Your kid will use AI for schoolwork. Teen usage keeps climbing, bans have quietly failed almost everywhere, and "just don't" is not a strategy โ it's an invitation to use it secretly, which is the worst of all outcomes: all of the shortcut, none of the guidance, plus a habit of hiding.
What works is a line that's clear enough for a ten-year-old to apply alone. Here it is.
The line: tutor, not ghostwriter
AI can teach you. AI cannot be you.
A tutor explains the concept again, differently. Checks your work and shows where you went wrong. Quizzes you. Suggests how to start when you're stuck staring at a blank page. All allowed โ that's the same thing a good human tutor or a helpful parent does.
A ghostwriter does the assignment while you watch. Writes the essay, solves the problem set, answers the reading questions. Not allowed โ not because a rule says so, but because of the argument below, which you should actually make to your kid.
Here's how to say it out loud, by age:
Ages 5โ7: "AI is like a helper who explains things. It doesn't do your work for you โ you do your work, and it helps you understand."
Ages 8โ10: "Think of AI like a tutor on the couch. It can explain, quiz you, and help you get unstuck. It can't write what you hand in. That's the line."
Ages 11โ14: "Use it like you'd use a tutor โ not like you'd use a ghostwriter. If you couldn't defend this in front of your teacher with the laptop closed, you crossed the line."
The argument (make it, don't skip it)
Homework isn't a product the teacher needs โ the teacher has the answers. Homework is a gym. Nobody pays someone to lift weights for them. When AI writes the essay, the essay gets done and the only point of the essay โ the reps โ vanishes. Kids find this argument annoying and fair, which is the sweet spot for any house rule.
If your kid pushes back with "but the teacher can't tell," don't argue detection โ argue learning:
You: "Maybe they can't tell. That's not the test. The test is: did you get stronger at writing, or did the machine get stronger at writing while you watched?"
Kid: "But everyone uses it."
You: "Some people do. Some people also copy off each other. We're not building our house rules around what other people get away with. We're building them around what actually makes you better."
Media-literacy and education researchers consistently find that kids who understand why a rule exists follow it more reliably than kids who only know what the rule is โ especially once they're old enough to use AI without you in the room.
The enforcement mechanism: explain it back
You can't audit every assignment, so don't try. Use the test that can't be faked:
"If you used AI, you should be able to explain the answer back to me with the laptop closed."
Can they? Then whatever they did was learning โ pass, regardless of how much AI was involved. Can't? Then it was ghostwriting, and the assignment gets redone the slow way. This shifts enforcement from surveillance (which breeds hiding) to demonstration (which breeds studying).
Run it like this:
You: "Walk me through this essay. What's your main argument?"
Kid: [explains]
You: "Where did you get stuck, and how did you get unstuck?"
Kid: [explains or can't]
You: "Close the laptop. Explain paragraph three to me again โ the one about the treaty."
If they can do it, you're done. If they stumble on their own sentences, you have your answer โ and now you have a repair plan, not a punishment plan.
Why teachers are moving toward oral defense
This isn't just a parent hack. Across schools, teachers are quietly adapting to AI by asking students to talk about their work, not just submit it. A five-minute oral defense โ "explain your thesis," "where did you disagree with your sources," "what would you change if you had another day?" โ is nearly impossible to fake if you didn't do the thinking. It also mirrors how knowledge actually works outside school: you have to be able to explain what you know, not just produce a document that looks like you know it.
Teachers who use this approach aren't trying to catch cheaters for sport. They're protecting the part of homework that matters โ the reps โ in a world where generating a plausible essay takes thirty seconds. Your explain-it-back test at home is the same muscle, rehearsed before the real thing.
Subject-by-subject quick guide
- Math: AI checks answers and re-explains methods; it doesn't produce solutions to hand in. (It also still makes arithmetic slips โ a feature, honestly: "find the AI's mistake" is a great exercise.)
- Writing: brainstorming, outlining against a blank page, and critique of their draft: allowed. Generating sentences that get submitted: not. A useful trick โ have the AI ask your kid questions about what they want to say, instead of saying it for them.
- Reading: summaries as a preview before reading or a check after โ never instead. The explain-it-back test catches this one instantly.
- Languages: conversation practice partner, yes. Translator of homework, no.
Age-band variations by subject
- Ages 5โ7: AI for homework is mostly off the table โ not because AI is dangerous, but because the homework itself is handwriting, reading aloud, and basic arithmetic. If they ask, keep it to "explain this word" or "give me another example like this one" on your account, shared screen.
- Ages 8โ10: introduce the tutor line with writing and math first. A good co-prompt: "Ask me five questions about what I want to write about โ don't write it for me." For math: "Show me a similar problem and explain each step, then I'll do mine."
- Ages 11โ14: they will push every boundary. Let them use AI for research structuring ("what are three angles on this topic I haven't considered?") but not for prose generation. The explain-it-back test becomes non-negotiable on anything over a paragraph.
The family AI agreement (print it, sign it, stick it up)
Rules kids help write get followed more than rules kids receive. Sit down once โ fifteen minutes โ and fill in a version of this together. Post it where homework happens.
Our Family AI Agreement
- AI is a tutor, not a ghostwriter. It explains, quizzes, and helps me get unstuck. It does not do my homework for me.
- My words go in; my words come out. AI can help me think, but what I turn in has to be mine โ my sentences, my ideas, my mistakes.
- Explain-it-back. If I used AI, I can explain my work with the screen closed. If I can't, I redo it.
- School rules win at school. If a teacher says no AI on an assignment, that assignment is AI-free โ even if our house usually allows tutoring.
- Honesty always. If someone asks whether I used AI, I tell the truth. No exceptions.
- No personal info. Real names, school names, addresses, and photos stay out of prompts. We keep our details for people we trust.
- Stuck means ask a person first. If I'm confused, I try the hard part myself, then ask a parent or teacher, then use AI as a tutor โ in that order.
- When in doubt, ask before submitting. If I'm not sure whether something crosses the line, I ask. Asking is never trouble.
Signed: _______________ Date: _______________
Read it out loud once. Let them suggest a ninth rule if they want โ ownership matters. The agreement isn't a contract to enforce with punishments; it's a shared reference for the moments when "everyone uses it anyway" shows up in the group chat.
Scripts for the hard conversations
These are the moments that decide whether your kid keeps coming to you or starts hiding AI use entirely. Repair, not escalation.
When you find a ghostwritten essay
Don't open with accusation. Open with the test:
You: "Hey โ walk me through this one. What's the main thing you're arguing?"
Kid: [stumbles, or gives vague answers that don't match the writing]
You: "I'm not mad. I need to understand what happened here, because this doesn't sound like how you talk about this topic. Did AI write parts of this?"
Kid: "Yeahโฆ everyone does it."
You: "I'm not interested in everyone. I'm interested in you. Here's what happens next: we figure out what you were supposed to learn from this assignment. You redo it โ with AI as a tutor if you want, but your words. And we talk about why ghostwriting felt like the move. Not today as a lecture. Today as a redo. Tomorrow as a conversation."
The redo is the consequence. Shame is not. A kid who ghostwrites is usually stuck, scared of a bad grade, or copying a norm they see everywhere โ address the stuck part first.
When your kid says "the teacher lets everyone use AI"
You: "Maybe they do for some things. Show me the assignment instructions."
Kid: "It doesn't say."
You: "Then we follow our agreement until it does. If the teacher meant 'use AI for anything,' they'll say so clearly. If they meant 'use AI to brainstorm,' that's tutoring โ allowed. If you're not sure, you can ask the teacher. I'll back you up."
Teaching kids to read assignment boundaries carefully is a skill that transfers to every workplace they'll ever enter.
When your kid says "you use AI at work"
Fair point. Answer it directly:
You: "I do. And here's the difference: at work, the deliverable is the product โ my job is to produce a good document efficiently. At school, the deliverable is you getting better. Different game, different rules. You'll use AI freely at work someday too. Right now you're in the gym."
When a younger kid asks "why can't I just ask it the answer?"
You: "Because your brain is the one that has to grow, not the computer's. The computer already knows. You're still learning. If I let it do your thinking, I'd be stealing your reps โ and you'd feel it later when something hard shows up and you haven't practiced."
When school rules and house rules differ
Schools range from bans to mandatory AI use, sometimes in the same building. The family rule that survives all of them: school rules win at school, and honesty wins everywhere. If the teacher forbids AI, that assignment is AI-free even though your house allows tutoring โ integrity is doing the boring thing when the rule applies to you. And if your kid is ever asked "did you use AI?", the answer is the true one. That's transferable to the rest of their life in a way no citation format is.
When your house is stricter than school, say so plainly: "Our rule is tighter because we care about the gym, not just the grade." When school is stricter, respect it without eye-rolling โ kids notice when parents treat teacher rules as optional.
"What if my kid..." โ troubleshooting the real moments
...got a good grade on AI-written work and wants to keep doing it? Name what happened without attacking: "The grade measures the essay. It doesn't measure you. Let's run the explain-it-back test on that assignment right now." If they can't defend it, the grade isn't the trophy โ the gap is. Then redo one assignment tutor-style together so they feel the difference between "done" and "learned."
...says AI helps them when no one else will? Take that seriously. Some kids genuinely lack access to tutoring, patient parents, or teachers with time. The response isn't "too bad, no AI" โ it's "let's figure out how to use it as a tutor, not a replacement for help." Offer the human backup where you can: sit with them for twenty minutes, connect them with a school resource, or schedule a weekly homework check-in. AI-as-tutor plus a human in the loop beats AI-as-ghostwriter alone in a bedroom.
...is anxious they'll fall behind peers who cheat freely? Acknowledge the fear โ it's real. Then separate two games: "Some kids are optimizing for grades this semester. We're optimizing for you being able to explain your work at 25. Different timelines." Kids with the explain-it-back habit consistently report feeling more confident in oral exams and interviews later, because they've practiced defending ideas out loud.
...is a teen who refuses any rules as "controlling"? Shift from rules to a deal: "You want more independence. I want to know you're actually learning. The deal: you handle homework your way, but once a week you explain one assignment to me with the screen closed โ your pick. If you can do that, I stay out of your process." Demonstration-based trust scales; surveillance doesn't.
...got caught by the school, not by you? Side with the repair, not the institution's drama: "Okay. What's the school's consequence? What's our redo plan? What would you do differently next time?" A kid who gets caught and finds you on their side of the learning line โ not the punishment line โ keeps talking. One who gets a second punishment at home learns to hide better.
The reframe that makes this easy
Every game and project in this pathway has taught the same posture: AI is an instrument, and your kid is the operator. Homework is just the highest-stakes place that posture gets tested. Kids who built the story machine already know the difference between directing a draft and passing one off โ you're not introducing a rule, you're naming one they've already lived.
Frequently asked questions
Is using AI for homework always cheating?
No โ using AI as a tutor (explaining concepts, quizzing, helping you get unstuck) is learning. Using AI as a ghostwriter (generating answers or prose you submit as your own) is cheating. The explain-it-back test tells you which one happened.
What age should we start homework AI rules?
Around eight to ten, when writing assignments and independent homework begin. Younger kids rarely need AI for schoolwork; keep any use to shared-screen tutoring on your account.
Should I tell my child's teacher we use AI at home?
You don't need to announce it, but if the teacher asks or the school sets a policy, be honest about your family's tutor-not-ghostwriter line. Teachers increasingly appreciate parents who reinforce learning-over-shortcuts โ it aligns with what many of them are trying to protect.
My kid's school requires AI on some assignments. Now what?
Follow the assignment. School rules win at school. Your family's ghostwriter line still applies to work the teacher expects to be genuinely theirs โ when in doubt, ask the teacher to clarify what's allowed on that specific task.
How do I know if they used AI without asking directly?
You often can't โ and trying to detect every instance turns you into surveillance. Use spot checks with the explain-it-back test instead of constant monitoring. Random "walk me through this one" conversations catch gaps without requiring you to read every sentence.
What if my child has learning differences and AI genuinely helps them access the material?
Talk to the teacher and, if you have one, their support team. Many educators support AI as an accommodation when it's documented โ text-to-speech, simplified explanations, step-by-step breakdowns. The tutor line still applies: the tool helps them learn, it doesn't replace their demonstration of learning.
Last stop in the pathway: putting it all together into an age-by-age roadmap from first questions to first real builds.