If that sounds familiar, you are in enormous company. This isn’t coming. It arrived. Pew Research Center’s national survey of 13–17-year-olds in February 2026 found that over 60% of them use AI chatbots, and 54% use them for schoolwork. About one in ten say they do all or most of their schoolwork this way. Almost 60% believe their classmates are using chatbots to cheat. Those are the ones who are willing to admit it to a stranger with a clipboard. Treat them as a floor. It's in your child’s pocket, it’s free, and it’s really, really good.

Norway just banned it. I think they’re wrong. In June, Prime Minister Jonas Gahr Støre announced that from August onwards Norwegian children in grades one to seven (ages six to thirteen) cannot use generative AI in school at all. Fourteen to sixteen-year-olds can use it under close teacher supervision. The government also announced more funding for physical books in June, reversing a decade of digital-first policy. Mr Støre said “The most important thing in school is that our children learn to read, write and do mathematics.”

I do have some sympathy for his position. But a blanket ban will not work, and here's why: unlike social media, AI is not an optional entertainment product your child could theoretically live with and not without. It is indisputably the basis of the working world they will have. By banning it outright, the Norwegian government have simply guaranteed that their children learn to use it secretly, badly, and without a single adult in the room.

The question was never 'whether'. It’s how. Go back in time, to when calculators were introduced at school. You may not recall, but there was a furious argument about calculators in classrooms in 1975. The proponents for a ban were also asking the wrong question. A calculator in the hands of a child who understands arithmetic is a fantastic labour-saving device. The same calculator in the hands of a child who never learned to multiply is a permanent prosthetic for a limb that was never allowed to grow.

The device didn’t change. The child’s relationship to their own thinking did. That distinction cuts through about ninety per cent of what you will read about AI in education, and it is worth holding onto, because the evidence does start to come through. In a 2025 MIT Media Lab study, students wrote essays with and without a chatbot. The chatbot group had significantly lower brain engagement and remembered almost nothing of what they had supposedly written. The researchers dubbed it ‘cognitive debt’. When the machine does the ‘thinking’, your child is borrowing against their future competence. And that debt is real. It doesn’t come due for a while.

When it really helps. The test is simple: is the child still driving?

• As a patient tutor. “Explain the water cycle like I’m twelve, then ask me three questions to check I’ve got it.” Used this way, it’s a patient teacher who never sighs, never checks the clock and never makes a child feel foolish for asking the same thing three times. I have known human tutors who could not clear that bar.
  
• As a first draft to argue with. If you have a rough outline and you dismantle it, you are thinking. The struggle goes on, and only the terror of the blank page is gone.
  
• As a study partner. “Turn my notes into ten practice questions.” The child still has to answer them, and that’s where learning lives.
  
• As a leveller. For a child with dyslexia, a child who is learning English, or a child whose parents genuinely cannot help with Year 11 maths, and I am including myself here, these tools quietly close a gap that used to determine entire life trajectories. That is not a small thing. When it quietly costs them.
• When it produces the finished product. An unread essay isn’t a shortcut. The essay was the thinking. Handing it over is a tiny act of self-sabotage with excellent formatting.
• When it removes the struggle. The discomfort of not-yet-knowing isn’t a bug in learning. It is learning. Strip it out and you remove the growth along with the pain.
• When it’s the only source. These systems are confident, articulate, and sometimes gloriously wrong. And they hallucinate, making up facts, quotes and whole studies with complete composure. Consider this: a $440,000 report to the Department of Employment and Workplace Relations, which looked at the automated penalty system in our welfare system, contained a fake quote from a Federal Court judge and references to academic papers that don’t exist. Sydney University researcher Chris Rudge has noticed it. Deloitte paid part of the fee and said Azure OpenAI had been used. If the Big Four can’t catch it on a job worth nearly half a million dollars, expecting a year 9 to catch it between netball and dinner is wildly optimistic. A child who can’t tell the difference is not being educated. They are being misinformed with beautiful grammar.
  

What not to say:

• “Using AI is cheating.” Too blunt to be true. It makes the honest child anxious and teaches the dishonest one to hide it better.
  
• “In my day we walked to the library.” Possibly true. At the very moment your teenager stops listening.
  
• “Just don’t use it.” You can’t enforce it. They know you can’t enforce it. Conversation over.
  
• “The school will catch you.” The detectors are unreliable, and your child knows that better than the school does. You’ve just started a cat-and-mouse game where you are Mickey.

What to say instead.

• “Show me how you used it.” Curiosity, not accusation. You will learn more and so will they.
  
• “Could you have written that yourself? Explain it to me now.” Still the best test we have. If they could, learning happened. If they can’t, it didn't happen.
  
• “What did it get wrong?” It is the single most valuable skill for the next twenty years, looking at the machine’s answer as a first draft to be checked, not a verdict to be copied.
  
• “Let’s write some rules together.” The rule a teenager helped draft is the rule that a teenager will sometimes follow.
  

You have not lost this race. Many parents tell me they feel outgunned, that their children understand these tools far more than they ever will, so what’s the point? I don’t accept it because it’s not true. Yes, your child is faster with the technology, but you are decades ahead on the only thing that counts: judgement. Knowing when an answer smells wrong. Knowing when one is too good to be true. Knowing when a shortcut is a trap. Those are the calls that someone with life experience makes infinitely better than a fourteen-year-old, however fast they type.

You don’t need to out-code your child. You need to do just what you have always done: teach them how to think, when to be sceptical, and how to tell help from dependency. Because an education has never been about producing a person who can just get answers. Answers have never been cheaper or more plentiful than they are now.

The point is to produce a child who can think, who can sit with a hard problem, who can tolerate not knowing and who can interrogate what you’re telling them. Those capacities are built the way they have always been built through effort that cannot be delegated. A tool that respects the struggle may be the best teacher your child ever has. A tool that removes it is stealing something you won’t notice is gone until it’s needed. And your job, again, is to know the difference and to help your child learn it too. That has never been a small thing.

Right now it might be everything. How do I raise a thinker?

Four ways to raise a thinker.

1. Ask, don’t forbid. “Show me how you used it” instead of “don’t use it” every time. You can’t monitor what you’ve driven underground.

2. Apply the explain-it test. If they can explain the work in their own words, learning occurred, whatever tool they used. If they can’t, it didn’t.

3. Teach scepticism, not trust. “What did it get wrong?” is the most important question you can teach a child to ask a machine.

4. Write the rules together, and follow them yourself. Agree as a family when AI is a tutor and when it’s off-limits. That includes you. Your example is the loudest rule in the house.

Poll: How do you feel about your child using AI for schoolwork?

1. Comfortable, it’s a tool like any other.
2. Cautious, depends entirely on how they use it.
3. Nervous, I’m not convinced they’re learning.
4. Honestly? I don’t know what they are doing.

    

   Dr Michael Carr-Gregg AO, MAPS is a child, adolescent and family psychologist. He has written books on raising young people and works with schools, families and organisations across Australia. He was responsible for leading the team that wrote the report that banned mobile phones in all government schools in 2018.