Re-reading feels productive. It is not.
In study-technique research, re-reading and highlighting consistently rank near the bottom for retention, while practice testing ranks at the very top. The reason is simple: recognizing material on a page is a different mental act from retrieving it cold — and exams only test retrieval.
The obstacle has always been effort. Writing good practice questions for yourself takes longer than most students are willing to spend. That is the part AI removes: modern AI tools can generate exam-style questions directly from your own lecture notes and slides, so the material matches exactly what your course covers.
Size the sets correctly
The most common mistake is generating one giant quiz for an entire course. Big sets get postponed, and postponed sets never get taken. Generate 10–15 questions per lecture instead — a set you can finish in under ten minutes, which means you will actually do it.
- Multiple-choice questions work best for facts, definitions, and formulas
- Open-answer questions force deeper retrieval on concepts and processes
- Include a couple of "explain why" questions per set — they expose shallow understanding fast
Schedule with spaced repetition
One pass through a quiz proves little. The schedule that works: take the full set once, then retake only the questions you missed two days later, then again seven days later. Each successful recall at a longer interval roughly doubles how long the memory lasts.
This is also why quizzing per-lecture beats cramming per-exam: by the time finals arrive, most of the material has already survived two spaced reviews, and your final prep shrinks to the handful of questions that kept failing.