“Spaced repetition” appears on every flashcard app's store page, but the details determine whether the schedule actually fits your memory. This post is the honest, mechanical account of how AiWords does it — what numbers live behind each word, how your answers turn into those numbers, and what the daily queue is made of.
Three numbers per word
AiWords schedules reviews with a variant of SM-2, the algorithm behind the first generation of serious flashcard software. Every word in your study record carries the classic SM-2 fields:
- Easiness factor (EF) — a multiplier between 1.3 and 2.5 that describes how readily this word sticks for you. Words you keep missing drift toward 1.3; easy words sit near the top.
- Interval — the current gap between reviews, in days.
- Repetitions — how many successful encounters the word has had in a row.
Alongside those sit two scheduling fields: a next review timestamp, which decides when the word is due, and a priority used to order words within a session. All of this lives in your study record on the server, so it survives reinstalls and follows your account between devices.
The quality score: how your answers are graded
The engine runs on one input: a quality score from 0 to 5 per encounter. It comes from one of two places.
You rate it yourself
During study you can press a rating button on each word, and the choice maps directly onto the scale:
| Button | Quality | Meaning |
|---|---|---|
| Hard | 0 | Forgot it completely — needs relearning |
| New | 3 | Correct, but it took effort |
| Seen | 4 | Correct and fairly easy |
| Know | 5 | Perfect recall |
Or the quiz grades you
If you just answer the multiple-choice questions, the app derives the score from your accuracy across the word's checks in that session:
| Accuracy | Quality |
|---|---|
| 100% correct | 5 |
| 80% or more | 4 |
| 60% or more | 3 |
| 40% or more | 2 |
| Above 0% | 1 |
| 0% | 0 |
A manual rating always wins over the derived one. And the record also carries your error count (zero to three wrong picks) plus which distractors you confused the word with — the server uses those to adjust the final quality, so a word you keep mistaking for the same neighbour is treated more strictly than one you failed at random.
What happens after a session
Study records upload in batches; when the server answers, it returns the updated state per word: a new next-review date, a new interval in days, a new easiness factor, and a mastered flag. The next time you open the app, the queue is already computed from those dates. Records are stored locally first and retried automatically if the network is down, and they also upload when the app moves to the background — so closing the app mid-day does not lose the morning's work.
The practical effect is the one spaced repetition is famous for: words you know well earn longer and longer gaps, while words you keep failing stay near the front with short intervals and extra repetitions. The schedule adapts per word, not per session.
Anatomy of the daily queue
Each day the server assembles two lists: new words, up to your daily allowance, and review words — everything whose next-review date has arrived. The home screen shows both counts and an estimated session time before you start. Inside the session, your study intensity setting controls how many in-session rounds each word gets:
- Normal — new words shown once with 2 reviews; due reviews pass through 1 round.
- Intensive — new words shown once with 3 reviews; due reviews go through 2 rounds.
- Heavy — new words shown once with 3 reviews; due reviews go through 3 rounds.
Two behaviours worth knowing. First, new words multiply: the in-app guidance is that each 10 new words generate 3 to 5 times that volume in reviews over the following days, which is why the recommended daily range is 10 to 50. Second, if the due pile exceeds your daily review cap, the overflow is not dropped — it waits for the next session. That keeps one bad day from turning tomorrow into a three-hour debt.
Where this lives
If you want to feel the mechanics instead of reading about them: watch what happens to a word you answer “Hard” three days in a row versus one you answer “Know”. The first keeps reappearing inside the same week; the second quietly disappears for days at a time and returns just as you start to wonder about it. That timing is the easiness factor and the interval doing their job — and it is the whole reason the daily session is a queue of mixed words rather than a chapter of a book.
The user-facing walkthrough of all of this — plans, modes, settings — lives in the AiWords guide. For the app's wider history, see how AiWords got from 1.0 to 3.0.