The most common way students use ChatGPT for studying is "summarise this lecture". It's also close to the least useful.
Reading a summary is rereading a shorter version. It looks familiar, so it feels like learning, and mostly it isn't. What builds memory is retrieval: trying to pull the answer out of your own head. So the trick with any chatbot (ChatGPT, Claude, Gemini) is to flip the direction. Make it ask you. Paste in your own material, tell it to stick to that, and have it quiz you one question at a time. Then check anything important against your lecture, because it can be confidently wrong. Why retrieval matters.
Prompts that actually help
Quiz me, one question at a time
Here are my lecture notes. Ask me one question at a time about them. Wait for my answer, tell me if I'm right and why, then ask the next one. Only use what's in these notes.
Explain it three ways
I don't get [concept]. Explain it simply, then with an analogy, then with a real example. Then ask me one question to check I've got it.
Find my gaps
Here's my explanation of [topic] in my own words. What did I get wrong or leave out?
Make flashcards properly
Make flashcards from these notes. One idea per card, phrased as questions, short answers. Don't add anything that isn't in the notes.
Exam-style practice
Write 5 exam-style questions on this material for a [year and subject] exam. Don't show the answers until I've tried all five.
Plan the week
My exams are on [dates]. I can study [hours] on these days: [list]. Spread each subject across several days and finish with a light review the day before each exam.
One more, if you're getting multiple choice from it: add "make every option the same length and detail, make each wrong answer a common misconception, and randomise which letter is correct." When we checked AI-written quizzes in our own app, the right answer was option B 72% of the time. What we found.
Where it falls short
- It can be wrong with total confidence, especially on specifics like drug doses, dates, formulas and citations. Check against your lecture or textbook.
- It doesn't know your course unless you paste it in. Otherwise it answers from general knowledge, which may not match how your lecturer taught it or how they'll assess it.
- It doesn't remember for you. Nothing tracks which questions you got wrong or brings them back in three days. Why that matters.
- Every session starts from a blank box. You re-paste, re-prompt and re-explain.
- Academic integrity. Using AI to learn is different from using it to write assessed work. Follow your uni's rules.
When a study tool is the better fit
If you find yourself pasting the same lecture into a chatbot every week and asking for flashcards and a quiz, that's the job study tools are built for. The Study Mill, which I build, takes each week's slides, PDFs or recordings and builds a lesson, flashcards, a quiz with written answers that get marked, and an audio walkthrough, all limited to your material. Then it keeps track of what you get wrong and brings it back later. There's a tutor in there too, for "explain that differently".
Use whatever you'll stick with. The method matters more than the tool: questions over summaries, spaced over crammed, your material over generic.
Try it
Upload one lecture and compare it with what you'd get from a chatbot. Free, about a minute.
More: the best AI flashcard makers · NotebookLM vs study tools
