"Is my grandchild's homework written by AI?" How to talk about it

You read the essay. It’s good — better than the last one. The sentences are tidy, the argument is neat, and something in you goes: did she write this?

Maybe. Maybe not. Before you decide what to do with that feeling, it helps to know what’s actually happening — because it’s more complicated, and more interesting, than “cheating.”

What’s really going on

AI chatbots can write a passable school essay in ten seconds. Every teenager knows this. Most have tried it at least once, in the way every generation tried something. That’s the honest starting point.

But “used AI” covers a huge range:

• Asked it to explain the topic before writing. That’s what a good teacher or a library book would do.
• Asked it to check spelling and grammar. That’s what a spell-checker does, and schools have allowed those for decades.
• Asked it for an outline, then wrote the essay themselves.
• Had it draft the essay, then rewrote it in their own words.
• Copied the whole thing and handed it in.

Only the last one is what you’re worried about — and it’s the least common, partly because teachers can usually tell, and partly because it’s boring. The others are, increasingly, just how homework gets done.

What the schools have decided

There isn’t one answer; it changes by country, by school, and by year. But the direction is clear: most schools have moved from “ban it” to “teach when it’s allowed.” Many now say explicitly: fine for research and checking, not fine for the final text, and always say if you used it.

So the first useful question isn’t “did you use AI?” It’s “what does your school say about AI?” Your grandchild will know. The answer might surprise you — and it gives you both a shared rule to talk about instead of your suspicion versus their defensiveness.

How to have the conversation

Don’t open with the essay. That’s an accusation, and you’ll get a wall.

Ask them to show you. “I keep hearing about this AI thing — can you show me how it works?” Teenagers love being the expert. In five minutes you’ll see what they actually do with it — and it’s usually more sensible than you feared. You’ll also learn something.

Ask what they think is fair. “Where’s the line, do you reckon? What counts as cheating?” You’ll get a thoughtful answer. Most young people have thought about this more than the adults have.

Talk about the point, not the rule. The reason to write an essay yourself isn’t that a rule says so — it’s that writing is how you learn to think. “If the AI writes it, what did you get out of it?” is a real question, not a trap.

If you’re worried, say so — once, plainly. “This one felt different from your usual writing. Did you have help?” Then listen. If they did, the interesting part is why: too busy? Didn’t understand the topic? Everyone else does? Each of those is a different conversation, and none of them is “you’re a cheat.”

What not to do

Don’t run it through a “detector.” There are websites that claim to spot AI writing. They’re unreliable — they flag careful human writing as AI and miss actual AI writing all the time. Accusing a child based on one is a fast way to a bad afternoon and a wrong answer.

Don’t make it about your generation. “In my day we went to the library” is true and unhelpful. In your day, calculators were cheating too.

Don’t ban it. You can’t, and it makes it secret. Better to be the grown-up they can talk to about it.

The thing worth passing on

If you take one idea to that conversation, make it this: AI is a brilliant assistant and a terrible replacement. Use it to understand, to check, to get unstuck. Don’t use it to skip the part where you learn. That’s true for the essay — and it’s exactly the same rule this site gives adults about letters, health and money. You’re not teaching them a school rule. You’re teaching them the habit that will matter for the rest of their lives.

The short version

Start with curiosity, not the essay. Ask what the school allows. Ask them to show you. Talk about why writing matters, not whether they broke a rule. And skip the “AI detectors” — they don’t work.