Your child is relying on ChatGPT too much when they can't start, explain or finish schoolwork without it. The fastest test takes two minutes: ask them to talk you through something they handed in. If the ideas fall apart out loud, the chatbot did more than help.
That's the short answer. The longer one matters, because over-reliance hides well. The homework looks great. The marks look fine. And the gap underneath grows wider every week, until an in-class essay or a test pulls the curtain back.
I'm an Ontario Certified Teacher with 14 years in education (more about the founder), and I spent three years as an AI evaluator, testing AI models for weak reasoning and flimsy writing. I know what machine writing looks like from the inside, and I know what a child's real writing looks like on the page. Here are the signs I watch for, and what to do when you spot them.
It's hard to spot because AI polish looks like progress. A chatbot doesn't leave spelling mistakes, crossed-out lines or half-finished thoughts, the usual fingerprints of a child at work. So the evidence parents and teachers rely on vanishes.
Kids know where the line sits, too. In a Pew Research Center survey of U.S. teens, 42% said using ChatGPT to write essays was not acceptable, and only 18% said it was. In my experience, over-reliance doesn't begin as a plan to cheat. It starts as one tired night, one tricky question, one "I'll just see what it says." Then the shortcut becomes the road.
One sign on its own can mean nothing. Three or four together form a pattern worth a calm conversation.
You know your child's voice. If a Grade 6 book response starts to sound like a polished magazine column, with words they've never said at dinner and not one of their trademark turns of phrase, trust that hunch. Real kid writing has quirks. Machine writing has a smooth, faceless sameness.
This is the clearest sign of all. Ask, "Why did you pick this example?" or "What does this sentence mean?" A child who wrote the work can answer, even if the answer is messy. A child who prompted it shrugs, guesses or changes the subject. Research points the same way: in a small MIT Media Lab study (an early preprint with adult writers), the people who wrote essays with ChatGPT struggled to quote their own work accurately.
Strong take-home work beside shaky in-class writing and test results opens a gap worth noticing. If the report card says one thing and the homework folder says another, believe the in-class work.
The blank page used to be a hurdle. Now it's a panic button. Watch for a child who won't write a single sentence until a laptop is open, or who melts down at the idea of writing by hand.
In English, AI summaries let kids skip the reading. Ask about a scene they liked or a character they couldn't stand. A child who read the book has opinions. A child who read a summary has a list of events.
A multi-paragraph analysis that used to take an evening is done in fifteen minutes, with no outline, no questions and no complaints. Speed isn't proof of anything. Speed plus silence is worth a gentle look.
Writing is messy work. There should be a scrappy plan, a rough draft, a sentence your child fought with. When every assignment arrives finished in one clean pass, ask to see how it got there.
A common pattern: book reports earn the best marks ever while the in-class reading journal holds two thin lines a week. The fix starts with talking instead of typing: your child gives an opinion, someone pushes back, and they defend it out loud before writing a word.
A mark your child didn't earn isn't a win. I'd take an honest B over a borrowed A every time.

Start curious, not accusing. In my experience, most kids who lean on AI aren't trying to fool anyone. They're stuck, tired or unsure they can do it themselves, and a lecture confirms that fear.
In my experience, a ban tends to backfire, and it skips a skill your child will need for life. A better rule fits in four words: thinking first, AI second.
For a full five-day routine, my guide to kids using AI for homework walks you through it step by step.
If the signs keep stacking up, the chatbot may be covering a real skill gap, not laziness. I've found that kids reach for AI hardest in the places they feel weakest.
Polished homework can hide that gap for months. The Learning X-Ray™ looks past the finished page: it studies your child's recent schoolwork alongside new work they produce on their own, with no help, and the Performance Report flags hidden micro skills such as depth of reasoning and self-correction, so you can see which ones the chatbot has been standing in for.
Raising kids who think for themselves is the reason behind our story, and it's why our online English courses, opening soon, treat AI as a partner to thinking, never a replacement. Start with one question this week: "Can you tell me about this?" The answer will tell you a lot.
Comments will be approved before showing up.
Rachel Esco
Author