,

AI Homework Helpers: A Parent Guide to Safe Tools (2026)

A protected inner boundary separating a child from outward data flows

The three AI assistants your child has already heard of — ChatGPT, Gemini and Claude — all require users to be 13 or older, and none of them claims COPPA compliance for direct use by children. That single fact decides most of what follows.

It doesn’t mean AI is off-limits for homework. It means the safe pattern is parent-mediated: you hold the account, you sit alongside for younger children, and anything marketed directly at under-13s gets checked properly before it touches your child’s data. This guide is the checking framework, plus what the rules actually say in 2026.

Key takeaways

  • ChatGPT, Gemini and Claude all require 13+ and none claims COPPA compliance for direct child use. Under 13 means a parent-mediated account, not the child’s own.
  • COPPA changed on 23 June 2025. Disclosing a child’s personal information to train AI now needs separate verifiable parental consent.
  • Look for an explicit COPPA compliance statement, not a vague privacy promise. If a tool aimed at children won’t say it plainly, treat that as an answer.
  • The best tools collect no child data at all. No account, no tracking, nothing to leak.
  • The real risk isn’t privacy, it’s learned helplessness. A tool that hands over answers teaches a child to outsource thinking.
A protective boundary between a small inner area and an outer data flow
Parent-mediated access is the boundary. The child works with the tool; the tool never holds the child’s account.

What Do the Rules Actually Say in 2026?

COPPA — the US Children’s Online Privacy Protection Act — was updated by the Federal Trade Commission on 23 June 2025 to address AI specifically. The change that matters for parents: disclosing a child’s personal information for the purpose of training AI systems now requires separate verifiable parental consent, distinct from any general consent to use the service.

The rule also expanded what counts as personal information, adding categories including biometric identifiers and government-issued identifiers, with compliance on those expanded definitions required by April 2026.

What changed Effective Why it matters to you
Separate consent for AI training on child data 23 June 2025 A tool can’t bundle “use our app” with “train on your child”
Expanded personal-information definitions Compliance by April 2026 Biometric and ID data now explicitly covered
Verifiable parental consent standard Existing, reinforced A checkbox saying “I am a parent” is not verifiable consent
COPPA is US law. If you’re elsewhere, your own regime may be stricter — UK and EU rules on children’s data are not weaker.

Practical translation: if a tool aimed at children is vague about whether your child’s conversations train its models, that vagueness is itself the finding.

Are ChatGPT, Gemini and Claude Safe for Kids?

They are not designed for children, and they say so. All three set a minimum age of 13, and none of them claims COPPA compliance for direct use by a child. That’s not a scandal — they’re general-purpose adult tools — but it does mean using them with a younger child is a parent-supervised activity on a parent’s account.

Tool Minimum age Claims COPPA compliance for child use? Safe pattern
ChatGPT 13+ No Parent’s account, parent present for under-13s
Gemini 13+ No Parent’s account, parent present for under-13s
Claude 13+ No Parent’s account, parent present for under-13s
Tools marketed for children Varies Check — many say nothing Verify before use, see checklist below
Age minimums come from each vendor’s own terms. The absence of a COPPA claim is the point — read it as “not built for this”.

Note the asymmetry in that last row. A general-purpose tool being honest that it’s for 13+ is safer than a kid-branded tool that markets to eight-year-olds and stays silent on compliance.

How Do You Check a Tool Before Your Child Uses It?

Six questions, answerable in about ten minutes from the tool’s own privacy policy and terms. If you can’t find an answer, that counts as a no.

Question Good answer Walk away if
What’s the minimum age? Stated plainly and matches your child Not stated anywhere
Does it claim COPPA compliance? An explicit statement, not a hint Only vague “we care about privacy” language
Is my child’s input used to train models? No, or opt-in only with real consent Buried, conditional, or unanswered
What data does it collect? Named, minimal, and justified “Information about your use of the service”
Is an account required? No account is best Requires real name, school, photo
Can I delete everything? Clear deletion process No mechanism described
The strongest answer to question five is that the tool collects nothing — no account, no tracking, nothing to breach later.

One shortcut that works: search the tool’s site for the word “COPPA”. Tools that are compliant say so prominently because it’s a selling point to parents and schools. Tools that aren’t, don’t mention it.

What’s the Bigger Risk — Privacy or Learning?

Learning, for most families. Privacy risk is real but manageable with the checks above. The harder problem is that a tool which hands over finished answers teaches a child that thinking is optional — and that damage is invisible until an exam.

The distinction worth teaching your child is between a tool that explains and a tool that completes:

Use Effect on learning Verdict
“Explain why this answer is wrong” Builds understanding of the error Good
“Give me three practice questions like this” More reps, self-directed Good
“Explain this paragraph in simpler words” Unblocks reading comprehension Good
“Check my working and tell me where I went wrong” Feedback loop, child still did the work Good
“Write my essay on this topic” Learns nothing, likely flagged No
“Just tell me the answer” Trains outsourcing No
The rule a child can remember: ask it to help you understand, never to do it for you.

This is also the honest answer to school policy. Most schools now distinguish between AI as a tutor and AI as a ghostwriter, and the second is treated as plagiarism. Teach the distinction at home before the school has to.

Two contrasting paths, one building up in stages and one skipping to the end
Explaining builds steps. Completing skips them. Same tool, opposite outcomes.

How Should You Set It Up at Home?

Parent-owned account, shared space, and a conversation about what the tool is for. The technical setup takes five minutes; the conversation is the part that actually works.

  • Use your account, not theirs. For under-13s this isn’t optional given the age terms. For teenagers it’s still the better default because you can see the history.
  • Work in a shared room. Not surveillance — availability. Most misuse happens because a child got stuck and nobody was there.
  • Turn off training on your conversations where the tool offers it. Most major assistants have this setting; find it once.
  • Agree the rule out loud. “Ask it to explain, never to write it for you.” Children follow rules they helped phrase.
  • Review together weekly at first. Look at what they asked. You’ll learn more about their schoolwork than from any report.

And check the school’s policy before term starts. Some schools ban AI outright for certain subjects, some require disclosure, and a child who breaks a rule they didn’t know about still carries the consequence.

What Should You Never Put Into an AI Tool?

Anything that identifies your child or anyone else. This is the one hard rule, and it applies to every tool regardless of its compliance claims — because the safest data is the data you never sent.

Never paste Why
Full name, school name, class or teacher Combined, these identify a specific child
Home address or phone number Obvious, and children don’t see it that way
Photos of your child or classmates Biometric data is now explicitly covered by COPPA
Anything about another child You can’t consent on their behalf
Medical or diagnostic information Sensitive category, indefinite retention risk
Login details for school systems Never, into anything
Teach one substitution habit: describe the problem, not the person. “A Year 6 maths question about fractions” instead of names and schools.

Frequently Asked Questions

Is ChatGPT safe for kids to use for homework?

ChatGPT requires users to be 13 or older and does not claim COPPA compliance for direct child use. For younger children the safe pattern is a parent-owned account with an adult present, not the child having their own login.

What is the minimum age for AI assistants?

ChatGPT, Gemini and Claude all set a minimum of 13. None of them claims COPPA compliance for direct use by children, which is why parent-mediated access is the appropriate arrangement below that age.

How do I know if an AI tool is COPPA compliant?

Search the tool’s site for “COPPA”. Compliant tools state it explicitly because it’s a selling point to parents and schools. Vague “we value your privacy” language without a direct compliance statement should be read as a no.

Does my child’s homework train the AI?

It depends on the tool and your settings. Since the FTC’s June 2025 COPPA update, disclosing a child’s personal information for AI training requires separate verifiable parental consent. Find the training toggle in your account settings and turn it off.

Is using AI for homework cheating?

It depends on how. Asking it to explain a concept, check your working or generate practice questions is tutoring. Asking it to write the essay is plagiarism under most school policies. Teach that distinction explicitly.

What should my child never type into an AI tool?

Their full name, school, address, phone number, photos of themselves or classmates, anything about another child, medical information, or any login details. Teach them to describe the problem rather than the person.

The Bottom Line

The privacy question has a clean answer: the major AI assistants are 13+ and don’t claim COPPA compliance for children, so under-13s use a parent’s account with an adult nearby, and any kid-branded tool gets the six-question check before it’s allowed near your child’s data.

The learning question is harder and matters more. A child who asks AI to explain their mistake is better off than one without the tool. A child who asks it to write the essay is worse off than if it had never existed. That difference is a habit, and habits are taught at home.

So set the technical boundaries once, then spend your effort on the one rule that does the real work: ask it to help you understand, never to do it for you.

If you’re also weighing AI tools for your own work rather than your child’s, our comparison of AI automation tools covers the business side, and AI agents for business explains where the technology genuinely holds up in 2026 and where it still fails. For the household-admin side of AI, making money with AI automation covers realistic expectations without the hype.

Want one practical automation you can set up in 15 minutes, delivered twice a month? Join the free AI automation newsletter.

Sources

All sources retrieved 26 July 2026. Age minimums are from each vendor’s own terms of service. This is general information, not legal advice — COPPA is US law and your jurisdiction may impose stricter obligations.