AI Prompts for Studying: Study Plans, Flashcards and Exam Prep

A study plan works backwards from the deadline, in hours that actually exist in the week. Reviews are scheduled at widening intervals from the start, because material is forgotten on a curve, first measured by Hermann Ebbinghaus, and a plan without scheduled review teaches everything twice.

• Hours that exist: six a week, on the days she named, minus the week she would lose. A plan built on ten hours a week would have failed in February for the same reason.
• 27% of the exam: risk and procurement carry more than a quarter of the marks and were her two weakest topics. The plan gives them four weeks, not a fair share.
• Every week ends in evidence: a quiz, a mock or a piece of work, so a week that produced nothing is visible at the time rather than in July.

Most courses recommend far more than anyone can read, and the skill is knowing which sources carry the marks. Sources are ranked by what they are for: the one you work through, the one you consult, the practice you do, and the extra you read only if time allows.

• Practice over reading: the question bank went in the core list. For this exam, questions answered predict the score better than pages read.
• The skip list has reasons: two books covered an older version of the syllabus and one repeated the official guide.
• Two gaps found: no source covered two syllabus topics properly, which is exactly what you want to know in week one.

Notes are only worth making if they are used again: a wall of transcription is a copy, not a note. Good notes compress a source into its structure and end with questions that force you to retrieve them later.

• Answers held back: the questions are useless if the answer is on the same page. They are kept separately so the note can be used as a test.
• Limits and exceptions: the conditions a method does not cover are where exam questions live, so they get their own section.
• Marked, not smoothed: two passages she could not explain were marked rather than paraphrased, and went straight to prompt 04.

An explanation works when the learner can restate it in their own words, which is the test at the heart of the Feynman technique. The prompt explains at three levels, then asks you to explain it back and corrects what you get wrong rather than repeating itself.

• Where the analogy breaks: a comparison that holds for the first sentence and fails in the second is worse than none, so the failure point is named with it.
• The separating question: the one question whose answer differs between the two ideas is what makes the distinction stick in an exam.
• It changes approach, not volume: asked again, the prompt explains differently rather than repeating the same words more slowly.

A good card tests one thing, asks for recall rather than recognition, and is answerable in seconds. Cloze deletion — a sentence with one term removed — suits definitions and formulas. Cards only work on a schedule of widening intervals.

• 60 scenario cards: the exam asks you to apply the idea to a situation, so 60 cards do that rather than asking for a definition.
• 18 confusion pairs: cards built specifically to force the distinctions she kept losing, including the two reserve types.
• One import: the CSV has Front, Back and Tags columns and goes into Anki in a single step, tagged by topic.

Good questions test understanding rather than recognition: plausible wrong answers, applied to a situation, with no clues in the wording. Bloom’s taxonomy is the ladder — remember, understand, apply, analyse — and most exams sit at apply, where most practice questions do not.

• Every wrong answer gets a cause: not known, misread, confused, or careless. Four causes need four different fixes.
• Strict marking: a nearly-right answer is marked wrong, with exactly what was missing. A generous marker in June is an expensive one in July.
• Harder, not easier: questions she got right easily came back one level up, because that is where the marks she was losing sat.

A mock is only useful under real conditions: the full question count, the real time limit, no notes. The marking matters more than the score, because every wrong answer has a cause and the causes cluster.

• 62% from two topics: the error log turned a syllabus of eleven topics into a plan with two, which is what the next four weeks attacked.
• Nine slips at the end: all in the last 20 questions, which is a pacing problem, not a knowledge one. The fix was timing, not study.
• 58 to 81: four mocks, each one changing the plan. The last came two weeks before the exam, with time to act on it.

An assignment is marked against its criteria, so the criteria are read before a word is written. A strong essay has one arguable thesis and paragraphs that each carry one claim with its evidence.

• The command word: "evaluate" is not "describe". Unpacking the question showed half her planned content was off-topic.
• Nothing invented: the prompt cites only the sources she supplied. Every unsupported claim is marked [NEED SOURCE] rather than given a plausible reference.
• Three gaps, filled by her: the marked gaps were filled from her own hospital's incident data, which is what the rubric rewarded most.

Time is words: spoken delivery runs at about 140 words a minute, so a ten-minute talk is roughly 1,400 words. The slides follow the words, at one slide per 60 to 90 seconds, in five layouts only: title, statement, list, number and closing.

• The budget comes first: her first draft was 2,700 words, which is nineteen minutes. The prompt says so plainly and cuts scope rather than speeding her up.
• Action titles, not labels: every title states the point of the slide - "Two topics caused 62% of my wrong answers", not "Results" - so the deck reads on its own.
• Occasions without slides: a toast, a eulogy or an award speech gets no deck at all, and the prompt says so rather than making one anyway.

Progress comes from input you can almost understand and from speaking before you are ready, what Stephen Krashen called comprehensible input. Vocabulary is chosen by frequency and by your own situation, not by textbook order.

• Her situations, not the syllabus: reception, appointment times and consent forms. The words she needs at 9am, rather than the ones an app teaches in week one.
• Corrected after, not during: a learner corrected mid-sentence stops speaking, so the corrections come at the end, grouped by pattern.
• Three patterns: twenty mistakes traced to three rules. One evening on those three was worth more than twenty separate corrections.