"No matter how many times I review, I can't remember." "The knowledge I crammed before the exam vanishes immediately." AI-scheduled spaced repetition, built on cognitive science, addresses these common learner frustrations. In Memly's internal study (n=648, 2025), learners using its flashcard learning system scored 74% higher on a retest than with traditional repetition.
74% is the kind of number that invites suspicion, so here are its conditions: it comes from an internal study, not a peer-reviewed trial, and it measures retest accuracy against traditional repetition.
Why Traditional Memorization Is Inefficient
Conventional study methods fail for a measurable reason. According to the "forgetting curve" discovered by German psychologist Hermann Ebbinghaus in 1885, humans forget learned information at a surprisingly rapid rate.
| Time Elapsed | Retention Rate | Forgetting Rate |
|---|---|---|
| After 20 minutes | 58% | 42% |
| After 1 hour | 44% | 56% |
| After 1 day | 33% | 67% |
| After 1 week | 25% | 75% |
| After 1 month | 21% | 79% |

The first row is the one that stings. 42% is gone in twenty minutes, before the desk is even packed up, and about 80% has followed within a month. Summary notes and a second pass through the textbook are more input, and the curve describes what happens to input.
How Spaced Repetition Overcomes the Forgetting Curve
The most effective method for combating the forgetting curve, established in cognitive science, is spaced repetition.
It gradually extends the intervals between reviews. Reviewing at the "optimal moment" just as memory begins to fade is what lifts retention.
Spaced repetition is one of the most reproducible evidence-based learning techniques in cognitive psychology, grounded in more than a century of research on long-term memory formation.
Why Reviewing at the Moment of Forgetting Is So Effective
Why would waiting help? Every instinct says to go over it again while it is still fresh. Three mechanisms explain why the instinct is wrong: the testing effect, desirable difficulty, and reconsolidation.
- Testing Effect: The act of trying to recall information itself strengthens memory. In Roediger and Karpicke's 2006 study, students who practiced recall remembered about 56% of a text after one week, against about 42% for students who reread it.
- Desirable Difficulty: Reviewing when some effort is required strengthens the neural pathways of memory. Reviews that are too easy add little.
- Memory Reconsolidation: Each time a memory is recalled, it is reconstructed and converted into more stable long-term memory.
Which is why the review that feels hardest is often doing the most work.
How AI Optimizes Review Timing
Whether spaced repetition works has never been the hard part. Working out when and what to review is. Done by hand across a few hundred cards, it becomes a second job.
Traditional Methods vs. AI Optimization
| Aspect | Traditional Methods | AI Optimization (Memly) |
|---|---|---|
| Review Timing | Fixed intervals (1 day, 3 days, 7 days) | Dynamic calculation based on each card's answer history |
| Difficulty Adjustment | Self-reported | FSRS sets each card's next review from its answer history |
| Study Order | Random or sequential | Prioritized based on forgetting probability |
| Personalization | None | Reflects each card's answer history (parameters shared by all users) |
| Prediction Accuracy | Low | Higher accuracy than fixed intervals with FSRS 6.0 |
The FSRS 6.0 Algorithm Used by Memly
Memly uses the spaced repetition algorithm FSRS 6.0 (Free Spaced Repetition Scheduler). FSRS 6.0 is a predictive model that combines machine learning with cognitive science insights, fitted on large public review datasets.
Thanks to this algorithm, the same study time goes further than under the traditional SM-2 algorithm (Anki's default; Anki has had FSRS built in since 23.10), because fewer reviews are spent on cards you still remember.
How Memly Puts AI Flashcards into Practice
Step 1: Automatic Card Generation from Your Content
Memly's standout feature is its ability to automatically generate flashcards from your learning materials using AI.
- PDF/Text: Simply upload your textbooks or notes, and key points are extracted to create Q&A-format cards
- Images: Charts, diagrams, and handwritten notes are recognized and converted into appropriate questions
- YouTube: The captions of a lecture video are read, and key points are turned into questions
The card-making step drops out of the evening. What is left is the upload, a scan through what came back, and then the reviewing, which was always the only part building memory.
Step 2: AI-Optimized Review Schedule
FSRS 6.0 calculates the forgetting probability for each card in real time and presents reviews at the optimal timing. Cards you learn easily have their intervals extended, while difficult cards are reviewed more intensively.
Step 3: Visualize Your Progress with a Dashboard
A dashboard that shows your learning progress, retention rates, and upcoming reviews at a glance helps you maintain motivation while studying efficiently.
What Memly User Data Shows
The following results come from Memly's internal study (n=648, 2025). It is an internal study, not a peer-reviewed trial:
| Metric | Result |
|---|---|
| Retest Accuracy | +74% (compared to traditional repetition) |
| Card Creation Time | -98% (compared to traditional methods) |
| Study Continuation | +89% (compared to traditional repetition) |
Read each figure with its condition attached. All three compare Memly learners with traditional repetition inside one internal study, so they describe that group and are not a promise of your result.
Use Case Guide by Scenario
Professional Exam Preparation (TOEIC, Bar Exam, Medical Licensing)
For certification exams that require efficiently memorizing large volumes of specialized terminology and regulations, AI flashcards are particularly useful. Uploading textbooks or past exam PDFs makes it easier to focus on the topics that come up most often.
Language Learning
Vocabulary, grammar rules, and example sentences are turned into flashcards and reviewed at optimal intervals based on the forgetting curve. Because every review is an attempt to recall a word, the same study time tends to go further than rereading a vocabulary book.
Programming Education
Programming also involves many elements that require memorization, such as API references, design patterns, and commands. AI automatically generates questions from technical documentation, helping you retain knowledge that is ready for real-world use.
Getting Started in 3 Steps
- Create an account (30 seconds, no payment method required)
- Upload your study materials (supports PDF, text, and images)
- Start learning with AI-generated cards (feel the difference with just 15 minutes a day)
On the free plan, you can try AI card generation with your own material and no payment details required. Try it and see the difference for yourself.
Conclusion
AI flashcard learning is a scientific answer to Ebbinghaus's forgetting curve. The curve itself does not move. What changes is that something is now watching it card by card and picking the day.
Which brings this back to where it started. "No matter how many times I review, I can't remember" is rarely about the number of reviews. It is about their timing. Fifteen minutes tomorrow, on cards the schedule picked, is the cheapest way to find out.
For a full overview of AI-powered memorization support, see our article "What Is AI-Powered Memorization? How It Works, Its Effectiveness & Best Tools [2026 Guide]." For a walkthrough of the available AI flashcard apps, see our complete AI flashcard app guide. For the best AI flashcard apps for students, see our top picks for every budget.
