The AI-education death spiral: when 'let them cheat' becomes the plan

A piece that made the rounds recently put the quiet part out loud. Its argument, picked up in a long Hacker News thread, was that some schools are not really fighting AI cheating anymore. They are surrendering to it: assigning the same essays and problem sets, knowing a chatbot is doing the work, and grading the output anyway. The writer called it a death spiral. Students stop learning, the credential stops meaning anything, and eventually nobody trusts the grade on the page.

The uncomfortable part is that the spiral is real. The comforting fiction is that there are only two ways out of it.

Banning the tool was never going to work

The first reflex is to wall the machine off. AI detectors, handwritten essays under supervision, laptops collected at the door, plagiarism software pointed at twelve-year-olds. Some schools have gone further, into something close to surveillance.

Set aside that the detectors do not reliably work and routinely flag children who wrote every word themselves. The deeper problem is the goal. You are spending enormous energy to protect an assignment whose entire value rested on the assumption that doing it was hard. The tool just proved it was not. Locking the door does not make the room worth defending.

The second reflex, the one the thread was really about, is to give up and let the kids cheat. Keep the worksheet, ignore where the answer came from, hand back a grade everyone privately knows is fiction. That is not a plan. It is the spiral, written down and signed.

The worksheet was the weak point

Here is the part that should change how a parent reads this whole argument. AI did not break learning. It exposed an assignment that was already hollow.

Think about what most homework actually asks. Produce a five-paragraph essay on a topic with a known shape. Fill in the blanks. Define these ten terms. Solve thirty problems that differ only in their numbers. Every one of those tasks has a single correct output and a page where it lives. They were built to be checkable at scale, one adult facing thirty children, not because they were the best way to learn but because they were the only way to grade that many people at once.

A task that can be fully captured in its final output was always going to be the first thing a machine could fake. The cheating is not a moral failure that arrived with the chatbots. It is what happens when the thing you are measuring is the output instead of the understanding.

So the real question is not how do we stop children from using AI. It is why we are still assigning work where copy-pasting an answer even counts as a shortcut.

Make the answer beside the point

There is a third way out, and it is not subtle. Change what you ask a child to do so that the final answer is the least interesting part of it.

Give a child a real project, something they choose part of, that does not have one correct response sitting at the back of a book. Have them build it, get it wrong, and try again. Then have them explain it out loud: walk through why they made each decision, defend the parts that are weak, answer a question they did not see coming. Ask them to take feedback and iterate, so the work has a visible history rather than appearing fully formed overnight.

Now watch what happens to cheating. A child can have a chatbot draft a paragraph. A child cannot have a chatbot understand the paragraph for them when you ask, in the moment, why did you choose that. They cannot outsource the third revision, or the defense of a tradeoff, or the small original connection nobody handed them. When learning is assessed by doing and explaining, the word “cheating” quietly loses its meaning. There is no answer to steal, because the answer was never the point.

This is also where the tool changes sides. A child who has to defend their reasoning uses AI the way a capable adult does, to check a draft, to argue against, to get unstuck, and stays the one in charge of the thinking. The goal was never to keep children away from these tools. It was to make sure the child is the one holding the pen, not the one being led by it.

What this means for your child

If your child is “cheating” on a worksheet, the worksheet is most of the story. It is asking for exactly the kind of output a machine now produces for free, then treating the child as the problem for noticing. Your child is not broken. The assignment is.

This is the whole reason Globeskool teaches one child at a time through real projects they build, defend, and revise, work where there is nothing to copy because the understanding lives in the child, not on the page. If you want a concrete sense of how your own child thinks, defends an idea, and works with a tool rather than hiding behind it, that is what a short, honest look can show.

See how your child works without the worksheet →