It depends on which part of the work the AI did. If it did the thinking your child was supposed to do, it is cheating, whether or not the school has a rule about it. If it did the doing around that thinking, it is help, and the same teacher who would mark down the first would be glad of the second. The line is not the tool. It is the part of the task your child was meant to learn from.
I taught in secondary schools for years before I trained teachers on this, and I want to give you the test teachers actually use, because it is not a detector and it is not a ban. It is a question you can ask at the kitchen table tonight.
What counts as cheating with AI?
Cheating is submitting thinking as your own that you did not do. Every school policy I have read says a version of this, usually in the words "unauthorised assistance", and the word that matters is authorised. A teacher who sets an essay wants to see your child build an argument. A teacher who sets a spelling list wants the spellings in your child's head. In both cases, a chatbot that produces the finished product has removed the thing the homework was for.
The confusion comes because the same tool can do the opposite. Asking an AI to explain why a paragraph does not flow, to test you on the spellings, or to give you three practice questions on fractions leaves the thinking exactly where it belongs. The principle I use with schools is outsource the doing, not the thinking. For families I put it more simply: let AI help with the doing; keep the thinking with your child.
The test teachers actually use
Good teachers have never relied on catching the product. They look at process and performance too: the drafts, the workings, the notes, and whether the child can explain the work out loud, answer a question about it, or do the next one unaided. I describe this as the three Ps, Product, Process and Performance, and for a parent it translates to three questions: what did my child produce, how did they produce it, and what can they now do?
That is why AI detectors are the wrong thing to hope for. They are unreliable, they flag the careful writers and the children who speak English as a second language, and a school that leans on them has stopped looking at the child. If a teacher asks your child to talk through a piece of work, that is not an accusation. That is the assessment working properly.
The question I ask school leaders to put to every task also works at home, and this is the version I give parents:
The Homework Test is one question to ask of any piece of homework before AI touches it: if the machine did this part, what would my child lose? If the answer is nothing, let it help. If the answer is the learning, that part stays with your child.
Formatting a bibliography: nothing lost. Writing the paragraph that argues why the character changed: everything lost. Once your child can run the test on their own, you have given them something more durable than a rule.
Does AI help or hurt learning?
Both, depending on how it is used, and the research is now clear enough to act on. Pew Research Center found in February 2026 that 54 per cent of US teens use AI chatbots for schoolwork, that one in ten do all or most of their schoolwork with them, and that 59 per cent believe pupils at their school use AI to cheat at least sometimes. Common Sense Media's June 2026 census found that 85 per cent of 9 to 17 year olds who use AI use it for schoolwork, and nearly half do so weekly.
The finding that should shape your rules comes from the University of Chicago's Booth School of Business. In a study published in April 2026, Leonardo Bursztyn, Alex Imas and colleagues told parents about research in which students who had used AI for their work scored nearly 20 per cent lower on quantitative reasoning when tested without it. Those parents became more supportive of school bans, and just as willing to pay for AI subscriptions, because they feared their child would fall behind other children who had one. The fear of falling behind is real. It is also the thing most likely to make you hand your child a tool that does the learning for them.
The tools are changing to meet this. On 18 August 2026 OpenAI launched ChatGPT for Teens, which puts under-18 accounts into a version with a Study Mode that asks guiding questions instead of giving answers, reminders that nudge a homework request back towards working it through, and "study hours" a parent can set so new chats start in Study Mode. It is a good design. It is not a substitute for you knowing which part of tonight's work is the thinking.
The Mistake I See Parents Make
The mistake is choosing a side. Some families ban it, which works until about age twelve and then teaches a child to hide it. Some families allow everything, usually with the words "they will use it at work anyway", and then are surprised at a parents' evening when a teacher says the essays are excellent and the tests are not. Both positions save the parent from the harder job, which is to sit down once and work out, subject by subject, which parts are the thinking.
The second mistake is treating a detector email from school as a verdict. Ask the teacher what they saw in the process and the performance, not just the product. A school that has done its homework will have looked at both.
Where help ends and doing begins, by age
Under 8. No AI for homework. Reading, counting and handwriting are the thinking, all of it. If they see you use it, say what you are using it for. You could say: "I'm asking the computer to check my spelling. Your job right now is to learn your spellings so you don't need to."
8 to 12. AI can quiz them, explain a concept a second way, or give practice questions. It does not write sentences that will be handed in. Sit beside them for the first few uses. You could say: "It can test you and explain things. It doesn't do the writing, because the writing is where your brain grows. If it wrote it, your teacher would be marking the computer, not you."
13 to 15. Give them the Homework Test and let them apply it, then check the process: ask to see the plan, the draft, or the workings. Turn on Study Mode. You could say: "Show me one piece of work where AI helped and tell me which bits were yours. If you can't tell me, that's the bit we need to fix."
16 to 18. The stakes are exams and coursework (GCSE and A level coursework in the UK, graded assignments and college applications in the US), where the rules are written down and the penalties are real. Ask them to find and read their school's policy, then talk about the one grey area they are unsure of. You could say: "Where's the line for you? Find me the school's rule and tell me whether you agree with it."
What to ask the school
Three questions get you further than "do you allow AI?":
- Which pieces of homework this term are the thinking, and which are the practice? Teachers know. Ask them to say.
- How do you assess process and performance, not just the product?
- What do you want us to do at home when a child asks to use AI?
A school that can answer all three has a policy that means something in the classroom. If yours cannot yet, this is the work I do with schools: training teachers to set work AI cannot do for a child, and to assess in ways that show what the child can now do. It is worth sending to your head of year. I set the fuller argument out in my books on AI in education, and I write a Sunday newsletter that keeps more than 44,000 teachers and parents up to date without the panic.
Dan Fitzpatrick is a former secondary teacher and assistant headteacher who has trained more than 150,000 educators on AI. Read Dan's story.
