Re-reading notes feels productive. Highlighting feels productive. They aren't. Decades of learning science show that passive review is the least effective way to prepare for an exam, while the methods that feel hardest — retrieving answers from memory, explaining ideas in your own words, being tested — are the ones that actually move grades.

AI is perfect for the hard methods, because it can generate infinite practice material from your notes, quiz you adaptively, and never get tired of your wrong answers. But only if you prompt it for retrieval, not delivery.

These 20 prompts do that. Use them in ChatGPT, Claude, Gemini, or any assistant the week before and the weeks leading up to exams.

Planning: where to spend your hours

1. The personalized study plan. Our Study Plan Generator does this formally — give it your subject, exam date, hours per week, and current level and it builds a week-by-week plan. The manual version:

I have an exam in [subject] on [date]. I can study [hours] per week. My current level: [what you know vs what's weak]. Build a weekly plan that front-loads my weakest topics, schedules active recall every 3 days, and leaves the last 20% for practice exams. Include one rest day per week.

The plan should tell you what to skip, not just what to cover. If it doesn't prioritize, tell it "cut 30% — what goes?"

2. Study guide from messy notes. Paste your actual notes, not a textbook chapter. Our Study Guide Generator was built for this:

Turn these notes into a study guide: [notes]. Structure it as: key definitions (one sentence each), cause-effect chains, common confusion pairs with how to tell them apart, and 5 practice questions that require application, not recall.

A guide that only summarizes is a prettier version of your notes. You need application questions — that's where the exam lives.

3. The 80/20 audit.

Here is what will be on the exam: [syllabus / exam description]. Here is how I've been spending study time: [breakdown]. Tell me where my effort is misaligned with likely point distribution. Which 20% of topics will produce 80% of the points, and am I under-investing there?

Students systematically over-study what they already know because it feels good. This prompt fights that bias.

4. Weakness map.

Quiz me on [topic] with 6 questions, one at a time. After each answer, tell me if I'm right, what's missing, and tag the sub-topic. At the end, give me a table: sub-topic, accuracy, and what to study next — ordered by lowest accuracy.

One-at-a-time matters. Batch quizzes let you pattern-match; sequential quizzes force true recall. Use our Socratic Tutor prompt for a more conversational version that probes gaps instead of just scoring.

Memorization that sticks: flashcards and recall

5. Flashcards that actually work. Our Flashcards From Notes prompt generates them, but quality beats quantity:

From these notes: [notes], create 12 flashcards. Rules: one fact per card, question requires producing the answer (not recognizing it), include 4 cards that ask "why" or "how," not just "what." Format as Q: / A:. Mark the 4 most important cards with ★.

Recognition cards ("Which of these...") feel easy and test nothing. Production cards ("Explain why...") feel hard and build durable memory.

6. The Leitner-style multiplier.

Take these 8 flashcards: [cards]. Quiz me on them in random order, one at a time. If I get one right, drop it. If I get one wrong, put it back in the queue and re-ask it later in this session with a slightly different wording.

This simulates spaced repetition inside a single session. The rewording prevents you from memorizing the question, not the concept.

7. Reverse flashcards.

For this concept: [concept], create 3 flashcards that go backwards — give me the application and ask me for the principle. Example: instead of "What is confirmation bias?" ask "A doctor orders extra tests that confirm their initial diagnosis and ignores contradicting labs. What bias is this?"

Exams rarely ask "define X." They give you a situation and ask you to identify X. Reverse cards train that direction.

8. The blurting check.

Without looking, I'll explain [topic] from memory. After I do, compare my explanation to [notes / textbook summary] and list: what I got right, what I misstated, and what I omitted. Grade omissions as critical vs minor.

Blurting — writing everything you can recall before checking — is the most honest diagnostic. AI as grader makes it scalable.

Understanding: not just memorizing

9. The Feynman drill. Our Feynman Technique Tutor structures this as a loop — explain simply, get quizzed, re-teach where you're weak:

I want to learn [topic] using the Feynman technique. Have me explain it in simple terms, then ask a question that probes where my explanation is vague. When I answer, tell me if I actually understand or am hiding behind jargon. One question at a time.

If you can't explain it to a 12-year-old, you don't know it. The AI's job is to catch the jargon that masks a gap.

10. Why and how, not just what.

Explain [concept] by answering three questions: What is it? How does it work step by step? Why does it work that way and not another way? Use an everyday analogy, then say where the analogy breaks.

"What" gets you recognition. "Why that way and not another" gets you transfer — applying the concept to a novel exam question.

11. The confusion pair.

I keep mixing up [concept A] and [concept B]. Explain both in parallel columns: definition, example, and a scenario where you'd use one but not the other. End with a 2-question quiz where each answer is A or B and the wrong answer is plausible.

Confusion pairs are where exams harvest points. If you can articulate the boundary, you've mastered both.

12. Explain at three levels. Our Explain at Three Levels prompt does this in one shot, and our Explain a Hard Topic is the simpler version:

Explain [concept] at three levels: one sentence for a 10-year-old, one paragraph for a first-year student, and one paragraph for an expert that includes the nuance the simpler versions omit. Then quiz me at the middle level.

The expert nuance tells you what the exam's "challenge question" will target.

Practice: training for the exam, not the notes

13. Practice exam generator.

Create a practice exam for [subject, exam format, time limit]. Include the same mix of question types as the real exam: [multiple choice / short answer / essay]. Make 30% of questions application-based (new scenario, same principle) and 10% harder than the real exam. Provide an answer key with brief reasoning for each.

Harder-than-real practice is deliberate overtraining. If your practice exam feels comfortable, it's too easy.

14. Timed essay simulation.

Give me one essay prompt for [topic] as it would appear on the exam. I'll write for 25 minutes. Then grade my essay on thesis clarity, use of evidence, and structure — not grammar. Give one thing to fix first.

Time pressure changes what you produce. Practice under it, not without it.

15. The distractor decoder.

Here's a multiple-choice question I got wrong: [question + my answer + correct answer]. Explain why the distractor I chose was designed to be tempting, what misconception it preys on, and how to avoid it next time. Then create a new question that tests the same misconception differently.

Wrong answers are designed. Understanding the design teaches you to see the trap, not just the fact.

16. Past paper reverse-engineering.

Here are 3 past exam questions for [subject]: [questions]. Deduce what the examiner values — is it definitions, applications, comparisons, or synthesis? Predict 3 questions likely to appear on the next exam and explain your reasoning.

Examiners have tendencies. Pattern-matching them is not cheating — it's studying the game.

The week before: sharpening, not cramming

17. The mixed-queue shuffle.

Quiz me on these topics in random, interleaved order: [topics]. One question at a time, no two consecutive questions from the same topic. After each answer, give feedback, then ask the next.

Interleaving feels harder than blocking (studying one topic at a time) and performs better. Exams interleave — so should practice.

18. Error log review.

Here are the questions I've gotten wrong this week: [list with my wrong answers]. Categorize my errors: misread the question, knew the concept but applied wrong, or didn't know the concept. For the biggest category, give me a specific fix I can use on exam day.

Most error logs reveal one dominant failure mode. Fixing that mode buys more points than learning new facts.

19. The teach-back final.

Let me teach you [topic] as if you're a classmate who missed the lecture. Interrupt me if I skip a step, use jargon without defining it, or say something inaccurate. At the end, rate my teaching as: could ace, could pass, or would fail — and why.

Teaching is the highest form of retrieval. If you can teach it cold, you can write it under pressure.

20. Night-before calibration.

Tomorrow's exam is [format, duration, weight]. I've studied [topics] to [level]. Give me: a 30-minute light review plan for tonight (no new material), a sleep-and-morning checklist, and the 3 most common mistakes students make on this type of exam. Be brief.

Night-before cramming hurts more than it helps. A constrained review protects sleep — the highest-ROI exam input.

The principle behind all 20

Every prompt here forces you to produce rather than consume. That is the whole science of exam preparation in one line. The discomfort you feel when you can't answer a good quiz question is the feeling of learning actually happening.

Start with Study Plan Generator and Study Guide Generator to set direction, then live in Feynman Technique Tutor and Socratic Tutor for the daily work. Browse the full Promptigo library — every education prompt is free to copy with variables you can fill in.