How our questions are written and checked
AI drafts our questions. People decide which ones you see. No question goes live until a subject expert has solved it without seeing our answer and agreed with it.
The four checks
- Draft. A question is drafted to a fixed format: the question, four options or a numeric answer, the solution, the topic and a target difficulty. Anything that does not fit the format is thrown away.
- Machine check. Quant answers are worked out a second time, independently, with code where the answer is a number. Every logical reasoning set is run through a solver that confirms each question has exactly one correct answer. If the checks disagree, the question is rejected before a person sees it.
- Solved cold by an expert. The reviewer sees the question without the answer, solves it, and records the time it took. Only then is our answer shown. If they disagree, the question is fixed or dropped.
- Difficulty from real solving time. We set difficulty from how long experts take and, later, from how students do. Never from the AI's own opinion of how hard a question is.
Why we are this careful
AI models get exam questions wrong, and logic puzzles like CAT's DILR sets are among the hardest things for them. A mock with wrong answers teaches the wrong thing, so every question passes all four checks before you see it. That matters most in DILR, where one wrong clue breaks a whole set. Try the result yourself in the free CAT mock test.
When we get it wrong
Every question has a Report button. Confirmed errors are fixed, scores are recalculated, and the fix is listed on the errata page, where anyone can see it.
Who checks the questions
The reviewers' names and backgrounds will be listed here at launch, and each question will show who checked it.
This site is independent. It is not affiliated with the IIMs or the CAT conducting body.
Last checked 8 Nov 2026. Found an error? Tell us and it goes on the errata page.