How to use ChatGPT to study without it doing the work
The best way to use ChatGPT to study is to have it ask you questions, not give you answers. That's the opposite of what most of us do: paste in the lecture notes, ask for a summary, read it, and feel done. The summary is comfortable, but it's just another form of rereading, and rereading is one of the weakest ways to learn anything.
Used well, ChatGPT is a study partner with endless patience. It explains what you didn't get in a different way, gives you examples, quizzes you, and corrects you. Used badly, it does the studying for you and leaves you with the feeling that you studied.
I've been using it for a couple of years in my computer engineering degree. Here's how to use ChatGPT to study: what it's good for, what it isn't, and the exact prompts I use.
The most common mistake: asking for a summary
The first instinct is to upload the lecture PDF and type "summarize this." You get back something tidy and clear, with bullet points. You read it in five minutes and it feels like you understood the lecture.
The problem is that it did the hard part. Summarizing means deciding what matters, connecting ideas, and putting them in your own words, and that's exactly what makes a topic stick. If the chatbot does it, all you do is read the result. It's the same trap I wrote about in does rereading work: it feels like studying and leaves very little behind.
There's a second problem: summaries eat the details. In a sixty-slide deck, what ends up on the midterm is often the worked example on slide 34 that the summary left out.
What it's actually good for
ChatGPT is best at what a good classmate would do if they had unlimited time:
- Explaining what you didn't get, differently. If a definition won't click, ask for an everyday example, or to explain it like you're fifteen. You can ask for a third version, then a fourth, and nobody gets tired.
- Giving you examples. Lecture slides often have the definition and no examples. Asking for three different examples of a concept, plus one that isn't, helps you see where the boundary is.
- Quizzing you. This is where it pays off most: having it test you on a topic and correct your answers.
- Checking your explanation. You write what you understood and it tells you what's wrong or missing.
- Building a study plan. You give it the topics and the exam date, and it suggests how to split them up.
What these have in common: you do the thinking, and the chatbot reacts.
The ChatGPT study prompts I use
There's no magic formula. What matters is telling it what its job is and what it shouldn't do. These are the ones I use most, adapted to whatever course I'm on:
- To understand: "I'm studying deadlocks in Operating Systems. I don't get why all four conditions have to hold at the same time. Explain it with a real-life example, then show me what it looks like in an actual system."
- To quiz me: "Ask me questions about this topic, one at a time. Wait for my answer before moving on, correct me, and tell me what I missed. Start easy and make them harder if I'm doing well."
- To check what I know: "I'm going to explain virtual memory in my own words. Don't give me the right answer yet. Tell me which parts are right, which are wrong, and what's missing."
- To practice like the real exam: "Here's a past paper. Write new questions in the same style on the topics I give you, and don't show me the answers until I've responded."
- To plan: "I have an Operating Systems final in six days covering these eight topics. I already understand the first three. Build me a day-by-day plan, with one day for practice and one for review."
The key phrase in almost all of them is "don't give me the answer yet." Without it, the chatbot tends to explain everything up front and saves you exactly the effort you needed to put in.
ChatGPT works when you're the one doing the thinking and it only asks, corrects, or explains things a different way.
What to watch out for
ChatGPT is genuinely useful, but it has limits you should know before trusting it with an exam.
- It can be confidently wrong. Sometimes it makes up a fact, a formula, or a source, and says it in the same tone as everything else. If it doesn't match your notes, your notes win.
- It doesn't know what your professor said. It explains the topic the general way, and your professor might use different notation, a different approach, or weight things differently. Your professor writes the exam.
- It gets lost in long documents. If you upload a whole slide deck and ask about one detail, it may answer from what it knows about the topic instead of what your material says. It works better if you paste in just the part you care about.
- What you upload leaves your computer. Your notes and your professor's materials end up on another company's servers. For a lot of people that's fine, but it's worth knowing.
What a study session looks like
Here's a typical afternoon on an Operating Systems topic, CPU scheduling:
- First, the material. I go through the lecture slides for understanding, no chatbot. If a slide has three words and I don't know what they mean together, I write down the question. How I do that pass is in how to study from lecture slides.
- Then, the questions. I ask ChatGPT only what I wrote down: why round robin needs a time quantum, what happens if it's too small. I ask for examples and compare them with what the slide says.
- Close everything and explain. I write out how each algorithm works in my own words and paste it in with the checking prompt. It almost always finds something I mixed up.
- Quiz myself. Ten questions, one at a time, with corrections. I note the ones I missed.
- Two days later, I redo only the ones I missed.
The chatbot comes in the middle, not at the start. If I begin by asking it everything, I end up reading its explanations instead of my professor's.
If your material is lecture slides
Almost everything I study comes as slide decks, and that's where a chatbot hits a practical limit: it doesn't walk through the deck with you. You ask, it answers, and you go back to the next slide on your own.
That's why I built Filmina, an app that runs in your browser and explains your lecture slides and PDFs out loud, slide by slide, pointing at the part of the page it's talking about. It has a chat about the slide you're on and another one with the whole lecture that tells you which slides each answer comes from. If you want to see how the two compare, there's a side-by-side with ChatGPT.
With or without that app, the rule is the same: use AI to make you think, not to think for you.
Frequently asked questions
Is ChatGPT good for studying?
Yes, if you use it to explain what you didn't understand in a different way, to quiz you, and to check your explanations. Less so if you use it to summarize everything, because it skips the effort that actually makes you learn.
How do I get ChatGPT to quiz me?
Tell it the topic, ask it to give you questions one at a time, and to wait for your answer before moving on. Add that it should correct you, tell you what you missed, and make the questions harder if you're doing well.
Can ChatGPT be wrong about college material?
Yes. It can make up facts or formulas and state them confidently, and it doesn't know how your professor teaches the topic. If something doesn't match your material, trust your material and ask your professor.
Can I upload my lecture PDF to ChatGPT?
You can, but with long documents it tends to answer from general knowledge rather than what your file says, and the file ends up on its servers. It works better if you paste in the specific section you're studying.