There is no age that makes a child ready for AI, and the guidance that exists is a floor rather than an answer. Common Sense Media says none under 5, supervised use from 6 to 12, and from 13 for schoolwork but never for comfort. Your child's readiness is a set of skills you can check tonight. Platform rules and your school's rules still come first.
What do the official guidelines actually say?
Three different bodies have put a number on this, and they are answering three different questions.
Common Sense Media is the closest thing to a standard for families. Its age and stage guidance for AI chatbots, last updated on June 8, 2026, sets out four positions. For ages 5 and under: "No AI chatbots, young children can't fully tell AI apart from real people." For 6 to 12: "Limited use, alongside a parent or caregiver." For 13 and over: "Supervised use is OK for homework and creativity, not for comfort or feelings." And for everyone under 18, no AI for emotional support, mental health advice or friendship, because, as they put it, "AI has no medical training."
UNESCO put the classroom number at 13, in guidance released on September 7, 2023 that set "an age limit of 13 for their use in the classroom". That guidance is three years old now, which in this field is a long time, and it was written for governments rather than for kitchen tables. It is still the only global figure anyone cites.
The companies set their own ages, and they are all different. OpenAI's help center, re-read on September 29, 2026, says that "ChatGPT is not meant for children under 13" and that it requires 13 to 18 year olds to get parental consent. Claude is 18. Gemini allows under-13s on a supervised account a parent switches on. We went through what each company actually requires, and what those numbers buy you, because the numbers differ from company to company and most children use more than one of these apps.
Here is the part worth holding on to. The company numbers come from privacy law and liability, not from child development. UNESCO's number was a policy recommendation to education ministries. Common Sense's bands are the only ones written about children, and even those describe a population rather than your child. None of the three has met the person you are deciding about.
Most children have already started
This is the finding that changed how I answer this question, and it is uncomfortable.
The Common Sense Media census of AI use by tweens and teens, a survey of 1,204 US children aged 9 to 17 with fieldwork from March 18 to 26, 2026 and published on June 8, 2026, found that 86 percent of them use AI. Broken down by age, that is 81 percent of 9 to 12 year olds, 89 percent of 13 to 15 year olds and 92 percent of 16 and 17 year olds.
Read the first of those numbers again. Four in five American children aged 9 to 12 are already using AI, which is the band Common Sense says should only use it alongside a parent.
The same survey found that 44 percent of children aged 9 to 17 say a parent or guardian has never talked with them about how to use AI safely. Among the 9 to 12 year olds, it was 47 percent.
Parents are also working with an incomplete picture. Pew Research Center's study of 1,458 US teens and their parents, with fieldwork from September 25 to October 9, 2025 and published on February 24, 2026, found that 51 percent of parents said their teen uses chatbots while 64 percent of teens said they do. About three in ten parents were not sure either way.
That gap is not a finding about dishonest teenagers. It is a finding about a technology that arrived inside apps children already had, without an installation moment a parent could notice.
My reading of it, and this is my reading rather than anyone's finding: for most families the question is not whether to start. It is whether anyone is going to talk about it now that it has started.
The Mistake I See Parents Make
Parents come to this looking for a number, when what decides it is a handful of things their child can or cannot yet do.
I have stood at a lot of parents' evenings, first with a class of my own and later as the person parents were sent to, and I train teachers on these same conversations now. The question arrives in the same shape almost every time: what age is it allowed, as though somewhere there is a form that gets signed. It is a reasonable instinct. It is how we handle films, phones, and being left in the house alone.
It does not transfer, for a simple reason. A film has a fixed content and a fixed running time, so an age rating can be about the film. A chatbot has no fixed content. What it produces depends entirely on what your child types into it, which means the rating has to be about the child.
The pattern I see most often in schools is a year group where the same tool is genuinely useful to one child and quietly harmful to the one sitting next to them, in the same lesson, on the same task. The difference is never the birthday. It is whether the child can tell when the answer in front of them is wrong.
Readiness is not a birthday. It is whether your child can tell you what they are not sure about, and whether they would tell you if something on the screen frightened them. You can check both tonight, and neither answer is in a terms of service page.
The Ready Test: three signs, whatever the age
So here is what I would actually look for, and you can check all of it this week without announcing that you are checking anything.
The Ready Test is the three things I would want to see before a child uses AI without an adult in the room: they can tell you what they are not sure about, they can be wrong out loud without it stinging, and they would come and tell you if something on the screen frightened them. Not one of those is about age. All three are about the child in front of you, and you can check all three at the kitchen table tonight.
Sign one: they can tell you what they are not sure about. A child who knows the edge of their own knowledge can catch a machine that does not. A child who cannot will take a confident answer at face value, because it sounds the way a right answer sounds. The same Common Sense census found that only about a third of children knew AI cannot reliably tell fact from fiction. To check it, ask your child something just past what they know and watch whether they say so, or fill the gap. Filling the gap is normal at 7 and worth working on at 12. If you want a way into it, the one sentence that explains AI from 5 to 15 is where I would start.
Sign two: they can be wrong out loud without it stinging. This one sounds soft and is the most practical of the three. Checking an AI answer means being willing to find out you were wrong, in front of someone. A child who finds that unbearable will not check, and will defend the answer instead. You already know where your child sits on this. It shows up over board games and spelling tests long before it shows up over a chatbot.
Sign three: they would come and tell you. Children do meet things on these apps that unsettle them, and the census found that of those who had, only about a third said anything to a trusted adult. That silence is the real exposure, and no filter reaches it. What reaches it is whether your child expects the conversation to go well.
Two signs out of three is not a failure. It tells you which one to work on, and it usually tells you to sit beside them for a while longer rather than to say no.
What changes at each age
The signs matter more than the numbers, but the numbers still shape what use looks like.
Under 8. No independent account, on anybody's guidance including mine. At this age a child cannot reliably tell a machine from a person, which is exactly why Common Sense draws its hardest line here. If AI appears at all, it is something you open together on your own phone for a few minutes and then put away. The Forbes piece by Kyle Smitley on September 24, 2026 that argued elementary schools should pause student-facing AI altogether is making a stronger version of the same point, and for this age band I think it is largely right.
8 to 12. The band where the decision is usually made by default, because most of what their friends use says 13. This is where the three signs earn their keep, and where supervised use means genuinely sitting there rather than being in the same room. Our guide to the years before a phone sets out the options in more detail.
13 to 15. They now meet most platform rules, and they get the teen version of these products, which is filtered rather than supervised. Those are different things. This is the age where what teenagers actually use AI for tends to surprise parents, and where sign three matters more than the other two put together.
16 to 18. The age question has stopped being useful. What matters now is whether they can catch a wrong answer that is well written, and whether you are still someone they bring things to. If AI has become the place they take their feelings, that is a different conversation, and we have written it up separately.
What to say to your child
Short, and not straight after a row about screens.
You could say: "What did it get wrong last time? There must have been something."
You could say: "How would you know if it made that up?"
You could say: "If it ever says something that makes you feel weird, come and tell me. You won't lose the app for it."
That last promise is the one to think about before you make it, because a child tests it once. If telling you costs them the tool, you have taught them to stop telling you, and sign three was the one worth having.
What to ask the school
Your child may be using AI at school at an age you would not have chosen at home, lawfully, because the school agreed to it on your behalf. That is normal, and it is worth asking about plainly.
- At what age or year group does AI use start here, and with which tools?
- Are children taught to check what it produces, and how often does that happen in a lesson?
- What were staff given by way of training before this was switched on?
The third question is the one that changes the answer, and it is often the one nobody has asked. Where a school cannot answer the third one, it is almost never because nobody cares. It is because staff were handed the tools before anyone trained them, which is the work I do with schools and a fair thing to send to a head of year.
What none of this can guarantee
The Ready Test is a judgment aid, not a safety system. A child who passes all three signs can still meet something upsetting, because no filter on any of these products catches everything, and the companies say so themselves. No parental control on the major consumer AI tools lets you read your child's conversations. Nothing here overrides a platform's stated age limit or your school's own rules, both of which take precedence over anything on this page.
And if your worry is about your child rather than about a tool, that is the other conversation. Common Sense is blunt that AI chatbots miss the warning signs of conditions like depression, anxiety and eating disorders, and no readiness test substitutes for a professional. Your family doctor or pediatrician and the school's safeguarding lead or counselor are the right first calls, and they are the right first calls before any of the advice above.
The number you were looking for does not exist. What you have instead is three things you can check tonight, and a child who will tell you more than any setting will.
If you want this sort of thing explained without the panic, I write a Sunday newsletter for teachers and parents. Parents are very welcome.
Dan Fitzpatrick is a former secondary teacher, assistant headteacher and Director of Digital Strategy who now helps schools and families make sense of AI. More about Dan.
