How do I know if my child is using AI for homework?

Ask directly, and ask without a penalty attached the first time — you're gathering information, not running a trial. Then check the work itself: ask them to explain one paragraph or one solved problem out loud, with the device closed. A kid who did the thinking can reconstruct it. A kid who pasted a result usually can't get past the first sentence.

How common is this, really?

In a Pew Research Center survey of 1,458 U.S. teens ages 13–17 (fielded September 25–October 9, 2025, released February 2026), 54% said they had used chatbots for help with schoolwork. Ten percent said they use them for all or most of their schoolwork. Meanwhile 64% of teens said they use chatbots at all, compared with 51% of parents who said their teen does — with another 29% of parents unsure.

Should I just ban AI for schoolwork?

A flat ban is hard to enforce and tends to push the use underground, where you lose the ability to shape it. Most schools have landed on a middle position — AI is allowed for brainstorming, explanation, and practice, but not for producing the work that gets submitted. A household rule that mirrors that is easier to hold to and easier for your kid to apply when you're not in the room.

What's a rule that actually works?

The explain-it-back rule: you can use AI to understand something, but you have to be able to explain whatever you turn in, in your own words, without it open. It's one sentence, it maps to what teachers are checking anyway, and unlike a screen-time limit it's something you can verify in ninety seconds at the kitchen table.

Does this apply to younger kids too?

The Pew numbers cover teens 13–17, so treat them as a picture of middle and high school specifically. But the habit forms earlier than the data does. If your fourth or fifth grader has a school-issued device, the same rule is worth setting now, while the assignments are still small enough that the conversation is low-stakes.

You Probably Don't Know How Much AI Your Kid Is Using. September Is When to Find Out.

About 64% of teens say they use AI chatbots while only 51% of parents say their teen does — and another 29% of parents aren't sure — so the most useful thing you can do in the first weeks of school isn't banning AI or pretending it isn't there, but setting one rule you can actually check: your kid has to be able to explain anything they hand in, in their own words, with the chatbot closed.

About 64% of teens say they use AI chatbots. Only 51% of parents say their teen does — and another 29% aren’t sure. So the most useful thing you can do in the first weeks of school isn’t banning AI or pretending it isn’t there. It’s setting one rule you can actually check: your kid has to be able to explain anything they hand in, in their own words, with the chatbot closed.

The gap is bigger than most of us think

In a Pew Research Center survey of 1,458 U.S. teens ages 13–17 and their parents, fielded in the fall of 2025 and released in February 2026, 64% of teens said they use AI chatbots. 51% of parents said their teen does. Another 29% of parents said they weren’t sure — which is arguably the more interesting number, because it’s not a disagreement, it’s a blind spot.

On schoolwork specifically, 54% of teens said they’d used chatbots to get help with it. Ten percent said they use them for all or most of their schoolwork; about a fifth said some; roughly another quarter said a little. More than four in ten had used one for research or math problems, and about a third for editing their writing.

And just over half of parents said they had ever talked to their teen about chatbots at all. Which means a large group of families is a year or two into this with no conversation on the record.

None of that makes your kid a cheater. Most of that 54% is a kid stuck on problem seven at 9:40pm doing the obvious thing. But it does mean that if your working assumption is “mine probably doesn’t,” the odds are against you.

”Is this cheating?” is the wrong first question

It’s the question that comes to mind first, and it’s a dead end — because it makes the conversation a verdict, and your kid will answer accordingly.

The better frame is about what got built. Two kids can use the same chatbot on the same assignment and end up in completely different places. One asks it to explain why the negative sign flips, tries the next problem herself, gets it wrong, asks again. The other pastes in the problem set. Same tool, same twenty minutes, and only one of them can do it again on Thursday without help.

That’s the actual line, and it’s the one worth teaching, because it’s the one that still applies when they’re 22 and no one’s checking. We wrote more about where that line falls in practice if you want the longer version.

Most schools have converged on roughly this position too — AI is generally fine for brainstorming, explanation, and practice, and not fine for generating the thing that gets submitted. A house rule that matches the school’s is far easier to hold than one that fights it.

The rule that holds up: explain it back

Here it is, in one sentence:

You can use AI to understand something. You have to be able to explain whatever you turn in, in your own words, with it closed.

Why this one and not a screen-time cap or an app block:

It’s checkable in ninety seconds. “Walk me through this paragraph.” You’ll know immediately.
It’s the same thing the teacher is checking. You’re not inventing a private standard your kid has to code-switch between.
It survives new tools. It’ll still make sense when the app they’re using is one neither of us has heard of yet.
It’s not a punishment, it’s a bar. Kids argue with restrictions. They argue much less with a standard they can clearly meet.

Apply it to work that came back with a good grade, too — not just the assignments that went badly. That’s where the surprises live, as we found when we looked at why solid grades don’t always mean grade-level understanding.

What to say in the first week

Keep it short and make the first pass amnesty. You want accurate information more than you want a confession.

“What are you allowed to use this year, and who told you?” Start with the school’s policy. It’s neutral ground, and half the time they don’t actually know — which is useful in itself.
“Show me how you’d use it on tonight’s homework.” Watching beats asking. You’ll learn more in five minutes of observation than in a week of questions.
“Here’s the one rule at our house.” Say the explain-it-back rule out loud. Once. Don’t stack five more on top of it — a single rule they remember beats a policy they tune out.

If you have a teenager and this conversation has a history, talking to your teen about AI and schoolwork goes deeper on the tone that keeps it from turning into an argument.

When the tool itself holds the line

The trouble with the explain-it-back rule is that it puts you in the position of auditor — checking after the fact, at night, when everyone’s tired.

That’s the gap TutorBuddi is built for. Buddi is deliberately designed not to hand over answers: it works Socratically, asking your child questions until they get there themselves, and it holds that line under pressure — it won’t produce a finished essay on request, and redirects to outlining, argument questions, and feedback on your kid’s own draft instead. Your child can’t paste in a problem set and get a completed one back, so “did you do this yourself” stops being a question you have to ask.

And because Buddi remembers across sessions, the parent dashboard shows you what your child worked on, where they got stuck, and what finally clicked — the visibility the explain-it-back rule was trying to get you, without you being the one running the check every night.

It’s free to try for one child — no card — and it covers grades 3–12.

The point of any of this isn’t to catch your kid. It’s to make sure that in June, the work they turned in and the things they can actually do are the same list.