How the CPA Exam Is Scored: The Scaled 75, Item Weighting, and the End of Adaptive Testlets

A 75 on the CPA Exam is not 75% of the questions right. The AICPA says so in plain words: the reported score "does not represent a percentage correct, nor can you interpret it as such." It is a point on a 0 to 99 scale that the AICPA builds from how many items you answered correctly and how hard each of those items was, and it is not curved against the people who sat beside you.

The second thing most candidates believe about scoring is also out of date. Until the 2024 exam, the multiple-choice testlets were adaptive: a strong first testlet earned a harder second one, and a harder second testlet was read, by a whole industry of forum posts, as a sign you were passing. The 2024 redesign removed multistage adaptive testing from the multiple-choice testlets. The second testlet is now no signal at all.

What follows is what the AICPA actually publishes about scoring, what it does not publish, and the handful of decisions on exam day that follow from a linear, difficulty-weighted exam.

What the AICPA publishes

Every section is built the same way: five testlets, the first two multiple-choice, the last three task-based simulations. The two halves are combined into one scaled score with a fixed weighting.

Section	Multiple choice	Simulations	Weighting
FAR	50	7	50% MCQ, 50% TBS
AUD	78	7	50% MCQ, 50% TBS
REG	72	8	50% MCQ, 50% TBS
BAR	50	7	50% MCQ, 50% TBS
ISC	82	6	60% MCQ, 40% TBS
TCP	68	7	50% MCQ, 50% TBS

Three sentences on the AICPA's scoring page carry the whole model. Scores "are reported on a scale that ranges from 0 to 99." To pass "you must score a minimum of 75." And the scaled scores on both halves "are calculated using formulas that take into account factors such as whether the question was answered correctly and the relative difficulty of each question."

That last sentence is the one to read twice. Difficulty is part of the formula. Two candidates with the same number of correct answers do not get the same score if one of them got the harder items right. This is the trace of item response theory, the statistical model behind most large licensing exams, and it is why the exam can still be fair across forms of different difficulty without any adaptive routing.

The AICPA also confirms that every section contains pretest items, "multiple-choice questions or task-based simulations," included "only for the purpose of collecting data." They do not count and they are not marked. You will not know which questions are pretest, so every item has to be treated as scored.

Why 75 is not a percentage

A scaled score is a translation. The raw result of your session, items right weighted by difficulty, is mapped onto the 0 to 99 scale so that a 75 means the same level of performance on every form of the exam, in every window, regardless of which items you happened to draw. The passing standard was set once, by a panel judging what a newly licensed CPA needs to know, and the scale is anchored to that standard rather than to a percentage or to the other candidates in your quarter.

Two consequences are worth holding onto. The first is that the exam is not graded on a curve. The AICPA says the scores "are not curved," and the pass rates that move from quarter to quarter, from 42.95% on FAR to 79.53% on TCP in the first half of 2026, are the outcome of who sat, not of a quota. The second is that there is no fixed raw number you can aim for. Sixty percent of the items right may pass on a hard form and fail on an easy one, because the hard form's items are worth more. The only reliable target is to answer as many items correctly as you can, with the hard ones counting most.

The end of adaptive testlets

Before 2024 the multiple-choice half of each section was a multistage adaptive test. Everybody started on a medium testlet; a good performance routed you to a difficult second testlet, a weak one to another medium testlet. Difficult items were worth more, so the harder route was the higher-scoring one, and a generation of candidates learned to guess their result from how the second testlet felt.

The 2024 exposure draft that redesigned the exam listed, among its changes, "eliminating multistage adaptive testing (where the difficulty of the Exam is dynamically personalized) in the multiple-choice question testlets." The blueprints that launched with the Core and Discipline model in January 2024 carried the change through. Since then every candidate on a given form sees testlets assembled in advance, not chosen in response to their answers.

Three things follow, and they are the practical content of this article.

The second testlet tells you nothing. If it feels harder, that is the mix of items on that form, not a verdict on your first testlet. Candidates who spend the second testlet monitoring its difficulty instead of answering it are paying attention to a signal that no longer exists.

Difficulty still weights the score. Removing adaptive routing did not remove item response theory. A hard item you get right is still worth more than an easy one. What changed is that the hard items are distributed across the testlets by design rather than served to you as a reward.

There is no advantage in the order you answer. On an adaptive exam a slow, careful first testlet had a theoretical payoff: it earned the harder, higher-value route. On a linear exam the first and second testlets are the same kind of thing. Pace them identically.

How simulations are scored

The AICPA applies the same formula language to simulations: scored on correctness and difficulty, then scaled and combined with the multiple-choice score at the section's weighting. It does not publish the grading scheme of an individual simulation, and it does not say how a partly completed one is credited, so the safe assumption is that every response field is a scored opportunity and none should be left empty.

Two of the older formats are gone. The written communication task, the essay that used to sit in BEC, was removed with the 2024 exam, and the research task that asked candidates to cite a paragraph of the authoritative literature is, in the AICPA's words, assessed "differently." What remains is the simulation family described in Task-Based Simulations on the CPA Exam: document review, journal entries, reconciliations and calculation grids, with a spreadsheet tool the AICPA describes as "similar but not identical to Excel."

The weighting makes the arithmetic blunt. In AUD, FAR, REG, BAR and TCP, the seven or eight simulations are worth exactly as much as the fifty to seventy-eight multiple-choice questions. A candidate who is strong on multiple choice and weak on simulations is fighting for half the marks with a fraction of the preparation. ISC uses 60% MCQ and 40% TBS.

What this means on exam day

The scoring model rewards a small number of unglamorous habits.

Habit	Why the scoring model rewards it
Answer every item	Nothing in the AICPA's published scoring describes a penalty for a wrong answer, and an unanswered item cannot score. A guess has a floor of zero and a ceiling above it.
Spend time in proportion to difficulty, not to anxiety	Harder items are worth more, so an extra minute on a hard item you can solve is worth more than an extra minute confirming an easy one.
Do not diagnose the second testlet	Its difficulty is a property of the form, not of you.
Treat every item as scored	Pretest items are unmarked. There is no way to spot and skip them.
Reserve half the session for simulations	They are half the score on every section except ISC, and they are where partly completed work is at risk.

The pacing that follows from this, testlet by testlet and section by section, is set out in CPA Exam Time Management by Section.

Reading a fail

A candidate who scores below 75 receives a Candidate Performance Report alongside the score. It describes performance by content area and by item type relative to candidates who just passed, and it is the only diagnostic the AICPA provides. A 74 and a 62 call for different retake plans, and the report is how you tell them apart; Failed a CPA Section? walks through it.

What the report cannot do is tell you which items you missed. Scores are released "on a rolling basis," with target dates the AICPA posts twice a year, and no candidate ever sees the scored items again. The practice questions you review before the exam are the only place the feedback loop closes.

Where to practise this

A difficulty-weighted, linear exam rewards two things in preparation: volume, so that the hard items are familiar rather than novel, and explanations, so that the hard items you got wrong stop being hard. Prepito's FAR, AUD, REG, BAR, ISC and TCP courses give every question a full explanation and track the ones you got right for the wrong reason, which on this scoring model are the ones that matter.

For the whole qualification from the Gulf, start with the CPA programme or the guide to taking the CPA Exam from Saudi Arabia and the Gulf.

---

Sources: AICPA, CPA Exam scoring and pass rates, AICPA, CPA Exam frequently asked questions, AICPA, Exposure draft on the 2024 CPA Exam changes, AICPA CPA Exam Blueprints. Item counts and weightings are those of the 2026 Blueprints and the pass rates are cumulative for the first two quarters of 2026; the AICPA revises both, so confirm the current figures for your testing window.