Short answer: use AI to write the questions and something else to decide when you review them. Give an AI tool your own notes and ask for short recall questions, check every one against the source, then put them in a scheduler such as Anki's FSRS or a simple dated list. At each review, answer from memory before you look. Spacing and testing are the parts that make it work; AI only makes the questions quicker to produce.
What spaced repetition needs
Spaced repetition rests on two well-tested effects. The first is spacing. Cepeda and colleagues (2006) reviewed 839 assessments of distributed practice across 317 experiments and found that spreading out study sessions improves retention, with the best gap between sessions growing as the time until the test grows. The second is retrieval. Karpicke and Roediger (2008) had students learn foreign vocabulary; once a word was learned, repeatedly studying it again had no effect on delayed recall, while repeatedly testing it produced a large benefit. Students' predictions of how well they would do did not match how well they actually did.
So a working system needs three things: questions that make you retrieve, reviews spread out over days and weeks, and honesty about what you got wrong. For years the bottleneck was the first. Writing hundreds of good flashcards by hand takes longer than many students have. That is the part AI changes.
Step 1: let AI draft questions from your notes
Give the AI your own material rather than a topic name. "Make flashcards on the cardiac cycle" produces generic cards that may not match your course; pasting your lecture notes or uploading the slides produces cards on what you will be examined on.
A prompt that works:
Using only the notes below, write recall questions for spaced repetition. One fact or idea per question. Answers should be short (a word, a phrase or one sentence). Include some "why" and "how" questions, not just definitions. After each answer, quote the line from my notes that supports it. Do not add anything that is not in the notes.
Three rules make the output usable:
- One idea per card. Long answers are hard to grade honestly, so you mark yourself right when you were half right.
- Mix question types. Definitions are easy to write and easy to recall. Questions that ask for a cause, a comparison or a step in a process are closer to what exams ask.
- Keep it small. Ask for questions on one lecture at a time. A thousand cards made in an afternoon become a review pile you abandon within a week.
Our guide to making good flashcards covers card design in more detail, and the same rules apply whether you or the AI writes them.
Step 2: check every question
This is the step people skip, and the one that matters most. Spaced repetition is designed to make things stick for months. If an AI-written card has a wrong answer, the system will help you remember the mistake very well.
Checking is quicker if you asked for the supporting line from your notes. For each card, ask:
- Is the answer actually in my notes, in those words or close to them?
- Is the question specific enough to have one right answer?
- Would my lecturer or exam board phrase this the same way?
Delete anything that fails. Reading and judging the cards is itself a useful first pass through the material, but it is not a review: the reviews start when you answer from memory.
This is the problem Ace was built around. Ace (acenow.app) is a study app that turns your own lecture PDFs, slides or YouTube lectures into multiple-choice questions checked against the source. Every answer is checked against your material, questions that fail the check are dropped, and under each answer it shows the sentence from your own file that backs it. You choose which chapters to study, and streaks and a study calendar help you come back on separate days. It does not replace a spaced repetition scheduler; you still decide when to review.
Step 3: schedule the reviews
Once you have checked questions, something has to decide when each one comes back. You have two good options.
A spaced repetition app. Anki is the best-known free option. Its manual describes FSRS, the Free Spaced Repetition Scheduler, as an alternative to its older SM-2 algorithm that aims to predict forgetting more accurately; the manual says it "can help you remember more material in the same amount of time". You set a desired retention, by default 90%, meaning cards are scheduled so you have a 90% chance of recalling each one when it comes due. The manual warns that pushing retention towards 100% sharply increases the daily workload and recommends staying below 97%. Most AI flashcard tools can export cards in a format Anki imports.
A dated list or calendar. If you have a few dozen questions per topic rather than thousands, you do not need an algorithm. Review a topic the day after you learn it, then at growing gaps, and bring any topic you struggled with back sooner. Our post on the spaced repetition schedule gives intervals that suit most exams.
Either way, the review itself should be a test. Read the question, answer it out loud or on paper, then check. Rating a card you only glanced at defeats the purpose.
Can a chatbot be your spaced repetition system?
Not on its own. A general chatbot does not reliably remember what you got wrong last Tuesday or prompt you when a card is due, and each new chat starts without your history unless you supply it. It can quiz you well within one session, which makes it useful for a review you have already scheduled: paste in the questions due today and ask it to ask them one at a time.
Some study apps now combine AI-generated cards with built-in scheduling. Whichever you use, the questions to ask are the same: are the cards made from my material, can I check them against it, and does the tool bring them back at spaced intervals rather than all at once?
If you want to go further with memory tools, our guide to memorising with AI covers mnemonics and other techniques that pair well with spaced reviews.
Frequently asked questions
Can ChatGPT do spaced repetition?
It can write the questions and quiz you in a session, but it does not track when each card is due. Use it to make and run reviews, and use an app or a dated list to decide the timing.
Are AI-made flashcards as good as ones you write yourself?
They can be, if they come from your own notes and you check each one. Writing your own cards has a learning benefit of its own, so a good compromise is to let AI draft and then edit the cards yourself.
What is FSRS in Anki?
FSRS is the Free Spaced Repetition Scheduler, an alternative to Anki's older SM-2 algorithm. It uses your review history to predict when you are likely to forget each card and schedules reviews to hit a retention target you choose, 90% by default.
How many AI flashcards should I make per lecture?
Fewer than you think. Enough to cover the main ideas and the details your exam asks about, often a few dozen. Every card adds to future reviews, so quality beats quantity.
Sources
- Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed practice in verbal recall tasks: a review and quantitative synthesis. Psychological Bulletin, 132(3), 354–380. DOI
- Karpicke, J. D., & Roediger, H. L. (2008). The critical importance of retrieval for learning. Science, 319, 966–968. DOI
- Anki Manual. Deck options: FSRS and desired retention. docs.ankiweb.net



