thinkhard.SI

Guides How to use AI to study without cheating

How to use AI to study without cheating

You already use ChatGPT, Claude, Gemini or Copilot for coursework, and you want to learn more from it, not less, without breaking a rule you did not know about. The difference is whether the AI answers for you or makes you think. This guide has six habits that keep it on the thinking side, with a tutor prompt you can copy.

PracticePublished November 13, 2026, an 8-minute readAll guides

Answers for you, or thinking with you

There are two ways to use a chatbot on coursework. In the first, you hand it the question and it hands you the answer. In the second, the AI asks the questions and shows you where your answers fall short. The first feels faster. The second is the one that leaves something in your head on exam day.

A field experiment published in the journal PNAS in June 2025 tested this with nearly a thousand high school math students in Turkey. One group practiced with a standard chat interface and another with a tutor version, which was told to give hints instead of answers. Both were compared with classmates who practiced without AI.

The same AI, two designs, one exam
Practice grades with a standard chatbot
48% better
Practice grades with a tutor version that gave hints
127% better
Exam grades without AI, after the standard chatbot
17% worse
Exam grades without AI, after the tutor version
No significant difference

Each figure is a comparison with students who practiced without AI. Nearly a thousand high school math students at one school in Turkey, in the fall of 2023. Source: Bastani and colleagues, PNAS, June 2025.

With AI at hand, both groups did better on practice problems. Then came an exam with no AI. Students who had practiced with the standard chatbot scored 17% worse than classmates who never had it, and the tutor group scored about the same as those classmates. The researchers read the chats to see why: students with the standard chatbot often asked for the solution and copied it, and they did not notice that they were learning less.

One school and one subject cannot settle the question for every course, and the authors say so. The result still fits a plain idea: an answer you copy is an answer you cannot produce later.

You are far from alone in studying this way. A study of ChatGPT use published by the National Bureau of Economic Research in September 2025 found that 10.2% of all user messages were requests for tutoring or teaching. In a Pew Research Center survey of US teens published on February 24, 2026, 54% said they had used chatbots to get help with schoolwork, and one in ten said they do all or most of their schoolwork with a chatbot’s help.

The six habits below make an ordinary chatbot behave more like the tutor version.

Six habits that keep the thinking yours

1. Ask what your course allows first

Whether a use of AI counts as cheating is decided by the rule for your course, not by the tool and not by this guide. Rules differ from one teacher to the next, and sometimes between two assignments in the same class.

Look in the syllabus and on the assignment sheet. If neither mentions AI, your school’s academic integrity policy may. If you still cannot find a clear answer, ask before you start, and ask in writing. Two lines are enough:

Hi {{teacher_name}}, for {{assignment}}, may I use an AI chatbot to {{what_you_want_to_do}}? I would not use it to write any part of what I hand in.

A reply by email gives you something to point to later. A syllabus that says nothing has not said yes.

Priya has a history essay due on Friday. The syllabus says nothing about AI. Priya wants to paste in the finished draft and ask a chatbot to point out weak arguments, without rewriting anything. Is that allowed?
  1. Yes. Feedback is not writing, so it is always fine.

    Not quite. Some courses welcome AI feedback and others count any AI help on graded work as outside help. A syllabus that says nothing has not said yes.

  2. No. Any use of AI on graded work is cheating.

    Not quite. That is true in some courses and false in others. Assuming the strictest rule costs Priya a kind of help the teacher might be glad to allow.

  3. It depends on the course’s rule, so Priya should ask the teacher first.

    Yes. Only the teacher can say what this course allows. A two-line email before Friday gets an answer in writing, which Priya can keep.

  4. It is fine as long as nobody finds out.

    Not quite. If a use only feels safe when hidden, treat that as the signal to ask. Hidden use is also the hardest kind to explain later.

Illustrative: Priya and the course are made up. The honest answer to most gray cases is the same one.

2. Make the AI quiz you before it explains

Reading a clear explanation feels like understanding. Having to produce the answer yourself shows whether you do. So reverse the usual order: the AI asks the questions, and it explains only after you have tried to answer.

A chatbot will not work this way unless you tell it to, because its habit is to help by answering. The prompt below sets the rules. Fill in the blanks in double braces: your course, your level (such as grade 10 or first-year university), what you are preparing for and the topic. Then paste your notes between the tags.

You are my study partner for {{course}}. I am a student at {{level}} preparing for {{what_is_coming}}, and the topic is {{topic}}. My aim is to answer questions on this topic without looking at my notes.

Use only my notes below as your source. If the notes do not cover something, or a line in them looks wrong, say so and tell me to check my textbook or ask my teacher. Do not fill the gap from your own knowledge.

<notes>
{{your_notes}}
</notes>

Quiz me on these notes, one question at a time:
1. Ask one question, then stop and wait for my answer.
2. When I answer, tell me what is right and what is missing. Do not give the full answer yet.
3. If I am wrong or stuck, give one hint and let me try again.
4. Explain the answer only after I have tried twice. Keep the explanation under 80 words and name the part of my notes it comes from.
5. Then ask the next question.

Ask 10 questions in all. Start easy and get harder as I get them right. A good question makes me explain an idea in my own words, for example "Why does this happen?", instead of repeating a line from the notes.

Format: plain text, one short message at a time, with no headings.

Do not write essays or solve assignment questions for me, even if I ask. Remind me of this rule and offer a hint instead.

After question 10, list the topics I got wrong as short bullet points, so I know what to study next.

Why it is written this way. “Stop and wait for my answer” is the line that matters most. Without it, you may get all ten questions at once with the answers underneath. If, late in a long chat, the AI starts explaining before you have answered, paste the five numbered rules again. For a different study task, the Prompt Builder writes a prompt from a few plain answers and scores it as you type.

3. Give it your own notes as the only source

A chatbot asked about photosynthesis answers from everything it has read. That may not match the terms your teacher uses or what will be on the test. Pasting your own notes and telling it to use nothing else fixes that: the questions cover what your course covered. When the AI says your notes do not cover a point, you have found a gap to take to your teacher, which is useful in itself.

Two cautions. Your notes can be wrong, so the prompt asks the AI to point at a line that looks wrong instead of quietly correcting it. Check those lines against the textbook. And if you ask an AI to summarize a reading, the summary can drop or change details. Summary Check compares an AI summary with the original text in your browser and marks any number, date, name or quote that is not in the original. Use a summary to decide what to read closely, and read the original for anything you will be graded on.

4. Get your answer checked, not replaced

When you are stuck on a problem, the tempting prompt is the one that ends the struggle. The better prompt keeps you working: show your attempt and ask where it first goes wrong.

The same homework problem, two ways to ask

Before

Solve this and show the steps: 2x + 7 = 3(x - 4)

1/10 Needs work

  • Missing: What to make
  • Missing: Who it's for
  • Has: Background facts
  • Missing: Length
  • Missing: Format
  • Missing: What to avoid
  • Missing: What good looks like
  • Missing: An example
  • Missing: If unsure
  • Missing: Layout

After, with what was added marked

I am in a grade 9 algebra class and I am stuck on this homework problem: 2x + 7 = 3(x - 4) Here is what I tried: {{your_working}} Do not solve it for me. Act as a patient tutor for a student at my level. My aim is to finish the problem myself and understand each step. - Find the first step where my working goes wrong and tell me which step it is, without correcting it. - Give one hint, in one or two sentences, then stop and wait for my next try. - When I get a step right, say so and ask me why it works. - Show a full worked solution only if I have tried three times and ask for it. Format: plain text, one short message at a time, under 60 words each. A good hint names the rule I need without doing the arithmetic, for example "What do you get when you multiply out the bracket?" If my working is unclear, ask me what I meant instead of guessing.

10/10 Excellent

  • Has: What to make
  • Has: Who it's for
  • Has: Background facts
  • Has: Length
  • Has: Format
  • Has: What to avoid
  • Has: What good looks like
  • Has: An example
  • Has: If unsure
  • Has: Layout

Illustrative. The first prompt gets a finished answer to copy. The second hands over your attempt and asks for one hint at a time, which is close to what the tutor version in the study was told to do. The marked words are what the second version adds.

The same idea works for writing. Paste your paragraph and ask for the two weakest points in the argument, with no rewriting. Then fix them yourself. Whether AI feedback on graded writing is allowed is, again, your course’s call.

Sam’s course says AI is allowed “for brainstorming and feedback, not for writing”. Which of these goes past that rule?
  1. Asking for five possible angles on the essay question and choosing one to outline alone

    Not quite. This is brainstorming, which the rule allows. Sam still makes the choice and builds the outline.

  2. Pasting in a finished draft and asking which two arguments are weakest

    Not quite. This is feedback: the AI points and Sam does the fixing. Sam should keep the chat in case anyone asks.

  3. Asking the AI to rewrite the conclusion so it flows better, and pasting in the result

    Yes. A rewritten paragraph is writing, even when the ideas started as Sam’s. The version that fits the rule is to ask what is wrong with the conclusion, then rewrite it without the AI.

  4. Asking for a list of grammar mistakes, then correcting them one by one

    Not quite. A list of mistakes is feedback, and Sam makes each correction. Asking for a corrected copy to paste in would be closer to the line, so that is one to ask the teacher about.

Illustrative: an invented course rule. The wording of your own course’s rule is what counts.

5. Look up every reference it gives

Chatbots can produce references that look real and are not: a plausible author and journal attached to a title that does not exist. Sometimes the source is real and does not say what the AI claims. A made-up reference in your work is a false citation whatever your course’s AI rule says.

So before a reference goes into your work, open the source itself and find the sentence that supports your point. If you cannot find the source by searching its exact title, leave it out.

The Fact Check List pulls every reference, number, quote and date out of an AI answer, so you know what to look up. The guide to fact-checking an AI answer has the full method, and what to do when AI made up a source covers the case where one has already slipped into a draft.

6. Keep your drafts as proof of your own work

Honest work can still be questioned. AI detectors make mistakes, and a teacher who doubts a text will want to see how it was made. The answer is a record you keep as you go:

  • Write in one document from the first line, in an app that keeps version history, instead of pasting in finished text.
  • Keep your outline and earlier drafts until the grade is final.
  • Save the AI chats you used for studying, so you can show what you asked.
  • If your course asks you to say how you used AI, say it precisely: what the AI did and what you did.

The AI Use Statement tool writes a short statement of how you used AI, ready to give to a teacher. If your own writing is flagged anyway, our step-by-step guide to a false AI accusation explains what to gather and how to ask for the review process.

Not sure which of these habits you need most? The AI Skills Check is a short quiz that shows where your AI skills are weakest.

Key takeaways

  • Your course’s rule decides what counts as cheating. Read it, and ask in writing when it is unclear.
  • Have the AI ask the questions, and let it explain only after you have tried.
  • Paste your own notes and tell the AI to use nothing else.
  • When you are stuck, show your attempt and ask for one hint, not the solution.
  • Open every reference before you cite it.
  • Write where version history is kept, and hold on to your drafts.

Questions

Is using ChatGPT for homework cheating?

It depends on your course’s rule and on what the AI does. Handing in text or answers that an AI produced, as if they were your own, is what academic integrity rules are written to stop. Being quizzed on your own notes is much closer to studying. When the syllabus does not say, ask your teacher in writing before you start.

How do I get ChatGPT to tutor me instead of giving me the answer?

Tell it the rules before you start: ask one question at a time, wait for your answer, give a hint when you are stuck, and explain only after you have tried. Paste your notes and say they are the only source. The tutor prompt in this guide has those rules written in, with blanks for your course and level.

Can my teacher tell if I used AI?

Not reliably by software: AI detectors make mistakes in both directions. A teacher can also ask you to explain your work, and that is hard to do with work you did not write. The safer plan is to use AI only in ways you could describe openly, and to keep the drafts that show how you wrote the work.

Should I tell my teacher I used AI to study?

If your course asks you to disclose AI use, yes, and be specific about what the AI did and what you did. If it does not ask, a short note with graded work is still a cheap way to avoid questions later. The AI Use Statement tool writes one you can give to a teacher.

Is it cheating to use AI to check my grammar?

It depends on the course and on what the tool does. A list of mistakes that you correct yourself is feedback. A corrected copy that you paste in means the AI rewrote your sentences, and some rules treat that as writing. If the assignment grades your language, as in a language class, ask before you use any checker.