How to Use AI to Study Without Outsourcing Your Learning
Practical guide
By MotivLearn
Published
You ask an AI tool to explain a difficult topic. The answer is clear. The example makes sense. You move on feeling that you understand it.
The next day, a similar question appears without the explanation beside it. You cannot find the first step.
A useful AI study session should leave you able to do something without the tool that you could not do before. Getting through the assignment and learning the skill can overlap, but one does not automatically prove the other.
Here is a routine for keeping your own thinking inside the process: attempt the task, ask for the smallest useful hint, explain your next step, and then try a related task independently.
What the research actually found
In a 2025 PNAS study by Hamsa Bastani and colleagues, nearly a thousand high-school mathematics students in Turkey used either ordinary course resources, a general GPT-4-based assistant, or a specially designed AI tutor. Both AI groups performed better during supported practice. When the tools were removed, however, the general-assistant group scored lower than the comparison group. The specially designed tutor largely avoided that disadvantage, without showing a clear advantage on the independent exam.
That is a finding about particular tools, mathematics lessons, and assessments. It does not establish that every AI tool harms learning, or that a good prompt recreates the study's tutor. The tutor also used teacher-provided information about the problems and common mistakes.
The practical question for your own studying is narrower: when the support disappears, what can you still explain or solve?
Start by naming the skill you want to keep
Before opening the chat, finish this sentence:
By the end of this session, I want to be able to ______ without looking at an answer.
Make the target observable. “Understand percentages” is difficult to check. “Calculate a percentage increase and explain which number belongs in the denominator” gives you a task you can attempt.
For a language lesson, your target might be writing five sentences using a new tense. For a work course, it might be explaining a concept to a colleague. Choose a target small enough to practise today.
If you need ideas for making the target testable, use our guide to turning study notes into practice questions.
Use a four-step study loop
1. Make an attempt before asking for the solution
Write what you know, what you tried, and where the attempt stopped. A diagram, incomplete calculation, or rough explanation is enough to begin.
This gives the conversation a specific problem to address. It also leaves a record of what you could do before receiving help.
Try this prompt:
I am practising [skill]. Here is the question and my attempt: [work]. I am stuck at [step]. Ask me one question or give one small hint. Please wait for my response before showing more.
Treat this as a request you must monitor. A tool may still reveal too much or make an error.
2. Ask for the smallest useful help
If a hint gets you moving, stop reading and work. If it does not, ask for an explanation of the particular step you cannot justify.
| What you need | A useful request | What to do yourself afterward |
|---|---|---|
| A starting point | “Which concept should I consider first?” | Choose and justify the first step. |
| Help with an error | “Identify the first incorrect step in my attempt.” | Repair that step and continue. |
| A clearer explanation | “Explain this relationship using a different example.” | Explain the original relationship in your own words. |
| More practice | “Create a similar question with different numbers. Withhold the answer.” | Solve it before checking. |
An AI-generated answer key is still an unverified answer key. Check important steps against your course material, a reliable reference, or your teacher.
3. Explain why the step works
After the hint, write a short explanation without copying the chat. Include the rule you used and why it applies here.
For example: “I divided by the original price because I am measuring the change relative to the starting value.” If you can only repeat the calculation, return to the meaning of the quantities.
4. Close the tool and try again
Use a fresh question that requires the same idea. Keep the answer hidden until you have committed to an attempt.
Research on retrieval practice by Henry Roediger and Jeffrey Karpicke found that recalling studied material improved delayed retention compared with repeated study in their experiments. That research did not test this AI routine; it supports making room for independent retrieval rather than judging learning only by how familiar an explanation feels.
Return to the skill in a later session as well. Immediate success is useful evidence, but it cannot tell you everything about what you will remember next week.
A worked example: percentage increase
This is an illustrative study conversation, not a report of a learner's measured results.
A price rises from $40 to $50. Your first attempt is (50 − 40) ÷ 50 = 20%.
Instead of asking for the answer, ask which quantity should be the reference point. The useful idea is that an increase is measured relative to the starting price.
You repair the calculation: (50 − 40) ÷ 40 = 25%. Then you explain why $40 belongs in the denominator.
For the independent check, close the conversation and calculate the increase from $60 to $75. The increase is $15 relative to $60, so it is also 25%. You should be able to show the reasoning, rather than recognize the number.
When this routine needs changing
If you cannot make any attempt, study one trustworthy worked example first, then try a closely related problem. If the AI keeps giving away answers, use your textbook exercises or ask a person for help. If explanations conflict, bring the disagreement to the original source or your instructor instead of collecting more confident guesses.
Follow your course's rules for AI use, particularly for assessed work. Do not put private student information or confidential work material into a tool unless you are permitted to do so.
Try one small session
Choose one concept and one practice question. Write your attempt, request one hint, and finish the question. Then close the tool and try a similar question.
Record three things: what you managed independently, what still required help, and what you will practise next. You can keep an optional note in your device notebook.
The most useful output from the session is evidence of your own next step: a calculation you can explain, a sentence you can construct, or a problem you can begin without waiting for the chat.
Sources and scope
- Bastani and colleagues, Generative AI without guardrails can harm learning: Evidence from high school mathematics, PNAS, published June 25, 2025. A specific classroom experiment, not a comparison of every AI learning product.
- Roediger and Karpicke, Test-enhanced learning: Taking memory tests improves long-term retention, Psychological Science, 2006. Experiments on retrieval and retention, not an evaluation of AI.
The prompts, table, and worked example above are MotivLearn's practical suggestions. They are not a separately validated teaching intervention.
Spotted an error or a source that needs updating? Contact MotivLearn about this article.
Find your next small step
What’s getting in your way?
Reflect on something that matters to you—learning, a hobby, a relationship, daily life, or work—and choose a small seven-day experiment.
Related Articles
Turn Your Study Notes into Useful Practice Questions
Make your notes easier to use by turning key ideas into questions, examples, and short practice tasks.
The Science of Effective Learning: Memory Techniques That Work
Discover evidence-based learning methods that can dramatically improve your ability to retain and recall information.
The Art of Deep Work: Achieving Focus in a Distracted World
Master the techniques of deep work to produce at your peak level and create extraordinary value in our distracted world.