How to do it
To figure out which side of the line you're on, ask yourself three questions before handing in anything done with the AI's help.
- Who did the intellectual work? If the AI produced the reasoning, the analysis or the text that was the point of the assignment, you handed in work that isn't yours. If instead you thought and wrote, and the AI only clarified a doubt or flagged a mistake, the work stays yours.
- What does your teacher allow? The rules aren't the same everywhere: the same use can be encouraged in one course and forbidden in another. The course's rule (often written in the syllabus or the assignment) wins over any general principle. If it isn't clear, ask beforehand, not after.
- Did you declare it, if required? Many courses ask you to indicate whether and how you used the AI, and sometimes to cite it. Using it in an allowed way but hiding it can still be a violation. When in doubt, transparency protects you.
Check: you're on the right side if you could explain and defend every part of what you hand in, and if your use of the AI is consistent with the course's rules and declared where required. If one of these is missing, stop before handing in.
A concrete example
Two students, the same assignment: an essay. Luca asks the AI to write the essay, barely rereads it and hands it in: he wouldn't be able to explain half the arguments, and the thinking isn't his. He cheated, even though the words were generated by a machine and not a classmate. Giulia, on the other hand, writes the essay herself; then she asks the AI to point out the unclear passages and the weak arguments, reviews them with her own head, and in the submission indicates that she used the AI for revision, as the course allows. Same tool, two opposite outcomes: the difference isn't having used the AI, but who did the work and with how much transparency.
When it does NOT work (and how to fix it)
If the course's rules say nothing about AI
Silence isn't permission. When the syllabus doesn't clarify, don't interpret it in your favor: ask the teacher directly what is allowed for that assignment. It's a legitimate question and it keeps you safe. Until you have an answer, stay on the safe side: use the AI to understand, not to produce what you have to hand in.
If you used the AI in an allowed way but don't know whether to cite it
When the use is permitted but you don't know whether it should be declared, declare it anyway, briefly: "I used an AI tool for X." The extra transparency almost never penalizes you, while the omission, if discovered, can cost dearly. Indicating how you used it also shows that you worked with method.
If you're afraid of being accused unfairly
AI detectors make mistakes and sometimes flag texts written by people as artificial: the risk of an unfair accusation is real. Defend yourself with the process, not with protests: keep the drafts, the notes, the document's revision history. Being able to show how you arrived at the result is the strongest proof that the work is yours, far more than any percentage from a detector.
A tip from someone who actually uses it
Adopt a personal rule stricter than the minimum required: hand in only what you could redo and explain without the AI beside you. It's a test that cuts every doubt. If you can reconstruct the reasoning, lay it out aloud, answer a question about a detail, then the work is yours and the AI was a study tool. If instead you hand in something you couldn't defend, it's copied, whatever the course's rule.
Frequently asked questions
If the AI only corrects the grammar of my text, is that cheating?
Generally no: correcting the spelling and grammar of a text you wrote yourself is comparable to using a spell-checker or having someone proofread it, and many courses allow it. Watch the threshold, though: if the "correction" becomes a rewrite that changes arguments and structure, the work stops being yours. And if the course forbids any use, its ban applies regardless.
Can teachers really tell whether I used AI?
Not infallibly with detectors, which err in both directions. But many shift the verification onto the process: they ask for spoken explanations, drafts, questions about your own text. That's where you can see who really worked. Counting on the fact that "they won't notice" is a risky bet, and in any case it doesn't change the substance: if you didn't learn, the harm you did was to yourself.