Study uses that help and hurt

The same tool can make you a stronger student or a weaker one, and from the outside the two uses look almost identical: a person typing to a chatbot about schoolwork. The difference is what is happening in your head while it happens, and there is a simple test for telling them apart.

Uses that build you

• Explaining: asking for a concept explained differently, then more simply, until it clicks.
• Practice questions: generating problems in the style of your exam, doing them yourself, then checking your answers.
• Feedback on your own work: handing over your draft or your solution and asking what is weak and why.
• Breaking down hard material: turning a dense reading you have done into a structure you can hang the detail on.

In each of these, the effort stays with you. You still wrestle with the idea, attempt the problem, write the draft. The tool speeds up the loop between trying and finding out how you went, and a faster feedback loop is one of the most reliable ways to learn anything.

Why does being quizzed work so much better than rereading? Because pulling an idea out of your own memory strengthens it in a way that looking at it again does not. Rereading feels productive because the material looks familiar, but familiar only means you recognise it on the page, and there is no page in the exam. Retrieval is harder, which is exactly the sign that it is working. A model that fires questions at you, tells you which answers were wrong and then drills the weak spots again is running one of the most reliable revision methods known, on demand, for free.

A study session, timed

Here is what the difference looks like inside a single hour. Version one: you ask for a summary of the chapter, read it in ten minutes, ask for a summary of the next chapter, and finish the hour feeling ahead. Version two: you spend twenty minutes attempting five practice questions the model generated in the style of your exam, with your notes closed. You get two wrong. You paste in your working and ask not for the answer but for where your reasoning went off the rails. You redo both, then ask for two fresh questions targeting the same weakness, and this time you get them right. Version one produced sixty minutes of recognition. Version two found and repaired two misunderstandings that were otherwise headed into the exam with you, and catching a misunderstanding before the exam does is the entire point of studying.

Uses that hollow you out

• Generating assignments and submitting them, reworded or not.
• Summarising readings you never read, so you can talk about a book you have not met.
• Doing your practice problems for you, so the homework looks done but the skill was never practised.

These feel efficient, and in the short term they are: the work gets handed in and the evening is free. The cost is invisible until it is not. Every one of them removes the exact struggle that was the point of the task. The assignment was never the product. It was the exercise. Getting a machine to do your push ups does not make you stronger, no matter how good the push ups look.

The worked solution trap

One use sits right on the line and fools almost everyone: asking for a worked solution and reading it. Watching a solution unfold feels like learning, because every step makes sense as you read it. But following a step and producing a step are different skills, and the exam asks for the second one. The honest sequence is attempt first, struggle for a while, and only then look, comparing the solution against your attempt line by line to find the exact step where you went wrong. Same worked solution, completely different result, because the version with an attempt in front of it turns the solution into feedback, and feedback only exists where there is an attempt for it to land on.

The exam room test

Here is the test that cuts through every grey area. Your exams almost certainly allow no AI. So ask, about any study use: did the way I just used this tool build something that will show up in the exam room? Understanding, recall, a practised skill? If yes, it helped. If everything it produced stays on the screen and nothing stays in you, it hurt, however productive it felt.

This matters more in Australia than the online debate sometimes suggests, because of how senior assessment works here. Whichever state you are in, your Year 12 exams (the HSC, the VCE, the QCE and their equivalents) are sat under supervision with no AI anywhere in the room, and a large share of your final result flows from those hours. So whatever policy your school runs on AI for assignments, the exam room policy is already settled, and it is the same everywhere: you, the question and whatever you built beforehand.

And the test is not only for school. The training modules at your casual job, learning a language, the knowledge test for your learner licence: anything with an assessment at the end where the tool is not allowed in follows the same rule. The next lesson picks up the other half of the problem, because even the helpful uses depend on the model being right, and it is not always right.