Using AI for CFA study: check the rules and answers

CFA Institute’s public website terms are marked updated 15 October 2026. They restrict using its site material with AI systems, including through third parties. Access to a reading is not, by itself, permission to upload it into a chatbot. Read the applicable material licence and ask the rights holder when the permission is unclear.

Do not assume that making an upload private, calling it a summary request or excluding one subject changes the permission. Use your own original examples for the checks below. This page reports the published terms; it does not decide an individual candidate’s contractual position or disciplinary outcome.

The official Level I page currently describes 180 multiple-choice questions, three choices per item and two 135-minute sessions. That is 270 minutes for 180 questions: 270 ÷ 180 = 1.5 minutes, or 90 seconds per question on average. A short, untimed quiz is a different task, even if its answers are correct.

CFA Institute lists curriculum readings, practice questions and mocks among its candidate resources. Use the version for your registered sitting. Keep a coverage list and a record of unfamiliar questions; a generated lesson sequence can omit a tested distinction without making that omission visible in its quiz score.

There is no evidence here for prescribing “50% AI study” or another universal split. Choose the next study action from the gap you can demonstrate: missing knowledge needs instruction, an incorrect setup needs reasoning repair, and a timing problem needs appropriate timed practice.

Fictional example: a generated set contains 20 questions. You choose all 20 answers marked correct by its key, so the displayed score is 20 ÷ 20 × 100 = 100%. An independent review of all stems and explanations finds four wrong keys. Assume each corrected question has one unambiguous answer and your original choice is wrong on those four.

Correct answers under the checked key = 20 - 4 = 16. Regraded score = 16 ÷ 20 × 100 = 80%. The difference is 20 percentage points on this same fictional set. Nothing about your knowledge changed during the regrading; the measurement changed.

These invented counts are not an observed AI error rate or a CFA pass threshold. If a stem is ambiguous or lacks necessary facts, fix or exclude it before grading and disclose the changed denominator. Asking the same system whether its own key is right is not an independent source check.

Original editorial example: a single USD 121 payment arrives exactly two years from today. Assume a 10% effective annual discount rate, annual compounding, no intermediate cash flows and no taxes or fees. Present value = future payment ÷ (1 + annual rate)^years = 121 ÷ 1.10² = 121 ÷ 1.21 = USD 100.

Suppose an explanation instead gives 121 ÷ 1.10 = USD 110. It has discounted one year, not two. Reverse the correct result: 100 × 1.10 × 1.10 = 121. The time line and reverse calculation identify the missing period without relying on confident wording.

The public 2026 time-value-of-money refresher identifies present-value analysis as a learning task. The amounts and deliberate error here are our own examples, not copied curriculum questions or an observed output from a named AI product. For your sitting, check the applicable current learning outcomes.

When a generated quiz feels easier than official practice, compare the tasks before drawing a conclusion. Record whether both were closed-book, unfamiliar and timed, and whether they covered the same topics. Different conditions can produce different scores without identifying a single cause.

Use a small verification log: topic, original answer, claimed rule, supporting source, independent working and a changed-fact check. If you cannot explain a step, return to the relevant instruction. If the generated question lacks a valid answer, repair the question rather than memorizing its key.

Finally, try fresh authorized practice without opening the explanation first. Explain both the result and why another answer would be wrong. Keep any uncertain rule on your review list. This is a study decision aid, not a validated diagnostic or a promise that a particular routine produces a pass.