Is ChatGPT safe for kids?

The honest answer is: it depends on what you mean by safe, and it depends on your kid. We know that's not the yes-or-no you searched for. But you've probably already noticed that most articles promising a straight answer to this question are trying to scare you or sell you something, and neither is much help at 10pm when you've just found ChatGPT open in your nine-year-old's browser tabs.

So here's the real picture, as plainly as we can put it. (Full disclosure: we make an AI product for kids, and there's a note about it at the bottom of this page. Everything above that note is meant to be useful whether or not you ever click on it.)

What the rules actually say

ChatGPT's terms require users to be at least 13, and anyone under 18 to have a parent's permission. In practice there is no meaningful age verification — a child ticks a box, or just uses the version that requires no account at all. OpenAI has been rolling out parental controls that let you link your account to a teen's, which is genuine progress, but they're recent, they're opt-in, and they don't apply when a kid uses the tool logged out or on a friend's account.

So the rules give you a floor, not an answer. If your child is under 13 and using ChatGPT, they're outside what the product is built or reviewed for. That's worth taking seriously — not as a panic, but as a fact.

What ChatGPT is genuinely good at

It would be dishonest to pretend the appeal is a mystery. ChatGPT is endlessly patient. It never sighs, never says “we already covered this,” never makes a kid feel slow for asking the same question four different ways. For a curious child, that's intoxicating — a grown-up who always has time. Plenty of kids have learned real things from it, and if your child is one of them, that curiosity is worth protecting, not punishing.

The risks that make headlines — and the ones that should

The headline fears are explicit content and dangerous instructions. Those filters aren't perfect, and determined teenagers trade tricks for getting around them, but for ordinary use they catch most of the obvious stuff. If that were the whole story, this would be a shorter page.

The quieter risks are the ones we'd actually lose sleep over:

• It doesn't know it's talking to a child. Unless told — and even when told, it drifts — ChatGPT talks to your 9-year-old the way it talks to a 40-year-old. The vocabulary, the examples, the assumptions about what the reader can handle: all adult-shaped.
• It is confidently wrong. It makes things up in a calm, authoritative voice. Adults find this annoying. Children, who are still learning that confident adults can be wrong, find it convincing.
• It hands over answers. Ask it to write the book report and it writes the book report. No judgment, no friction, no “shouldn't you try first?” The path of least resistance is always fully paved.
• It can become a confidant. An always-available, always-agreeable listener is a powerful thing for a lonely kid. Some of that is fine. Some of it quietly replaces the harder, better work of talking to actual people.
• You can't see any of it. The conversations live inside an account, behind a login. There's no parent view. Whatever your child is asking it — about school, their body, their friends, you — happens somewhere you have no window into.

If your kid already uses it

Resist the reflex to ban it on the spot. A ban teaches one lesson — use it where I can't see — and kids learn that lesson fast. You'll have traded visibility for the feeling of having acted.

Instead, sit down next to them and ask for a tour. Kids are usually delighted to show off what they've figured out, and ten minutes of watching them use it will tell you more than any article, including this one.

“Show me the coolest thing you've ever gotten it to do.”

Then set up something workable. For younger kids: a shared account on a family device, used in the kitchen or living room, with the understanding that you'll read the history — openly, not secretly. For teens: link accounts with parental controls where available, and keep the conversation going instead of assuming the settings will do the parenting. None of this is airtight. Chat histories can be deleted; incognito windows exist. The setup isn't the protection — the relationship is. The setup just keeps the relationship honest.

The bottom line

ChatGPT is a powerful general-purpose tool that was not designed for children, used by millions of children. For a teenager with good habits and a parent who stays curious, it can be fine — even great. For younger kids, “a very smart adult stranger with infinite patience and no supervision” is exactly as mixed a proposition as it sounds. You know your kid. Now you know the tool.