The failure mode in IB revision is notes that are complete but useless: everything summarised at the same depth, no command terms, no HL-only detail. The fix is to generate notes per source (one topic, one document), keep the command term attached to each point, and use flashcards only for things that are actually recall items.
Why IB notes need a different shape
Three things make IB different from most exam prep:
- Command terms change the answer. Define, describe, explain, discuss, evaluate are graded differently. A note that says "globalisation is important" is useless for a 22-mark discussion — but it is exactly what a generic summary produces.
- HL and SL diverge. The same topic has extra depth, extra case studies and different weightings at HL. If your notes merge the two, you waste revision time and miss HL-only detail.
- The assessment is spread out. IAs, TOK essays, the extended essay and papers are four different outputs from the same body of knowledge. Notes that only work for paper revision leave you re-doing work for the IA.
The workflow that actually works
1. One source at a time
Upload one chapter, one slide deck or one past paper — not your whole subject folder. Notes generated from a 200-page PDF are averages: they keep whatever appears most often and drop the specific example your examiner wants. A single 15-page chapter keeps the detail that matters.
Text, a URL, or a PDF all work: text goes in directly, a URL is fetched and cleaned, and an uploaded PDF is parsed to text first. If the PDF is a scan, the text layer may be thin — check what came through before generating.
2. Generate the summary first, then interrogate it
Start with the summary (it is the cheapest step) and read it as an outline, not as your notes. The question to ask is: which of these headings would I fail to write half a page on? Those are the ones you go back into the source for.
3. Flashcards only for recall items
Definitions, formulas, dates, named studies, command-term meanings — these are recall. Arguments, structures and evaluations are not: they need practice, not flash cards. Making a card that says "discuss the causes of X" just teaches you that the topic exists.
4. Practice questions in the exam's language
This is the step most people skip. Generate practice questions from the same source, then rewrite each one to start with an actual command term from your subject guide, and time yourself. Multiple-choice questions are good for checking that you know the content; they do not prepare you for a 22-mark essay. Use both.
5. A mind map per topic, at the end
Build the map after you have studied the topic, not before. Its job is to show the shape — how the sub-topics hang together, where the HL-only branch sits — so that you can sketch it from memory in the exam room.
Where AI notes quietly lose marks
- Syllabus invention. A model can produce a confident point that is not on your syllabus, or that belongs to a different subject guide edition. Anything that will be graded should be checked against the guide itself.
- Flattened command terms. Summaries turn "evaluate" into "describe". If a note does not carry the verb, it is not an answer.
- Missing HL depth. Unless you tell it, the model does not know which level you are studying.
- Generic examples. Real marks come from the case study or experiment you actually studied. Add your own, every time.
Use it as a first pass, never as the finished notes
Everything here is faster than writing notes from scratch, but nothing replaces checking the points against your own textbook and subject guide. Notes you did not write are notes you cannot defend in a viva, and the person marking your IA will read them closely.
中文摘要
IB 溫書最易犯嘅錯，係筆記「齊但冇用」：所有內容同一深度、冇 command term、分唔清 HL 同 SL。做法係一個來源一份筆記（一章、一份 slide、一份 past paper），先出 summary 當大綱睇，再針對自己寫唔出半頁嘅位返去原文；flashcard 只留定義、公式、研究名呢類真係要背嘅嘢；練習題要用你科目 guide 真正嘅 command term，並且要計時。Mind map 留到最後，用來記住整個 topic 嘅形狀。AI 出嘅內容一定要對返自己嘅 textbook 同 subject guide — 特別係會計分嘅 IA 同 TOK。