A study plan for exam season: how to spread preparation for several exams across a few weeks

In short: A good study plan for exam season is built backwards. List all your exams with their dates and the amount of material, distribute the material so that each topic comes up several times at spaced intervals, and study several subjects per day rather than one subject all at once. The last few days before each exam should be reserved for practice exams and review, not new material. Also schedule one unallocated buffer day each week.
Why plans fail during exam season
Most study plans fail not because of laziness, but because of three planning mistakes. They underestimate the material because nobody counted how many lectures there actually are beforehand. Students study one subject completely, then the next, forgetting the first in the process. And they have no buffer, so one bad day derails the entire plan.
Learning research shows a better way. A review in Nature Reviews Psychology (Carpenter, Pan, and Butler, 2022) summarises the evidence: Two methods consistently work in laboratories, lecture halls, and vocational training, and they work best together.
Actively retrieve instead of rereading. More than 200 studies show that you retain more when you retrieve knowledge from memory, for example with flashcards, practice questions, or practice exams. Highlighting and rereading feel more productive. The authors call this an illusion of learning.
Space out reviews instead of cramming. People who spread the same material over several days or weeks retain it longer. You should return to it when you have already forgotten some of it. A large study with more than 1,350 participants shows how long the gap can be (Cepeda and colleagues, 2008): If the test was one week away, a gap of one to three days worked best. The further away the exam, the longer the gap can be.
There is also a third finding: Mix subjects and problem types. People who practise different types of problems in a mixed order rather than in blocks perform significantly better on later tests (Rohrer and Taylor, 2007). Mixing feels more demanding because you have to recognise anew which method fits each time. That is exactly what the exam requires.
Step by step to your study plan
1. Take stock: What is coming up, and when?
Put all your exams in a table: subject, date, time, number of lectures, weighting (for example, ECTS credits), and how confident you currently feel. Actually count the lectures. “Biochemistry, about 12 sessions or so” is not enough. This list is the foundation for everything that follows.
2. Plan backwards
Enter the exam dates in a calendar and work backwards. The last two to three days before an exam belong exclusively to that subject: practice exams, past exam papers, and clearing up unanswered questions. Before that, you need to have worked through all the material at least once.
3. Break the material into study units
Divide each lecture into study units of 25 to 50 minutes. A study unit is a topic you can work through or review in one sitting. This shows you how many units you need before the exam and lets you distribute them across the days.
4. Schedule each topic several times
Studying a topic once is not enough. Schedule at least three encounters with each topic: work through it once, review it for the first time after a few days, and revisit it shortly before the exam. Here, review means retrieval, for example with flashcards or practice questions, not rereading.
5. Mix subjects each day
Study two to three subjects in a day rather than one. A typical day: new material in the subject with the next exam in the morning, review in the second subject around midday, and flashcards from all subjects in the late afternoon. This keeps every subject in mind and lets you benefit from interleaving.
6. Build in buffer time
Schedule one day per week without assigned material. Anything left unfinished goes there. If nothing is left unfinished, it is a day off. A plan without a buffer becomes useless after your first illness.
7. Make a brief adjustment every evening
Check off what you have completed and move the rest to the buffer day. Five minutes is enough. What matters is not that the plan works out perfectly, but that you always know what comes next.
Example: three exams in four weeks
Suppose you have biochemistry at the end of week 2, physiology at the end of week 3, and statistics at the end of week 4.
| Week | Main focus | Alongside it | Final days |
|---|---|---|---|
| 1 | Biochemistry: work through all the material | Physiology: work through the first half; statistics: create flashcards | Sunday buffer |
| 2 | Biochemistry: review, past exam papers | Physiology: short reviews | Thursday and Friday: biochemistry only; exam on Friday |
| 3 | Physiology: second half, then past exam papers | Statistics: mixed practice problems; biochemistry on hold | Thursday and Friday: physiology only; exam on Friday |
| 4 | Statistics: mixed problems, practice exam | Brief review of statistics flashcards | Thursday and Friday: statistics only; exam on Friday |
Week 1 is crucial: You start all three subjects then, even though the last exam is still four weeks away. If you do not touch statistics until week 4, there is no room left for spaced repetition.
The most common mistakes
- Starting too late. Spaced repetition needs weeks, not days.
- One subject after another. Much of what you learn in week 1 and do not revisit until week 4 will be gone.
- Only reading and highlighting. This feels productive but does little. Retrieval does a lot.
- No breaks. Schedule a short break after every study unit and a longer break several times a day. Sleep is part of the plan, not the buffer.
- Never adjusting the plan. A plan is a tool, not a contract.
- Producing material instead of practising. Creating summaries, mind maps, and AI texts feels like progress. What counts is the exercises you have done yourself.
AI in your study plan: a shortcut or a helping hand?
AI can help you study or keep you from studying. What matters is who does the thinking.
The largest study on this so far comes from this year. Researchers analysed the grades of 26,811 school students over 30 months (Strömberg, Lei, and Wu, 2026). Those who used AI for homework completed it faster and performed 18 percent better on it. In classroom tests without AI, results fell by around 20 percent after six months, and by 18 to 24 percent in final exams. Four out of five users let the AI solve the problems. The one-fifth who used AI like a tutor and continued thinking for themselves showed no loss.
When AI solves the problems, you learn less. In a study with almost 1,000 school students (Bastani and colleagues, PNAS 2025), groups practised mathematics using a ChatGPT-like chat. During practice, they performed 48 percent better than the group without AI. In the subsequent exam without AI, they performed 17 percent worse. A second group used an AI that provided only hints rather than complete solutions. This group did not experience that harm.
When AI guides you through the problems, you learn more. At Harvard University, 194 students learned a physics topic either in an active in-person class or with an AI tutor that guided them through the problems step by step (Kestin and colleagues, Scientific Reports 2025). Learning gains with the AI tutor were more than twice as large, with less time spent.
An experiment with university students from July 2026 (Contractor and Reyes, not yet peer-reviewed) confirms this: Those who used AI as a tutor still retained more a week later. Those who let AI write the text lost that advantage again.
An initial study from the MIT Media Lab (2025, not yet peer-reviewed) points in the same direction: People who wrote essays with ChatGPT showed lower brain activity and could barely remember their own text shortly afterwards.
For your study plan, this means: Let AI create exercises for you, not answers. Schedule time for exercises you solve yourself, and at the end of the day, count how many questions you answered, not how much material was produced.
How Learngine does it
Learngine is built for exactly this division of labour. The AI creates exercises from your slides and notes: flashcards, quizzes, and practice exams. You have to solve them yourself. Learngine shows you how many exercises you have actually completed and how confident you are in each topic, not how much material you have collected.
The plan is based on your exams. You enter the exam dates, and Learngine distributes exercises and reviews so that every subject is covered in time and the next exam gets the largest share. Every day, Lumi tells you what is up next in 10, 25, or 40 minutes, mixing subjects along the way. If you miss a day, Lumi replans without you having to move anything around.
Frequently asked questions
When should I start studying for exams?
Early enough for every topic to come up at least three times. For several exams, that usually means four to six weeks before the first exam. Ideally, you should already be reviewing the previous week's lectures every week during the semester.
How many hours a day should I study?
That depends on the material and the person. Plan realistically rather than ambitiously: A plan with fewer hours that you stick to beats a plan with many hours that you abandon after three days. Count only focused study time, not time spent at your desk.
Several subjects a day, or one after another?
Several. The subject with the next exam gets the largest share, while the others continue with short reviews. Only in the last two to three days before an exam should you concentrate on that one subject.
What should I do if I fall behind?
Do not push everything back. Prioritise according to weighting and likelihood of appearing in the exam: Start with what appears frequently in past exam papers. Cut details, not entire topics. And use the buffer day.
Sources
- Contractor, Z. & Reyes, G. (July 2026, preprint, revised October 2026): Experimental Evidence on the Learning Impact of Generative AI. arXiv
- Strömberg, D., Lei, V. & Wu, Y. (2026): The generative AI learning penalty: Evidence from Chinese secondary education. CEPR Discussion Paper 21577. Summary at VoxEU
- Bastani, H. et al. (2025, PNAS Vol. 122, Issue 26): Generative AI without guardrails can harm learning: Evidence from high school mathematics. PNAS, 122, e2422633122. Summary from Knowledge at Wharton
- Kosmyna, N. et al. (June 10, 2025, preprint, revised December 2025): Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. MIT Media Lab. Project page
- Kestin, G., Miller, K. et al. (June 2025): AI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting. Scientific Reports, 15. Article
- Carpenter, S. K., Pan, S. C. & Butler, A. C. (2022): The science of effective learning with spacing and retrieval practice. Nature Reviews Psychology, 1, 496–511. Article
- Cepeda, N. J., Vul, E., Rohrer, D., Wixted, J. T. & Pashler, H. (2008): Spacing Effects in Learning: A Temporal Ridgeline of Optimal Retention. Psychological Science, 19(11), 1095–1102. PubMed
- Rohrer, D. & Taylor, K. (2007): The Shuffling of Mathematics Problems Improves Learning. Instructional Science, 35(6), 481–498. Summary
Learngine. As of November 6, 2026.
More from Knowledge
All articles on study methods and guides for university and school