what came back, explain it to someone. A child who has practised that a hundred times in one product will apply it instantly to a product neither of you has seen. A child who has sampled six has practised nothing. So the useful question isn't which tool is best in the abstract — it's which one your child will actually keep using long enough for the habits to become theirs.

Run the four-question test on us

Luchi is built for 7–12s specifically: restricted by design, projects with an ending, and a parent view of what your child practised. Free to start.

Start free

Part of the series: Teaching kids about AI

Related guides in this series

• How to explain AI to a child — Simple analogies and sample scripts by age (5–12), plus the three things every kid should understand about AI.
• AI activities for kids — 10 free, screen-smart activities to explore how AI works together — no coding, no sign-up, mostly off the screen.