The task
Final exam is in 3 weeks. The syllabus has 12 topics; you’ve covered 6 of them confidently, 2 you understand on a good day, and 4 you have barely started because the homework dragged. You can study 2 hours on weekdays and roughly 4 on Saturday and Sunday — that’s about 22 hours a week minus the time when life happens. You want a week-by-week revision plan that respects two things: the topics you actually don’t know yet need more practice than reading, and the topics you sort-of know need spaced retrieval, not a single end-of-week review session. Plus at least 2 full mock tests before exam day.
Where AI helps — and where it does not
AI is good at building the week-by-week schedule, balancing new-material vs. revision slots based on forgetting-curve principles, weighting practice over re-reading for low-confidence topics, and spacing mock tests so they actually give you signal. It can also build in catch-up afternoons so a missed Tuesday doesn’t cascade.
What AI cannot do: diagnose what you actually don’t understand. “Confidence 2” on a topic could mean “I forgot one formula” or “I never understood the foundation.” Feed it real signal — a recent practice test score or your honest self-assessment per topic. AI is also bad at predicting your real attention span; if you say “2 hours weekday” but your actual focused time is 80 minutes with 40 of phone distraction, the plan will look right and execute badly.
A specific failure mode: AI defaults to a daily mixed-topic schedule that switches every 30 minutes. For exam revision, longer blocks on the same topic (60-90 min) plus daily 15-min retrieval on prior topics beats constant context-switching. Tell it the block structure you want.
What to feed the AI
- Exam date, subject, and exam format (multiple choice / problem set / essay / oral)
- Full topic list with self-rated confidence 1-5 per topic (1 = barely started, 5 = could teach it)
- Recent practice-test or mock-exam scores per topic, if you have them — beats self-rating
- Hours available per day, honestly — weekday and weekend separately
- Days you cannot study at all (other deadlines, social commitments, recovery)
- Your strongest retrieval methods (problem sets, flashcards, redoing past papers, teaching a peer)
- Methods that haven’t worked for you in past exams (often re-watching videos, highlighting)
- One specific topic you’re avoiding (every revision plan has one) — name it so the model schedules it early, not last
Copy-ready prompt
Build my revision plan for an exam on {date}.
Subject + exam format: {paste}
Topic list with confidence 1-5 (or practice-test scores): {paste}
Days I cannot study: {paste}
Hours per day: weekday {N}, weekend {N}
My most effective retrieval methods: {paste}
Methods that haven't worked: {paste}
The one topic I'm avoiding: {paste}
Structure:
1) Week-by-week table for the {weeks} weeks until exam.
2) Each study day: topic, mode (read / practice / mock test / retrieval review), minutes per slot. Use 60-90 min focus blocks, not 30-min switches.
3) Daily 15-min retrieval review on a prior topic at the start of each session — spaced repetition, not end-of-week-only.
4) Mock-test spacing — at least 2 full mocks, one at the end of week (weeks total) - 2, one at the end of (weeks total) - 1. Full timed conditions, plus 60 min review of wrong answers.
5) One built-in "catch-up afternoon" per week (suggest Sunday afternoon) — if I hit the week, that's rest; if I missed days, that's recovery.
6) Rest days — schedule 1 fully off per week, not the day before exam.
Rules:
- For any topic at confidence ≤ 2, practice slots must be at least 2x the read slots. Reading without practice does not lift confidence on weak topics.
- The avoided topic must appear in week 1, not week 3.
- Leave 20% slack per day for slippage. Do not pack the plan to the minute.
- Past exam papers count as both practice and mock; tell me which I should treat as which.Shorter variant — one-week sprint
1 week to exam. Topics still weak: {paste with confidence}. Hours: weekday {N}, weekend {N}. Give me a 5-day plan that hits each weak topic at least 3 times via practice (not re-read), plus a final mock 48 hours before exam with 24 hours of wrong-answer review after. Skip rest days this week.Sample output
A useful week-1 row: “Mon: 60 min Ch 3 practice problems (confidence 2, the avoided topic — start here) + 15 min retrieval review on Ch 1 (confidence 4). Tue: 90 min Ch 4 practice + 15 min retrieval review on Ch 3. Wed: rest evening. Thu: 60 min Ch 5 read + 30 min Ch 5 practice + 15 min retrieval on Ch 4. Fri: 60 min Ch 4 deep practice (still confidence 2 after this week) + 15 min retrieval on Ch 5. Sat: 120 min full mock for Ch 1-3 + 60 min wrong-answer review. Sun: catch-up afternoon (2-3 hours if needed) or rest.”
A useful “avoided topic” note: “Ch 3 (Markov chains) is your stated avoidance. Scheduled Monday week 1, not end of week 3. Confidence 2 today — you need at least 3 practice rounds before exam; I’ve placed them on Mon week 1, Thu week 2, Sat week 3.”
A useful mock-test setup: “Mock test 1: end of week 1, covers Ch 1-3. Time it strictly (120 min). Spend 60 min the same day reviewing wrong answers — not the next day; consolidation matters. Mock test 2: end of week 2, covers Ch 1-7. Same protocol. Save the week-3 mock optional, as confidence calibration.”
How to refine
- Force practice-over-read on weak topics: “For every topic at confidence ≤ 2, the schedule must show practice minutes ≥ 2x read minutes. If a weak topic has 60 min read and 30 min practice, swap the weighting.”
- Place the avoided topic in week 1: “Re-check: the topic I’m avoiding must appear in the first 3 days, not the last week. If you placed it later, move it forward and explain what to bump.”
- Use 60-90 min blocks: “Re-read the daily schedule. If any day has 4 topic switches across the 2 hours, consolidate into 1-2 blocks. Switching every 30 min is read-and-forget; sustained blocks build.”
- Add daily 15-min retrieval: “Each study session should start with 15 min of retrieval on a topic from 1-3 days ago. Add this if missing. Spaced retrieval is what makes weak topics stick.”
- Build in 20% slack: “Re-check: every day should have ~20% unscheduled buffer. If a day is packed minute-to-minute, cut the lowest-priority slot. Plans without slack don’t survive contact with real life.”
Common mistakes
- Treating “read again” as revision — re-reading is the lowest-yield encoding method; for exam recall, problem-solving and retrieval beat reading 3-to-1
- No mock tests until the last week — mocks done late tell you what you can’t change; the value is in week 1-2 when you still have time to fix gaps
- Cramming the weak topic on day 1 and never returning — once is not learning; spaced repetition across at least 3 sessions is what moves confidence
- Re-distributing missed days across the whole remaining plan — overpacks every day, you miss them all, the plan dies; use the catch-up afternoon, then resume normal
- Packing the plan to the minute — life happens; 20% slack per day is the difference between a plan you execute and a plan that breaks on Wednesday
- All-mixed-topic days — context-switching every 30 minutes hurts encoding; longer focus blocks plus short retrieval reviews beats it
- No rest day — fatigue compounds and you lose the last week to a fog you cannot push through; one full rest day per week is part of the plan, not a luxury
- Confidence based on hope, not data — “I’m a 4 on this” without a practice score is wishful; whenever possible, use actual scores
FAQ
- What if I miss a planned day?: Use the built-in catch-up afternoon (usually Sunday). Do not redistribute the missed work across all remaining days — that overloads every day and you miss all of them. If you miss two days, one catch-up afternoon handles ~half; accept that the lowest-priority topic gets less coverage and continue.
- How do I handle topics I haven’t started?: Give them confidence 1 and the model will schedule them as high-priority new-learning with both read and practice slots, not just revision. Don’t wait until week 2 to start them; the foundation needs time to settle.
- Should I do mocks under timed conditions?: Yes. Untimed mocks are a different cognitive task — you can’t learn pacing or pressure management without the timer. Time it strictly, same conditions as the real exam, no looking at notes mid-mock.
- My exam is open-book — does this change the plan?: Yes. Reading slots become “build your reference document” slots — the exam tests how well you can navigate the material under time pressure, not what you memorized. Mocks become “practice flipping to the right section in 90 seconds.”
- The model keeps under-scheduling mocks — what changes?: Add: “Mock tests are non-negotiable. Schedule at least 2 full timed mocks, the first at the end of week (total-2) and the second at end of week (total-1). Each mock includes 60 min same-day wrong-answer review. Do not let other study displace mocks.”