Judgment

Should you let your kid use AI for homework?

By the Keep Your Head desk · About a 6-minute read

Wrong question. The right one isn't "did they use AI" — it's "did they learn." A kid can use AI for two hours and walk away smarter, or paste one prompt and walk away with nothing, and from the outside the homework looks identical. Here's a way to actually tell the difference, and a rule of thumb you can hand your kid instead of hovering over their shoulder.

The number that matters here

This isn't a hypothetical you can wait out. Pew Research Center surveyed U.S. teens in the fall of 2024 and found that 26% of teens aged 13–17 had used ChatGPT for schoolwork — double the 13% who said the same in an earlier poll from 2023. Whatever your household rule is, it's competing with a habit that's already doubled in a year and shows no sign of slowing. A flat ban you can't enforce isn't a policy; it's a wish.

That's not an argument for handing over the keys either. It's an argument for a rule specific enough that your kid can actually follow it when you're not in the room — which is most of the time homework happens.

The test that actually works: could they redo it without the tool?

Forget "AI" versus "no AI" as the dividing line. Use this instead: if you took the answer away right now, could they reconstruct the reasoning themselves? That single question sorts almost every homework use into "fine" or "not fine" without needing a subject-by-subject rulebook.

Used it to understand, can reconstruct: asked AI to explain a concept a different way after the textbook explanation didn't land, then solved the actual problems themselves. This is what a good tutor does — except this one's available at 9pm on a Tuesday.
Used it to produce, can't reconstruct: pasted the assignment prompt, got an essay or a worked solution back, copied it in. No reconstruction possible, because there was nothing to reconstruct — the thinking never happened on their end.

The gray middle — used it to outline, then wrote the essay themselves; used it to check an answer after solving it independently — usually leans fine, because the effortful part still happened. The test isn't "purity." It's whether the struggle that builds the skill got skipped.

A traffic-light system you can actually hand your kid

Vague rules ("use it responsibly") don't survive contact with a deadline at 11pm. Specific ones do:

GREEN — use freely
- "Explain this a different way" when the textbook
  explanation didn't click
- Checking your own finished answer for mistakes
- Generating practice problems to test yourself
- Brainstorming essay topics or angles (then you write it)

YELLOW — use, but do the next step yourself
- Getting an outline or structure, then writing every
  sentence yourself
- Asking it to quiz you on material before a test

RED — don't submit this as your own work
- Having it write paragraphs or the essay itself
- Having it solve the exact assigned problems for you
  to copy
- Anything your teacher's syllabus already says is
  off-limits — that rule wins regardless of this list

Put this where your kid actually sees it — not a lecture once, a list taped somewhere near where homework happens. Specific beats sincere every time a deadline is involved.

Why "they'll get caught" is the wrong deterrent

A lot of parents lean on detection as the backstop: "the teacher will know." Don't build the rule on that foundation — it's shakier than it sounds. AI-detection tools have a well-documented false-positive problem; OpenAI built its own detector, then shut it down after finding it mislabeled human writing — including Shakespeare — as AI-generated. We cover this in detail in can teachers tell if you used AI?, and the short version matters here too: a detector being wrong in either direction is a bad reason to make a choice. Teach the actual standard — did you do the thinking — not "don't get caught," which teaches your kid to optimize for undetectability instead of honesty the moment you're not checking.

Where AI is genuinely a good tutor

This part is worth saying plainly, because it's easy to only talk about the risk: a kid who's stuck on a concept now has something that will explain it a third and fourth different way, at whatever hour they're actually doing the work, without getting impatient. That's a real gain, not a loophole. The skill worth teaching is how to use it that way — asking for an explanation and a check, not an answer to hand in. One caution worth teaching alongside it: AI gets facts wrong confidently, so for anything with a specific right answer — a date, a formula, a citation — the habit in how to check if an AI answer is actually true is worth teaching a kid early, not just adults.

What to actually say to your kid

Not a lecture about academic integrity in the abstract — a specific ask: "Before you turn this in, could you explain it to me out loud without looking at your paper?" If yes, they did the work, whatever tools they used along the way. If no, that's the actual conversation to have — not about AI, but about whether they understood the assignment at all. AI just made it possible to skip that step invisibly. The check still works.

Keep your head

The rule isn't "AI or no AI." It's: if you took the answer away right now, could they still explain it? That question works whether the shortcut is a chatbot, a friend's answers, or an old sibling's notebook — AI just made it available at 11pm, alone, with no one to ask "did you actually get this?"

Get one of these a week. Our free newsletter sends one genuinely useful AI habit and one judgment check every week — no hype, four-minute read. Subscribe on the home page.