AI for university students: the use that raises your grade and the use that ends it

I will state the line at the start, because most articles circle it without saying it:

If you submit something you cannot explain, you cheated. Regardless of the tool, and regardless of the policy.

That line is clearer than any regulation, and you know which side of it you are on before any professor does.

Uses that deepen understanding#

Explanation at two levels. "Explain this concept as if I were in secondary school, then as if I were in third year." Moving between the two is understanding itself.

Self-testing. Give it the material, ask for twenty questions with no answers, answer them, then ask for marking. This is the most effective revision method in the educational research literature, and it just became easy to do.

The opponent. "Write the strongest possible attack on my argument." A student who has defended their thesis against a tireless opponent walks into the viva prepared.

Structure, not writing. Give it your scattered notes and ask for an outline. The ideas are yours and the ordering is assistance, which sits inside the line.

Getting through a hard paper. "Summarise this in ten lines, then explain figure three to me." See deep research.

Uses that end your degree#

Text you submit and do not understand. The clearest case.

Invented references. Models generate references that look real, with authors and page numbers that do not exist. This is caught in a minute and costs a semester. Check every reference yourself, without exception.

Generated data or results. In empirical work this is fabrication, not assistance.

Translating a paper and claiming it. Even if you translated it yourself, the attribution is the problem.

A complete revision system#

"Do not be encouraging" matters. These models lean toward flattery, and flattery before an exam is harmful.

What this means for you#

If you study: use it to test you, not to answer for you. The first raises your grade; the second raises it this term and ends you the next.

If you are in your final year: see graduation projects that impress.

If you teach: students are already using it. The only useful question is whether your assessment measures what you meant to measure.

In closing#

The tool multiplies whoever understands and exposes whoever does not. The difference is not the tool but what you asked it for: to explain to you, or to answer for you.

Try this today: take a topic you think you have understood and ask for ten hard questions on it. The result will surprise you in one direction or the other.

Common questions#

When does using AI count as cheating at university?

When you submit something you cannot explain. The practical test is to close everything and explain your work aloud for two minutes; if you can, it is yours whatever you used.

What is the best way to revise with AI?

Self-testing: ask for questions without answers, answer them, then ask for strict marking that names the misunderstanding behind each error rather than only the correct answer.

What is the most dangerous mistake students make?

Invented references. Models produce citations that look real with authors and page numbers that do not exist, and a professor catches one in thirty seconds. Verify every reference yourself.

Can I ask it to structure my ideas?

Yes. The ideas are yours and the ordering is assistance, which is inside the line. The problem starts when the ideas themselves stop being yours.