Open ChatGPT and type this: "I'm studying [topic]. Quiz me on it, one question at a time. Don't show me the answer until I've tried." Then paste in whatever you're supposed to be learning.
That changes the interaction. Instead of asking ChatGPT to explain the material, you're asking it to make you retrieve it. For studying, that difference matters.
Why Studying Feels Productive (and Isn't)
Cognitive science has an uncomfortable message about studying: the methods that feel best tend to work worst.
A 2013 review led by John Dunlosky and published in Psychological Science in the Public Interest examined ten common learning techniques (Dunlosky et al., 2013). Two of the least effective were rereading and highlighting, techniques students report relying on heavily. They feel productive because the material gets familiar as you go over it. Familiarity with a page isn't the same as being able to recall what's on it.
What works better is retrieval practice: closing the book and pulling the information out of memory. In a 2006 study in Psychological Science, Henry Roediger and Jeffrey Karpicke found that students who practiced recalling material remembered substantially more of it a week later than students who spent the same time rereading (Roediger & Karpicke, 2006). Pulling information out of memory strengthens it, and a retrieval that takes effort tends to stick better than an easy one.
The catch is that retrieval practice is hard to do alone. You need someone to ask the questions, wait while you struggle, and check the answers. A textbook won't do it. Flashcards help, but someone has to write the cards first.
ChatGPT can handle this across most subjects, as long as you give it enough context and check its answers when accuracy matters. It can also generate questions at whatever difficulty you want, whenever you want to study.
Three Ways to Study With ChatGPT
1. Have it quiz you
Quizzing tests what rereading doesn't: whether you can retrieve the information without cues.
Paste in your notes, a chapter, or a list of key concepts, then ask for questions one at a time. Answer out loud or in writing before you look at anything. Ask it to grade your answer and point out what you missed.
One at a time matters. If you ask for ten questions at once, the answers arrive with them, and it's easy to skip the attempt.
A prompt that works:
Quiz me on the material below, one question at a time. After each of my answers, tell me if I got it right and explain anything I got wrong. Don't move on until I've answered.
The method only works if you grade yourself honestly, because the benefit comes from the attempt rather than the confirmation. Write your answer down before you look, even if it's just a word or two.
2. Explain it back
Explaining exposes the gaps that familiarity hides.
The physicist Richard Feynman liked a simple test of understanding: explain the idea in plain language, as if to someone with no background, and notice where you stumble. The stumbles are the gaps.
ChatGPT is a patient audience for this. Try:
I'm going to explain how [concept] works. Ask me follow-up questions like a curious professor until you find something I can't explain clearly. Then tell me what I should review.
You explain. It asks "why?" and "what happens if...?" and "how does that square with...?" until the explanation runs out. Whatever it exposes can become your study list, and it's usually shorter and more specific than the whole chapter.
A chat is also better suited to this than a stack of pre-written flashcards: it can follow your specific confusion instead of working down a fixed list of questions.
3. Ask it to write you problems
Practice problems show you what to review next.
Understanding a concept and using it on a problem are different skills. Ask ChatGPT to generate practice problems:
Write me five practice problems on [topic], ordered from easiest to hardest. Give them to me one at a time. After I answer each one, show a full worked solution.
Do them cold, on paper if the exam will be on paper. The problems you miss usually point toward what to review next, which is more useful than another full pass over material you already know.
For subjects where an answer can be checked (math, science, coding), also ask for a fully worked, step-by-step solution, so a wrong answer turns into a lesson instead of just a score.
The Setup Prompt That Makes It Work
Whichever method you use, set the ground rules up front. A short setup like this saves a lot of correction later:
I'm studying [subject] at [level: high school, intro college, professional certification]. My exam is [multiple choice / essays / problems / oral]. Act as a study partner, not a tutor. Don't give me answers unless I ask. Push back on weak explanations. If I'm getting everything right, make the questions harder.
Four things are doing work in that prompt:
- Your level and exam format. Quizzing someone for a multiple-choice final is different from quizzing someone for an oral exam. Say which one you're facing and the questions will match it.
- "Study partner, not tutor." Left to its defaults, ChatGPT tends to explain things to you. Explaining is the tutor's job. Your job is the recalling, and the prompt needs to protect that.
- "Push back on weak explanations." This is permission to challenge you, which it won't reliably do otherwise.
- "Make the questions harder." In our experience the questions run easy at first. Difficulty that adjusts upward keeps the retrieval effortful, which is the point.
None of this is prompt engineering. It's just clear instructions in plain language, which is the prompting approach we prefer. (We wrote about why in how to write better ChatGPT prompts.)
Where It Falls Short
A study partner this cheap must have flaws. It does.
It's agreeable. Ask "did I get that right?" and you'll often get an encouraging yes wrapped around a gentle correction, or sometimes just the yes. We've watched it bless an explanation we knew was half wrong, in the same encouraging tone. Ask a harder question instead: "What's wrong or incomplete in this explanation?" Give it permission to be tough and it usually will be.
It can be wrong. For common course material it's solid, but for niche topics, recent research, or anything specialized, it can state errors with total confidence. For exam-critical material, check its questions and answers against your course materials, the same way you'd check any source. Our guide on how to verify AI output covers a simple framework for deciding when checking is worth the time.
It doesn't know your course. Your syllabus, your teacher's emphasis, the exact topics on the exam: it won't know these details unless you paste them in. The more of your own material you share, the better its questions tend to get.
It will answer if you let it. The failure mode to watch for is letting ChatGPT do the thinking. You ask a question, it gives you a polished summary, and the session feels productive without testing what you can recall. Keep the attempt on your side of the conversation.
One more honest limitation: this works for understanding and remembering material. It won't do your homework for you (well, it will, but that's the one use we'd talk you out of, and it's the subject of a different post).
FAQ
Is using ChatGPT to study cheating?
Not if you use it to practice rather than to produce. Using ChatGPT to quiz you, generate practice problems, or check your explanations is closer to hiring a tutor than copying an answer. The line is whether you're doing the thinking or outsourcing it. Writing an essay with ChatGPT and submitting it as your work is a different category, and most schools treat it as one. When in doubt, check your school's or program's AI policy.
What is active recall?
Active recall (also called retrieval practice) is the practice of pulling information out of memory instead of reviewing the material again. Answering questions before looking at the answer, explaining a concept from memory, and using flashcards are examples of it. Research including Roediger and Karpicke's 2006 study supports stronger long-term retention from retrieval practice than from rereading.
Can ChatGPT quiz me on my own notes?
Yes. Paste your notes into the chat and ask it to generate questions from them, one at a time. It can work from lecture notes, textbook chapters, slide decks, or a list of terms. The more complete the material you give it, the better the questions tend to be.
Is ChatGPT better than flashcards?
They're good at different things. Flashcards are excellent for memorizing discrete facts: vocabulary, dates, formulas. ChatGPT tends to be stronger on conceptual material, because it can ask follow-up questions, chase your specific confusion, and generate new problems on the fly. Many study sessions benefit from both: flashcards for the facts, ChatGPT for the ideas and the practice problems.
Does ChatGPT remember what we studied last time?
Not reliably between chats. A new conversation starts without the context of your previous ones, and whether details carry over depends on the memory settings in your account. The dependable approach: keep a short document with what you're studying and your setup prompt, and paste it at the start of each session.
Want to go further? See what ChatGPT can do for more uses like this, or start with our beginner's guide to AI literacy if you're building your fundamentals. For skills you perform rather than material you retain, How to Learn Any Skill Faster with AI is the skills version of this loop.