Is ChatGPT safe for kids?
ChatGPT is not designed for children. OpenAI requires users to be at least 13, and 13–17s need parental consent. The risks are not mainly shocking content — they are open-ended scope, confident wrong answers, and a system built to be maximally agreeable to an adult who can push back. Younger children need a tool with fixed boundaries.
What the age rules actually say
OpenAI's terms require users to be 13 or older, and users under 18 to have parental permission. That is a terms-of-service position, not a technical control — there is no meaningful age verification, so in practice the rule is enforced by parents or not at all.
It is worth being precise about what that means. A ten-year-old using ChatGPT is not doing something dangerous by definition. They are using a product that was never designed with them in mind, whose safety work was scoped around adult users, and whose failure modes were never tested against how a young child actually behaves.
The real risks are subtler than parents expect
Most parents arrive worried about explicit content. The major systems are genuinely quite good at refusing that, and it is not where the problem sits for younger children.
- 1It is open-ended by designA general assistant will follow the conversation anywhere. That is the product. For an adult it is the whole value; for a seven-year-old it means there is no floor and no shape to the interaction, and no adult decided in advance what topics were in bounds.
- 2It is confidently wrongLanguage models state incorrect things in exactly the same tone they state correct ones. Adults have the background knowledge and scepticism to catch some of it. Children largely do not, and a fluent confident voice is very persuasive to them.
- 3It is built to be agreeableThese systems are tuned to be helpful and to go along with the user. With a child that can slide into validating things that need pushing back on — a bad plan, a distorted account of a fight with a friend, an unkind conclusion about themselves.
- 4It has no idea who it is talking toIt does not know your child is six, or anxious, or has just had a hard week, unless it is told and keeps that in context. It has no relationship and no memory of what matters to them.
- 5There is no parent viewYou cannot see what was discussed, set topic limits, or get told that something worrying came up. For a young child, that visibility is not optional.
When is a child old enough for a general AI assistant?
Around 13 and up, with supervision, ChatGPT can be genuinely useful — homework, curiosity, drafting, arguing with. Teenagers already have some scepticism and can be taught more, and the conversation about how to check what it tells you is itself worth having.
Under about 11, the honest answer is that a general assistant is the wrong shape of tool. Not because it will say something terrible, but because a young child cannot supply the judgement the product assumes the user has.
Common Sense Media rates and reviews AI tools for family use and publishes age guidance for individual products — a useful second opinion before handing anything to a child.
— Common Sense Media
What to look for in a kids' AI instead
- A fixed character with a fixed scope, rather than an assistant that will go anywhere the conversation leads.
- Hard topic boundaries decided by adults in advance — and sensible redirection of big life questions back to you.
- A parent view: what was talked about, what skills are developing, what to set limits on.
- A clear business model. If it is free and ad-supported, your child's attention is what is being sold.
- An explicit statement on voice and conversation data — what is kept, and whether it trains anything.
- Age-appropriate pacing. Talking with a five-year-old means tolerating long pauses and half-finished sentences, which general systems handle poorly.
Where Wizzly fits
Wizzly is the narrow version of this. It is one character — a slightly ridiculous wizard — with a fixed personality, a developmental curriculum underneath, and hard content walls that do not move no matter how the question is phrased. Off-limits topics stay off-limits; big life questions get handed back to you.
It is voice-first and screen-free, built for ages 3–15, with a parent dashboard for topics, time and progress. There are no ads and never will be — it is subscription-funded, and free during early access.
It is not a replacement for a general assistant when your teenager needs to argue with something clever. It is what you give a six-year-old instead.
Wizzly is completely free during early access — every story, no card required. Runs in the browser — no download.
Common questions
What is the ChatGPT age limit?
OpenAI requires users to be at least 13, and users under 18 to have parental permission. There is no real age verification, so enforcement in practice falls to parents.
Can a 7-year-old use ChatGPT?
It is below the terms-of-service age of 13, and more importantly it is the wrong shape of tool for that age — open-ended scope, confident errors and no parent visibility. A child-specific product with fixed boundaries is a better fit.
What is a safe AI alternative to ChatGPT for kids?
Look for a fixed character rather than an open assistant, adult-set topic boundaries, a parent dashboard and a subscription rather than ad funding. Wizzly is built to that shape for ages 3–15, is screen-free, and is free during early access.
Is AI bad for children generally?
No. The question is whether a specific tool was designed for children. AI that is scoped, bounded and supervised can be genuinely good for practising conversation and emotional vocabulary. AI built for adults and handed to a young child is the risky combination.