The Parent's Operating Manual for Raising a Kid With AI

Most advice about children and AI is a list of rules. Screen-time limits. Approved-app lists. "Always supervise." The rules are not wrong, exactly. They are just aimed at the wrong problem.

Rules answer "how do I stop something bad from happening?" The question that actually determines how your child turns out is different:

When my child sits in front of an AI in ten years — at university, at work, alone — what habits will they reach for?

Nobody builds those habits from a rule. They get built the way every other habit gets built: by doing the thing, repeatedly, with someone who notices.

This is the operating manual we use. It is not a curriculum and it will not take over your evenings. It is a small set of moves that fit into the fifteen minutes you already have.

Last reviewed: August 2026.

The three ways this goes wrong

Before the system, the failure modes — because if you recognise one of these in your house, that is the thing to fix first, and everything else can wait.

Failure 1 — The wish machine

Your child asks the AI for something. The AI produces it. Your child looks at it, says "cool", and asks for the next thing.

This is the most common failure and the hardest to see, because it looks like engagement. The child is busy, the output is impressive, everyone is pleased. But nothing is accumulating. Ten sessions in, the only skill that has improved is wishing.

The tell: ask your child "what did you make?" If the honest answer is "I asked it to make a dragon", not "I made a dragon that…", you are in failure 1.

Failure 2 — The forbidden fruit

The other extreme. AI is banned, or so heavily fenced that it never happens. The intention is protection, and for a five-year-old it is defensible.

The cost arrives later, and it is not the missing skill — skills can be learned quickly. The cost is that your child's first real relationship with AI forms somewhere you are not. On a friend's phone, at school, in a bedroom at fourteen. The first time they meet a confidently wrong machine, you will not be in the room to say "check that."

Failure 3 — The homework laundry

The child photographs a worksheet, sends it to a chatbot, copies the answer. Everyone knows this one. Most families respond with a rule against it.

The rule fails for a specific reason: it asks a ten-year-old to choose the harder path with no visible reward. The child is not confused about whether it is cheating. They have simply, correctly, noticed that the cost of not doing it is invisible and the benefit is immediate.

The only thing I have seen actually work is not a rule at all. It is a moment — the first time a child catches the AI being confidently, provably wrong about something they know. After that, "check it" stops being a rule you enforce and becomes an instinct they own. It is worth engineering that moment on purpose, and it is easier than it sounds.

Why the age policies matter more than you think

One piece of ground truth that reframes all of this, because most parents have it backwards.

The three assistants your child has heard of are, by their own terms:

Read that table again if your child is under thirteen. Every one of these is off-limits to them as an account holder. The three free-for-teachers programmes that launched into US schools through 2025–2026 are teacher tools; two of them explicitly exclude students.

Which means the legitimate paths for a 6-to-12-year-old are exactly two:

1. Tools built for children — narrower on purpose, no open-ended chat.
2. An adult-operated session on a grown-up tool — your account, you present, your child directing.

Path 2 is not a loophole. It is the richest version, and it is the one this manual is built around. The friction is the feature: your presence is what turns a wish machine into a workshop.

The one principle

Everything below comes from a single idea, and if you read nothing else, read this:

Optimise for what stays in the child, not for what appears on the screen.

Every general-purpose AI tool is optimised for the opposite — remove friction, produce output, delight the user. That is correct for adults doing work. It is exactly backwards for a child who is supposed to be growing.

So the whole job of a parent in an AI session is to reintroduce, deliberately and cheerfully, the small amount of friction where learning lives. Not to make it hard. To make it theirs.

Which sounds abstract until you see what it means on a Tuesday evening.