Revising with AI, honestly: active recall, Feynman, quizzes

Using AI to revise is tempting: it can read your course material in seconds and hand you back a tidy summary. The problem is that reading a well-written summary creates a feeling of mastery that doesn’t hold up against a real exam question. It’s not cheating, just a method that doesn’t work very well: passively rereading a text is one of the least effective ways to make knowledge stick in memory.

What actually works, on the other hand, has been documented in cognitive science for a long time. Here are three well-established methods, and how an AI tool can help you apply them without ever doing the memorization work in your place.

Active recall: testing yourself beats rereading

The principle is simple to state and yet counterintuitive: it’s better to try to remember something, even if you get it wrong, than to reread it one more time. An educational resource from the Académie de Guadeloupe (a regional education authority) on cognitive-science-based practices sums it up this way: training regularly with tests improves long-term retention more than passively rereading your course notes.

In practice, this changes how you work through a chapter. Instead of rereading your notes a third time before an exam, you close the notebook and try to answer a question without any support, then check afterward. The same resource actually recommends rephrasing your course material as questions without the support material, then checking afterward, rather than simply rereading.

An AI can help here in a very concrete way: generating questions from your course material, asking them one at a time, and only giving the correction after your answer. That’s the difference between a tool that helps you test yourself, and a tool that just recites a summary you’ll passively reread.

The Feynman technique: explaining to check you’ve understood

The Feynman technique starts from a simple idea: if you can’t explain a concept in simple words, you haven’t really understood it yet, even if you think you know it. The principle is to explain the concept out loud or in writing, as if presenting it to someone who doesn’t know it, then spot where you get stuck, where you use jargon without defining it, or where your explanation is wrong.

This is exactly where AI can play a useful role, as long as it doesn’t do the work in your place. A good Feynman-technique skill should listen to you explain a concept in your own words, spot the gaps and undefined jargon, then ask you follow-up questions — rather than immediately giving you the full explanation. The goal isn’t to get a neat, ready-made explanation: it’s to discover, through your own successive attempts, what you don’t yet master.

Spacing: several short sessions beat one long one

The third principle concerns the timing of your revision rather than its content. According to researcher Shana Carpenter, cited on the reference site retrievalpractice.org, there’s no single, universal optimal spacing between two revision sessions: spacing is flexible, and that’s exactly what makes it a practical tool you can adapt to your schedule.

What is well established, though, is the comparison between cramming and spreading things out over time: cramming (reviewing everything at once, right before the exam) improves very short-term memorization, but the information is then forgotten quickly. Spreading the same amount of revision over several days or weeks, for an equivalent total time, noticeably improves long-term retention and your ability to reuse the knowledge in a different context. The general rule suggested is to let enough time pass between two revision sessions that the information is no longer fresh in memory — which can mean ten minutes or several days, depending on what you’re revising.

Translated into practical use: three hours of revision crammed in the night before an exam are worth noticeably less than the same three hours spread over six days, at half an hour a day.

What an AI should never do in your place

These three methods share one thing in common: they require that you be the one making the effort of retrieval, explanation, or answering. An AI that hands you the perfect sheet directly, the explanation already chewed over, or the answer before you’ve even tried to respond, short-circuits exactly the mechanism that makes these methods work. The useful role for an AI tool here isn’t to replace you, but to play the part of the examiner or the classmate you’re explaining things to: the one who asks the questions, spots the gaps, and sends you back to the effort — not the one who answers in your place.

One last habit to keep, whatever tool you use to revise: check your institution’s own AI usage rules, especially if certain exercises or assessments explicitly regulate its use.

Key takeaways

• Testing yourself (active recall) reinforces long-term memorization better than simply rereading your course material.
• The Feynman technique reveals gaps in understanding by forcing you to explain a concept in your own words.
• There’s no universal optimal spacing, but spreading revision over several days is noticeably more effective than cramming, for an equivalent total time.
• A useful AI revision tool asks questions and waits for your answer; it should never give the correction or explanation before you’ve tried.
• Always check your institution’s AI usage rules before relying on one for graded work.

This principle of prompting you further without giving the answer right away is exactly what the methode-feynman skill applies.