In 2006, Henry Roediger and Jeffrey Karpicke published a study that should have changed how every student prepares for exams. They had two groups study the same passages.
One group studied the material twice. The other group studied it once, then took a practice test on it. Five minutes later, the re-readers performed better. One week later, the self-testers retained 80% while the re-readers had dropped to 36%.
The finding wasn’t new. Psychologists had been documenting the “testing effect” since 1909. But this study laid it out so cleanly that it became one of the most cited papers in educational psychology. Students who test themselves learn more than students who re-study, even when they spend the same amount of time.
And yet most people still prepare for exams by reading their notes over and over.
The Psychology Behind Why This Works
When you re-read something, your brain does very little work. You recognize the words, you follow the logic, and you get a comfortable feeling of familiarity. Psychologists call this the “fluency illusion”: the material feels easy to process, so you assume you know it. You don’t, not in the way an exam will require.
When you’re testing yourself, your brain has to actively reconstruct the information. That reconstruction process is difficult and often uncomfortable. You sit there thinking, “I know this, I just studied it, why can’t I remember?” But that struggle is exactly what strengthens the memory trace. Robert Bjork calls these “desirable difficulties.” The harder your brain works to retrieve something, the more durably it’s stored.
There’s a second benefit that gets less attention. Testing yourself gives you accurate feedback about what you actually know versus what you think you know. Without that feedback, you’ll spend hours re-reading material you’ve already mastered while neglecting the topics you’re weak on. Practice questions show you the gaps in real time.
The Practical Problem With Retrieval Practice
If self-testing is so effective, why doesn’t everyone do it? Because making good practice questions is tedious.
I ran into this during my NCLEX prep. I had a mountain of lecture notes, clinical rotation observations, and textbook highlights. I knew from the research that I should be testing myself on this material instead of re-reading it. But writing practice questions from my own notes was painfully slow. I’d spend 30 minutes creating 10 questions that I could answer in 5 minutes, and the questions I wrote were predictably biased toward what I already understood.
Some students form study groups where they quiz each other. That helps, but it depends on finding people who study the same material at the same pace, and their questions have the same bias problems yours do.
This is where AI quiz generators have become genuinely useful, not as a gimmick but as a practical tool for implementing what the research has been recommending for over a century.
Using an AI Quiz Generator for Retrieval Practice
The concept is straightforward. You take your study notes, upload them to a tool like Quizgecko’s AI quiz generator, and it produces practice questions based on that content. The questions target the concepts in your notes, but they’re phrased differently from how you wrote them, which is important because it forces recall rather than pattern matching.
I started using this approach about two months before my NCLEX. After each study session, I’d run my notes through the generator and immediately test myself. The first time I did this, I got roughly 60% right on material I had just finished reading. That was humbling and informative. The fluency illusion had been fooling me for weeks.
Over time I noticed something interesting. The topics where I performed worst on the generated questions were almost always the topics I’d later struggle with on full-length practice exams. The AI questions were catching my weak spots faster than I could identify them on my own.
What The Research Says About AI-Generated Practice Tests
There’s a growing body of evidence on this, though it’s still early. A 2024 study in the British Journal of Educational Technology found that students who used AI-generated practice questions scored significantly higher on final exams compared to students who only used instructor-created materials. The researchers noted that the key factor wasn’t the quality of individual questions but the volume and immediacy, students were simply testing themselves more often because the barrier to creating questions had been removed.
This aligns with what we know about spacing and frequency. Testing yourself once before an exam helps. Testing yourself repeatedly, across multiple sessions, with slightly different questions each time, helps a lot more. AI makes the “slightly different questions each time” part easy in a way that wasn’t possible before.
Being Honest About The Limitations
AI-generated questions are not as good as questions written by subject matter experts. For nursing and medical exams, the best questions test clinical reasoning, the ability to synthesize information from multiple domains and make decisions under uncertainty. AI tools are getting better at this, but they still produce questions that are sometimes too straightforward or that miss the clinical nuance.
I used Quizgecko alongside other study tools, not instead of them. Khan Academy for reviewing concepts I was shaky on. Handwritten flashcards for drug names I kept mixing up. A study group for talking through clinical scenarios. The AI quiz generator was one tool in the system, probably the one that saved me the most time, but not the only one.
Putting It Together
If you’re studying for any high-stakes exam, here’s what cognitive science actually recommends: spend less time re-reading and more time testing yourself. The ratio that worked for me was roughly 40% learning new material and 60% practicing recall through questions.
The barrier to testing yourself used to be that creating good questions was time-consuming. That barrier is largely gone now. Whether you use an AI quiz generator, a study group, or old-fashioned flashcards, the principle is the same. The uncomfortable moment of not being able to remember something you just studied is not a sign that studying isn’t working. It’s the moment when the actual learning happens.
Roediger and Karpicke showed us this twenty years ago. The tools to act on it have finally caught up.