ChatGPT for Kids: Rules That Actually Work

Most advice on this lands in one of two places. Either ban it outright, or shrug and let a child use it like any other app. Neither is much use to a parent deciding what happens the next time homework is open in one tab and ChatGPT is open in the other.

What follows is the practical middle: the age rules the tools actually set, what parental controls do and do not cover, and the one habit that matters more than any rule you put in place. Most of it is already written down in documents the companies publish themselves.

Is ChatGPT actually allowed for a child under 13?

No. OpenAI's own help center states plainly that "ChatGPT is not meant for children under 13," and that children aged 13 to 18 need parental consent before using it.

The OpenAI Terms of Use say the same thing as a condition of the account: "You must be at least 13 years old or the minimum age required in your country to consent to use the Services. If you are under 18 you must have your parent or legal guardian's permission to use the Services."

The help center adds a detail parents rarely see reported. "If you are using ChatGPT in the education context for children under 13, the actual interaction with ChatGPT must be conducted by an adult," according to OpenAI's article on age requirements. The same page advises "caution with exposure to kids, even those who meet our age requirements."

That last line is worth sitting with. The company that built the tool recommends caution for children who are old enough to use it, which is a more honest starting point than anything a coding school is going to tell you.

Why does every AI tool set a different age rule?

Because the companies have not agreed on one. The rules come from each company's own legal and product decisions rather than any shared standard, and they land in three different places.

ChatGPT sets 13 as the floor. Google allows a child under 13 to reach Gemini, but only through a supervised account: "A parent must enable access before their child under 13 (or the applicable age in your country) can use Gemini Apps with a supervised account," according to Google's guidance for parents. Anthropic goes the other way entirely, stating that "we require all users to be at least 18 years old to create and use a Claude account," per Claude's minimum age policy.

Thirteen, under thirteen with a parent, and eighteen. Three companies, three answers, for products a child would struggle to tell apart.

None of this is settled science about the age at which a child can handle a chatbot. It is three risk appetites, which is exactly why a parent has to make the actual call rather than borrowing one company's age gate as a verdict.

Do those age rules actually stop anyone?

No, and it is better to know that than to assume otherwise. There is no meaningful age check at sign-up, so the rule functions as a statement of intent rather than a gate.

This changes what you are actually deciding. You are not choosing whether your thirteen-year-old could use ChatGPT, because they can, today, on any device with a browser. You are choosing whether they use it in the open, with you aware of it, or quietly.

That is the practical argument against a flat ban in a house with a teenager. A ban you cannot enforce mostly buys the illusion of one, and it costs you the chance to teach the part that actually transfers.

What is ChatGPT actually good for in a child's learning?

Explaining a concept a different way when the first explanation did not land, and working through a practice problem one step at a time rather than handing over the answer. That is the genuine use, not a replacement for learning the material.

OpenAI's own description of its teen product draws the same line parents should be drawing at home. "A teen studying for a test, for example, can ask ChatGPT to explain an unfamiliar concept, work through a practice problem, and check their understanding," according to OpenAI's announcement of ChatGPT for Teens. The same page describes a feature built to catch the opposite behavior, one that can "recognize when a teen appears to be trying to shortcut an assignment and redirect them toward Study Mode for step-by-step collaborative problem solving."

Read that as a signal rather than a feature list. OpenAI built a mechanism to detect assignment shortcutting because shortcutting is the default behavior, not the exception. That is the real reason it matters at your kitchen table too.

What is the one habit that matters more than any rule?

Checking its work. A child who treats every answer as a first draft to verify is reasonably safe with almost any tool, and a child who treats every answer as finished is not, whatever the age rule says.

Here is the part that lands differently with a teenager, because it is not a parent's opinion. OpenAI's own terms require it of every user. Under Accuracy, they state: "Output may not always be accurate. You should not rely on Output from our Services as a sole source of truth or factual information, or as a substitute for professional advice." The following line is an obligation rather than a warning: "You must evaluate Output for accuracy and appropriateness for your use case, including using human review as appropriate, before using or sharing Output from the Services."

Verifying the answer is not you being suspicious of a tool your child likes. It is a condition of use they agreed to, written by the company that built it.

Common Sense Media points at why this is mechanically hard, noting in its guide to ChatGPT for parents that the tool "does not give sources for its information." A child cannot check a claim against a source they were never shown, so the habit has to be active rather than passive.

The version that works at home is a question, not a lecture. When your child shows you something it produced, ask how they know it is right. If the answer is that it sounded right, that is the teaching moment, and it will come around again within the week.

What does the research say about kids offloading the thinking?

That the concern is real and comes from the people building these tools. Anthropic's researchers, studying how students actually use Claude, wrote that "there are legitimate worries that AI systems may provide a crutch for students, stifling the development of foundational skills needed to support higher-order thinking."

The same analysis found the behavior in the data, not just in theory. Students "were asking Claude for direct answers almost half the time with minimal back-and-forth engagement," according to The Hechinger Report's coverage of the research.

Two honest caveats. This studied university students rather than children, and the reporting notes that Anthropic was hesitant to claim direct evidence of cheating, since it could not tell whether a request for multiple-choice answers involved a real exam or practice. So treat it as a description of how people use these tools rather than proof of harm to a specific child.

Even with those caveats it points somewhere useful. The risk is not the age of the account holder. It is the habit of asking for the answer instead of the explanation, and that habit forms at any age.

What does checking its work look like at different ages?

For a younger child, a parent reads the answer alongside them and asks where it might be wrong before anyone treats it as true. For a teenager, it means cross-checking against the textbook, a teacher, or a second source before it goes on the page.

Common Sense Media draws its lines by developmental stage rather than by product policy. Its chatbot guidance by age recommends "No AI chatbots" for children five and under, "Limited use, alongside a parent or caregiver" for ages 6 to 12, and from 13 up, "Supervised use is OK for homework and creativity, not for comfort or feelings."

That last distinction is the sharpest thing on this page. Homework and creativity are reasonable uses for a supervised teenager. Comfort and feelings are not, and a parent who only polices the homework half has missed the part that matters more.

Do the parental controls do what parents think they do?

Mostly not. ChatGPT has real parental controls now, and the important thing to understand is what they deliberately exclude.

Linking your account to your teenager's lets you manage certain settings, schedule quiet hours when ChatGPT cannot be used, and receive safety notifications in limited situations. OpenAI's page on managing parental controls confirms the settings side holds: "Your teen cannot change settings managed by a parent or guardian while your accounts are linked."

But the same page is direct about the boundary: "Parental controls do not let parents or guardians read or monitor a teen's conversations."

Absorb that before deciding this problem is solved. You can shape when and roughly how your child uses it. You cannot read what they said, and you should plan on never being able to. Which returns you to the only thing that scales, which is what your child does when nobody is checking.

The schedules are still worth setting. Study hours can make eligible new conversations start in study mode without blocking access, so the tool behaves more like a tutor and less like an answer machine during homework time.

Should AI be part of how a child learns to code?

Yes, once a child already understands what code does, because that is what turns typing a prompt into reading what came back. A child who can write a loop can look at a model's output and ask whether the logic holds rather than accepting it because it sounds confident.

This is why our Senior Track teaches AI alongside Python rather than on its own. A teenager who can already write code can understand what a model is doing rather than just typing prompts at it, which is much closer to the checking habit above than to a coding shortcut.

If your child is curious about what is happening underneath rather than just how to use it, our explainer on how AI actually works covers learning from examples rather than rules, and our wider parents' guide to AI covers the literacy side beyond this one tool.

What is the practical rule to set at home?

Treat the age gate as a floor rather than an answer, and make "did you check that" the question you ask every time rather than only when something looks wrong. The account rules tell you when a tool is allowed in your house. They tell you nothing about whether it is being used well once it is there.

Three rules cover most of what goes wrong. Under 13, an adult does the typing. Over 13, the account is known about and the parental controls are linked. At every age, the answer gets checked before it is used or handed in.

The thing not to do is treat this as settled once the rules are set. The tools change every few months, your school's policy will change, and your child's use will outgrow whatever rule you wrote. A conversation twice a year beats a policy set once.

If you would rather your child built the checking habit alongside someone who does it with them every week, a free trial class is one live session with a mentor, no card details and no commitment. Ask them to bring something ChatGPT got wrong, because there will be one.