Let's start with the honest part: AI is genuinely good at high school homework. It will produce a competent Social 20-1 paragraph, solve a Math 30-1 question with working shown, and explain the process convincingly enough that a student could defend it if asked.
If your kid is using it that way, the assignments come back fine. Nothing looks wrong. That's the whole problem — the marks that would warn you are the marks it protects.
Compare assignment marks against test marks.
Assignments happen at home, with everything available. Tests happen in a room, with nothing. A student who understands the material scores roughly similarly on both. A student whose homework is getting done for them scores well on assignments and noticeably worse on tests.
That gap is the signal. It's not proof — plenty of kids have test anxiety, or are fine at home and freeze up under time pressure. But it's the thing worth looking at, and it's visible on any report card that separates the two.
The Grade 12 version of this gap is expensive. The final mark in a diploma course is mostly school work, with the diploma exam about a third, and Alberta's own guidance notes the two marks should land reasonably close, since both measure the same curriculum. When they don't, you find out in the summer — after post-secondary offers have gone out on the strength of the school mark alone.
Three reasons.
Enforcement is impossible. It's on the phone, the laptop, the school Chromebook, and a friend's device. You are not going to win this on access.
It's not all cheating. A student asking AI to explain a concept a different way is doing something a good tutor does. A student having it write the essay is not. Ban the category and you ban both, so the useful version goes underground with the other one.
It makes it a secret. The moment it's forbidden, you stop hearing about it, and you've traded a manageable problem for an invisible one.
Don't open with an accusation. You probably can't prove it, they'll deny it, and then you're arguing about honesty instead of about school.
Open with the exam.
"In June you're going to sit in a room for three hours with no phone, no laptop and no help. Whatever you can do in that room is the whole thing. Are you ready for that?"
That's not an accusation. It's a fact they can't argue with, and it puts you both on the same side of the problem — you're not the person trying to catch them, you're the person pointing at a deadline.
Then the practical question: "Are your test marks lower than your assignment marks?"
If yes, you've got a shared, observable thing to work on. If no, you've learned something and the conversation cost you nothing.
Worth being specific, because "use it responsibly" means nothing to a sixteen-year-old.
Fine: - Explain this concept a different way - I got this wrong — where did my reasoning break? - Quiz me on this unit - What is this question actually asking me to do?
Not fine: - Write this - Solve this - Give me the answer
The line isn't AI versus no AI. It's whether the student did the thinking. In the first list they're still doing the work and getting help doing it. In the second, something else did the work and they were present.
That distinction is teachable, and it's a much better thing to install than a ban, because it's the distinction they'll need for the rest of their working life.
I build an AI study tool, so treat this as declared interest rather than neutral commentary.
Here's the argument anyway. The problem isn't that students have access to a machine that can do their homework. It's that they have access to one that will do it for them by default, and nobody has offered them a version that won't.
A tool that makes the student explain the question, choose the first move, do the work and check it is doing the same job with the same technology and producing the opposite outcome. Same access. Different design.
Which is a decision someone makes when they build the thing — not a moral failing in a sixteen-year-old who took the easy option that was placed in front of them.
Joseki is an AI that won't give your kid the answer. They photograph the question, say what it's asking, pick the first move, and do the work. Help arrives when they're stuck. Free while it's in beta.
Note for review: the declared-interest paragraph stays. This post makes an argument that benefits the product; saying so is what makes the rest of it credible to a sceptical parent.