If you're waiting for someone to hand you the right age for kids to start using AI, I'll save you some time: there isn't one. Anyone who gives you a single number is guessing. I'm a parent, not a child psychologist, so this leans on people who study it properly.
There are really two questions in here. Is it safe, and what does it do to the way they learn? This post covers both. I've gone deeper on the second in a companion post, The 10,000 Steps Problem.
It's also a slightly out-of-date question. When Australia's eSafety Commissioner surveyed 1,950 kids aged 10 to 17, 79% said they had already used an AI assistant or AI companion. [1] For a lot of families, the question isn't when to start. It's how to make sure what's already happening is going okay.
What does AI do to the way kids learn?
Adults who use AI well usually already have the basics, so they can tell when it's wrong and steer it. Kids are still building those basics. Hand them an answer machine too early and it can hide the gap.
There's early evidence for this. In a randomised trial at a Turkish high school, students with a standard ChatGPT-style assistant scored 48% higher in practice and then 17% lower in an exam without it. Students with a version set up to give hints instead of answers did about the same as students with no AI. [2] It was one school and one subject, and nobody has long-term evidence on young children. But it points to a simple rule that holds up while we wait: do it yourself first, then use AI.
Is there a right age for kids to use AI?
Not a single one, no. Think about swimming. Nobody says "kids can swim at seven." You start in the shallow end with your hands on them. Then you loosen your grip but stay within arm's reach. Later they're swimming laps while you watch from the side. And they never learn to swim by being carried.
AI works the same way. There are two questions to ask. How close should I be standing? And what should they be able to do on their own first?
What should it look like at each stage?
Here's a rough version. The ages are a guide, not a rule, and kids move through at different speeds.
Stage 1: You use it, they watch (the youngest kids). Common Sense Media recommends no AI chatbots for kids five and under, because young children can't fully tell AI apart from a real person. [3] At this age, the foundations are reading together, talking, and play. If AI comes into it, it's on your side of the table: a bedtime story idea, or a "why is the sky blue" argument settled while they're the audience.
Stage 2: Together, at the same screen (roughly primary school). For ages 6 to 12, Common Sense's guidance is limited use, alongside a parent. [3] Sit next to them. Have them write the story first, then ask the AI what it would add. Work out the maths themselves, then ask it to check. Ask it weird questions together, and get it to say something wrong so they can catch it. That lesson will stick better than any lecture.
Stage 3: Their turn, but out in the open (the early teens). Common Sense says supervised use is okay for homework and creativity, but not for comfort or feelings. [3] This is where the trial matters most. Show them how to ask for hints and to be quizzed instead of asking for answers. Keep it in shared spaces, and keep asking the question Common Sense suggests: "Is this AI helping me learn, or just helping me finish?"
Stage 4: Mostly on their own (the later teens). Fewer rules, more conversation. They have their own judgement by now, and the test is whether they can explain what they hand in. Your job shifts from supervising to staying someone they'd tell if something felt off.
How do you know when they're ready to move up a stage?
Use readiness instead of a birthday. Four quick checks:
- Do they understand it's a machine, not a person?
- Can they spot when it's wrong, or at least know to check?
- Can they do the task without it first?
- Would they tell you if something it said made them uncomfortable?
If the answer is yes to all four, loosen your grip a little. If not, stay where you are. That's completely fine, and nothing says everyone has to move at the same pace.
What do the AI tools themselves allow?
It's worth knowing, but it shouldn't be the thing you rely on. The big tools set their own minimum ages and they don't agree with each other. Some say 18 and over. Some allow teens with parental controls. Google has announced a supervised version of Gemini for under-13s, managed through Family Link. [4] These rules change, so check the current terms for whatever your kid is using.
And an age rule is only as good as the checking behind it. When eSafety looked at four AI companion apps in 2026, its Commissioner said none had meaningful age checks. [1] Australia's age-restricted material codes, which took effect on 9 March 2026, now require chatbots that can produce explicit or harmful content to check users' age before showing restricted material. [5] That's a good step, but I wouldn't hand the job to the app.
What stays the same at every stage?
Whichever stage you're at, three things don't move:
- It isn't a friend, a therapist, or a source of comfort. Common Sense's advice for everyone under 18 is no AI for emotional support, mental health advice, or "friendship." It's built to agree with you, which can reinforce unhealthy thoughts instead of pushing back. [3] This is the one category I'd steer clear of entirely, and it's where the AI companion apps sit.
- It can be wrong. Treat anything important as something to check, whatever the age.
- Personal details stay out. Full name, school, address and photos don't belong in a chat.
Where this leaves you
The goal isn't to keep kids away from AI forever, and it isn't to throw them in the deep end either. They're going to grow up with it. The goal is for them to reach the deep end already knowing how to swim, and knowing you're still standing on the side if they need you.
So start close, loosen up as they show they're ready, and keep those three rules constant the whole way. Let readiness decide, not the birthday.
Sources
- eSafety Commissioner, eSafety report shows AI companions are putting children at risk. Survey of 1,950 children aged 10 to 17.
- PsyPost, Unrestricted generative AI harms high school math learning by acting as a crutch, summarising "Generative AI without guardrails can harm learning: Evidence from high school mathematics" (PNAS).
- Common Sense Media, AI Chatbots and Your Child: Age and Stage Guidance.
- NBC News, ChatGPT rolls out new parental controls (teen accounts, ages 13 to 17); YourStory, Google's Gemini AI now accessible to kids under 13 via Family Link.
- CADE Project, Australia expands age verification rules to AI chatbots, app stores and online platforms.
alltools.solutions builds free and paid tools, sometimes in the same space as what's discussed here. Nobody mentioned in this post paid for or reviewed it, and nobody pays us to mention them. We hold our own tools to the same rubric we'd apply to anyone else's.