If you take one thing from this entire guide, take this: once a week, ask your kid to teach you something they finished — with the screen closed.
Two minutes. No warning needed, no app required. It is the single highest-signal thing a parent can do, and it works because retrieving something from memory is a different act from recognising it on a page. A kid who outsourced the thinking cannot do it. A kid who learned it can barely be stopped.
Do this once
The first sitting: twenty minutes, side by side
Not a lecture. You're establishing that this is something you do together and are allowed to talk about — which is the whole ballgame, because the alternative is that it happens in a bedroom and you never hear about it again.
Let them drive
They open it, they type. You sit next to them. Start with something they genuinely find hard this week — a real assignment, not a demo. If you take the keyboard, you've made it your tool instead of theirs.
Set the tutor behaviour together
Paste in the instruction below and watch what changes. This is the moment most families miss: out of the box these tools answer; told to teach, they teach. Your kid should see that switch happen with their own eyes.
Act as my tutor for [topic]. Ask me one question at a time. After each answer, tell me if I'm right and why — but never give me the answer before I've tried. If I'm wrong twice, give me a hint, not the solution.Catch it being wrong, on purpose
Ask something you can verify and it's likely to fumble — a specific local fact, a recent event, a page number in a book you own. When it invents something, don't gloat. Just say: "Okay. So how would you know?" That question is the entire curriculum.
Name the rule out loud, in their words
Don't recite yours. Ask: "So what's the line between using this and cheating with it?" Let them define it. Kids defend rules they wrote; they route around rules you imposed. Then write it on the fridge — Part Two has a printable agreement.
End by running the check once, playfully
Close the laptop. "Okay — teach me what you just did." Make it light. You're installing a ritual, not springing a trap. If they can't do it, that's not a failure of the evening; that's the evening working exactly as designed.
The playbook
What good use looks like, by assignment
"Use AI wisely" is useless advice. The honest rule changes completely depending on what's in front of them. Pick the kind of homework on the table tonight.
The hard moments
What to say when it goes wrong
These are the four conversations that actually happen. How you handle the first one determines whether you hear about the next three.
They clearly didn't write it, and you both know it.
Do not open with an accusation — it converts a teachable night into a lie you'll both defend for a week. Open with the check, calmly, as though it's routine. Because it is routine.
"Walk me through this one out loud."
Let the silence do the work. When it comes, don't fill it with a lecture. Give them the exit that costs them nothing:
"Okay. That's the part we need to fix — not the grade, the fact that you'd be stuck if this were a test. Want to redo it the slow way with me?"
You're separating the offence from the consequence. The consequence they actually fear is a proctored exam, and you're on their side against it.
"Everyone does it. You're making me the only one who can't."
Largely true — 59% of teens say AI cheating happens regularly at their school. Denying it costs you credibility you'll need later.
"You're right, most of them are. I'm not asking you to be the only kid not using it — I'm asking you to be one of the ones who can still do it without it."
Then make it concrete and self-interested, because at this age fairness arguments lose and self-interest wins: the exam is proctored, the interview is live, and the thing being graded then is what's in their head.
They're frustrated and want the answer, now. It's late.
Sometimes the right call is to give it to them. A kid crying at 10:40pm is not learning anything, and the productive struggle stopped being productive twenty minutes ago.
"Take the answer. Tomorrow, when you're not fried, do one just like it from scratch and show me."
You've kept the practice without breaking the kid. Rigidity here is how the whole agreement gets abandoned in week three.
They ask you something about AI you can't answer.
Say so, then find out together — in front of them. This is worth more than a correct answer, because it demonstrates the exact habit you're trying to install.
"No idea. Let's look it up and see if it's telling us the truth."
Your credibility here doesn't come from knowing the technology. It comes from being the person who checks.
The rollout
Four weeks, one change each
Introducing everything at once is how this fails. The research on AI tools for students with learning differences found the same thing: slow sequential adoption beat rapid multi-tool rollout on long-term outcomes.
The sitting, and nothing else
Do the twenty minutes above. Add no rules, no monitoring, no follow-up. This week is only about establishing that this is discussable in your house.
Introduce the check
Once, mid-week, on something they're proud of — not something you suspect. Two minutes. Frame it as interest, not inspection. If the first time you ever run it is the night you're suspicious, it will read as an accusation forever.
Hand over one prompt
Pick the assignment type they struggle with most from the playbook and give them that prompt. One. Let it become a habit before you add a second.
Write the agreement and review the setup
Now that you both have real experience, the rules will be specific instead of theoretical. Print the agreement from Part Two, re-check quiet hours, and set a reminder to revisit at the end of term.
When it isn't working
Troubleshooting
| What you're seeing | What's usually going on | What to change |
|---|---|---|
| They refuse the check | It's landing as surveillance, not curiosity — usually because the first one came on a suspicious night. | Drop it for two weeks. Restart on something they're proud of, in a subject they like. |
| Homework is instant, grades are sliding | Classic cognitive debt. Output is fine; nothing is being retained. | Move to the studying playbook. Retrieval practice, not explanation. The grade recovers before the habits do. |
| It teaches for two turns, then just answers | The instruction decayed over a long conversation. Very common and not their fault. | Start a fresh chat per assignment, or turn on Study Mode by default in the parental controls. |
| They've stopped telling you anything | Something got punished that should have been received. | Say plainly that telling you costs nothing. Then prove it once, at real cost to your own position. |
| It's become a friend, not a tool | Different problem, higher stakes — this is the companion risk, not the homework one. | Go back to Part Two's risk section today. This one doesn't wait for the weekend. |
Signals
What it looks like when this is working
Not grades. Grades were the last thing to move and will be the last thing to tell you anything.
- They can explain finished work without looking at it — and they don't dread being asked.
- They volunteer that the AI got something wrong. This is the big one. It means they're checking.
- They use it and then close it, rather than leaving it open through the whole evening.
- They start saying "I want to try it first" without being prompted.
- They tell you about a chat that got weird, unprompted, without appearing to weigh whether to.
One more thing
The hard part isn't knowing what to do. It's Tuesday.
Everything on this page works. The problem is that it depends on you remembering, at the right moment, with the energy to do it — after work, on an assignment you haven't read.
If you want to tell me what would actually help on a Tuesday night, or what in this guide missed, reply to the email that brought you here. I read every one.
— Hussain