The Question Underneath the Question
Parents ask whether AI is good or bad for kids, and that question has no answer. A better one: what was this piece of work supposed to build in my child, and does the tool still leave room for it? A ten-year-old formatting a bibliography and a ten-year-old writing a paragraph about a book are doing different jobs, and only one of them is the point of the lesson.
You are in a better position to make this call than any classroom teacher. You know which skill each task is meant to grow, you watch your child work, and you can change the rule tomorrow if it turns out wrong. Nobody is grading your child on output you did not see them produce, so you can drop the whole cheating frame and ask the only question that matters at home, which is what your child walks away knowing.
Age Limits Most Parents Never Check
These tools carry age restrictions, and they differ enough to surprise you. Checked in August 2026: OpenAI sets a minimum of 13 for ChatGPT and asks that anyone under 18 have a parent's permission, and since late 2025 parents can link their own account to a teen's to set quiet hours, switch off voice mode and memory, and opt the teen out of model training. Google takes a different route and lets kids under 13 use Gemini Apps through a supervised account a parent enables in Family Link, with image generation blocked for anyone under 18. Anthropic requires Claude users to be 18, with no parental consent path, and the company has been enforcing that with age checks.
Two practical notes. If you are in the UK or most of Europe, the minimums run higher and some supervised options are not offered at all, so what you read on a US parenting blog may not apply where you live. And these policies move, more than once a year in some cases. Check the provider's own help pages before you set an account up rather than trusting a post, including this one.
Choosing what to teach gets easier when you have a plan for the whole year, not just this week.
Get the GuideA simple step-by-step plan for getting started.
What These Tools Do Reasonably Well
Explaining something a second way is the strongest use. Your child hits a wall on long division, you have explained it twice, and a third framing helps. Asking for practice problems on a specific skill works too, as does having your child talk through what they understood and getting questions back that test it. In each case your child is still doing the work, and the tool is rephrasing or prompting.
Older kids can use it as a first reader on their own writing. Ask what is unclear in this paragraph, or where the argument jumps, and the answer is often worth considering, which is different from worth accepting. Kids who push back on the response, keep what helps, and ignore the rest are practising judgement. That habit is the whole skill, and it does not appear on its own.
Where It Takes Over the Thinking
Researchers studying older students have found that heavier reliance on these tools tracks with weaker performance once the tool is removed, and they point to cognitive offloading as the mechanism, meaning the work gets handed off rather than done. The same studies found that students who interrogated and reworked the output did not show the pattern. This is research on university and secondary students rather than young kids, so hold it loosely, but the shape of it matches what parents describe.
The risky uses are the ones where struggling is the lesson. Writing a first draft, working out what you think, sitting with a hard problem long enough to find the way in: hand any of those over and your child gets a finished product plus no new capability. Reading is the sharpest case. A summary of a chapter is not a substitute for reading the chapter, and a child who learns that summaries are close enough has learned something that will cost them later.
Confident and Wrong
The failure that catches families out is not obvious nonsense. These systems produce answers that read beautifully and are sometimes false, invent book titles and citations that were never written, and misjudge reading levels badly enough to hand a seven-year-old a text meant for twelve. A wrong answer arrives in the same calm, well-organized voice as a right one, so your child has no signal to work from.
Teach the check as part of using the tool rather than as a separate lesson. If it names a book, look the book up. If it gives a date, confirm it somewhere real. If it explains a method, work one problem by hand and see whether the answer matches. Kids find this more interesting than it sounds, because catching the machine being wrong is satisfying and it teaches exactly the scepticism you want them carrying into the rest of their reading. There is more to this than one habit, and teaching kids to use AI well covers how the skill builds at different ages.
What Your Child Types Goes to a Company
Every conversation lands on somebody's servers. Providers differ on how long they keep it and whether it trains future models, and several let you turn that off in settings, so open them when you set up an account instead of leaving the defaults. This is worth one clear conversation with your child and no drama.
Give them a short rule they can remember: no full name, no address, no details about the family, nothing they would not put on a public page. Kids treat these tools as private because they feel like a conversation, and the feeling is the thing to correct.
Setting a Family Rule You Can Hold
Write down which tasks allow AI and which do not, and tie each one to the skill it builds rather than to the subject. Many families land somewhere near this: fine for explaining a concept, generating practice, checking work your child already finished, and admin like formatting. Not for first drafts, not for reading that was assigned to be read, not for anything your child will be expected to do unaided later.
Then keep an eye on the direction of travel. The question to ask yourself every few weeks is whether your child is using it to get unstuck or to get finished. If a task takes a fraction of the time it used to and your child cannot explain what they produced, you have your answer. Ask them to walk you through their reasoning now and then, out loud, with the screen closed. That takes two minutes and tells you more than any rule you can write. Your own use is a separate question with separate rules, and using AI to plan your homeschool covers what it does for the parent doing the planning.
Software is one piece of the week. The guide covers the rest: what to teach, in what order, and how to plan a year you can keep.
Get the GuideA step-by-step plan for your first weeks at home.
A Note from Homeschool Teacher Guide: What This Really Means for You
We think the loudest voices on this are both wrong. One side tells you AI will teach your child better than you can, which is a claim about software nobody has earned. The other tells you to keep it out of the house entirely, which is advice your child will route around the moment they have a phone and a question.
What we would rather you take from this: you are the teacher, and the tool is a tool. Decide what each piece of work is for, let your child use help on the parts that are not the point, and protect the parts that are. Check the age rules for yourself, because they change and nobody will tell you when they do. Then watch your own child, who will show you within a month whether this is helping them think or helping them stop.
Did this save you hours of research?
We keep these guides free and updated. If this one helped, you can say thanks.
☕ Buy us a coffee ($5)Frequently Asked Questions
How Old Does My Child Have to Be?
It differs by company, and the differences are larger than parents expect. Checked in August 2026: OpenAI sets 13 as the minimum for ChatGPT and asks that under 18s have a parent's permission, Google lets under 13s use Gemini through a supervised Family Link account a parent enables, and Anthropic requires Claude users to be 18 with no parental consent route. Minimums run higher across the UK and much of Europe. Check the provider's help pages before you set anything up.
Will Using AI Stop My Child From Learning to Think?
It depends what your child hands over. Studies of older students link heavy use to weaker performance once the tool is removed, and researchers describe it as offloading the work rather than doing it. The same studies found students who questioned and reworked the output did not show that pattern. The rule that follows is to allow it on work that is not the point of the lesson and protect the work that is.
Is It Accurate Enough for Schoolwork?
Not reliably, and the way it fails is the problem. These systems invent book titles and citations, misjudge reading levels, and deliver every answer in the same confident voice whether it is right or wrong. Your child gets no signal that something has gone astray. Build the habit of checking claims against a real source into how they use it.
Should I Let My Child Use It for Writing?
Draw the line at who produces the sentences. A child who writes a paragraph and asks what is unclear about it is still the writer. A child who asks for the paragraph is not, and no amount of editing afterwards puts that back. You have the advantage of watching the work happen, so you do not have to guess which one is going on.
What Happens to What My Child Types?
It goes to a company. Providers differ on how long they hold conversations and whether they use them to train future models, and several let you switch that off in settings. Open those settings when you create the account. Then give your child a short rule: no full name, no address, nothing about the family they would not put on a public page.
Sources
This guide was written in August 2026 from the following sources. Age policies and product features change often, so confirm current rules with the provider before setting up an account.