Standard single-best-answer practice will carry you a long way through the RACGP Fellowship written exams, and no further. Three things it cannot assess sit at the centre of the fellowship: key-feature reasoning under the KFP's own conditions, the applied use of Australian primary-care guidance in context rather than as recalled facts, and the consultation skills examined in the Clinical Competency Exam. This article names each gap precisely and gives you an observable behaviour, a deliberate-practice task, a feedback source and an exit standard for training it. iatroX covers the AKT and KFP knowledge layer; it is explicitly not a CCE simulator, and that boundary is drawn honestly below.
The RACGP Fellowship format map
The fellowship has three assessments, and they test different things.
- Applied Knowledge Test (AKT): 150 single-best-answer questions applying knowledge in the Australian general-practice context. Currently paper-based with computer-readable answer sheets; verify current delivery and duration on racgp.org.au.
- Key Feature Problem (KFP): 70 multiple-selection questions, each case independent of the others, testing decision-making at the critical steps of a case. This is not standard single-best-answer format, and that difference is the whole point below.
- Clinical Competency Exam (CCE): the clinical and consultation assessment — the performance stage. The RACGP does not publish a full internal structure in a form worth paraphrasing here, so treat it as the applied clinical stage and do not invent its mechanics. Confirm the current format on racgp.org.au.
Results for the written exams are reported as a pass or fail with a score band rather than a raw mark. Keep the distinction between the three assessments sharp: a question bank prepares the AKT and KFP knowledge layer; it does not prepare the CCE, and no chatbot or bank is a consultation simulator.
Knowledge versus performance
A correct selected answer proves you could recognise the right option among the ones offered, on that day, in that format. It does not prove you would generate the same answer with no options in front of you, that you would gather the information the vignette handed you for free, that you would communicate the plan to a patient, or that you would do any of it inside a ten-to-fifteen-minute consultation while a patient reacts to what you say. The fellowship deliberately separates knowing from doing, which is why passing the AKT tells you little about the CCE. Everything that follows is about the doing.
Take a worked contrast. A KFP-style case on a febrile returned traveller might reward you for selecting the right investigations from a menu; the CCE version of that same patient requires you to elicit the travel history yourself, notice the detail that lifts malaria up the differential, arrange the tests in a system where access and cost are real, and explain the plan to an anxious patient inside the consultation time. The selected answer proves the first step. The fellowship tests all four, and three of them are only visible in a room.
The three under-tested skills, broken down
Key-feature reasoning
Key-feature problems test decision-making at the two or three steps of a case where getting it wrong changes the outcome, and they use multiple-selection formats without a tidy list of five options to lean on. Ordinary MCQ practice trains recognition and elimination; the KFP punishes both, because there is nothing to eliminate against and partial or over-inclusive answers are penalised.
- Observable behaviour: at each key step, you commit to the smallest correct set of actions — no scattergun lists, no omissions.
- Deliberate-practice task: work KFP-style cases with the options hidden; write your own answers, then compare against the marking approach.
- Feedback source: the official RACGP marking guidance and a peer or supervisor for the reasoning, not a generic explanation.
- Exit standard: you consistently identify the true key steps and answer them precisely, neither over- nor under-calling, across unfamiliar cases.
Australian primary-care guidance
The written exams reward guidance applied in the Australian context — Therapeutic Guidelines, RACGP's own guidelines and preventive-care material, and the practical realities of the PBS, MBS and Australian referral pathways. A candidate trained on overseas banks will carry thresholds and drug availabilities that are simply wrong here.
- Observable behaviour: you apply the current Australian recommendation, including its access and funding context, without defaulting to a UK or US position.
- Deliberate-practice task: for each topic, state the Australian source and date, and note where it diverges from overseas guidance.
- Feedback source: primary Australian guidance and a clinician who works in the system; not an AI answer taken on trust.
- Exit standard: you can name the current Australian position and source for your high-frequency topics from memory, and you catch overseas-flavoured distractors.
Consultation skills (the CCE)
The consultation is a performance: gathering data efficiently, reasoning under time pressure, managing complexity and safety-netting, and relating to the person in front of you. None of this is visible in a selected answer.
- Observable behaviour: a structured, patient-centred consultation that reaches a safe, justifiable plan within the time and communicates it clearly.
- Deliberate-practice task: timed role-play consultations with a peer or supervisor playing the patient, mapped to the exam's assessment areas.
- Feedback source: an examiner, supervisor or trained peer against the official assessment criteria — a human, not automated scoring.
- Exit standard: you reliably meet the assessment criteria across varied, unfamiliar cases within the consultation time, not just on rehearsed scripts.
A four-week modality ladder
Train each skill up a ladder from isolated to integrated, so you never practise only the easy rung.
- Week 1 — isolated skill: drill one component at a time. Hidden-option KFP steps for reasoning; source-and-date drills for Australian guidance; single consultation micro-skills (opening, safety-netting) for the CCE.
- Week 2 — coached case: work whole cases slowly with a supervisor or peer who stops you at each decision point and asks for your reasoning. Feedback is the point; speed is not yet.
- Week 3 — timed integrated case: run full cases and consultations under exam time, still with human feedback afterwards. Now the clock is part of the task.
- Week 4 — unseen simulation: fresh, unseen cases with no coaching during the attempt, scored afterwards. This is the readiness read. For the knowledge layer, a fresh unseen RACGP baseline in iatroX measures AKT and KFP transfer; the consultation rung still needs a human examiner.
A worked example: strong on paper, shaky in the room
Consider a registrar who has passed the AKT comfortably and scores well on KFP-style banks, but whose practice CCE consultations keep stalling. The scores say the knowledge is there; the performance says something else. Working up the ladder exposes the real gap. At the isolated-skill rung she safety-nets inconsistently and runs over time. At the coached-case rung her supervisor notes that she gathers data thoroughly but rarely signposts or checks the patient's understanding. At the timed rung she abandons structure under pressure and reverts to a rote history-taking script. None of that is a knowledge deficit, and none of it would surface in a bank percentage. The fix is targeted consultation practice — timed, human-scored, against the CCE criteria — not another hundred multiple-choice questions. Her AKT and KFP time becomes maintenance; her marginal hour goes to the consultation rung. That reallocation, driven by where the ladder breaks rather than by what feels productive, is the entire reason for separating knowledge from performance before you plan your weeks.
When AI feedback helps, when it does not, and when a human is required
AI feedback is genuinely useful at the isolated-skill rung: generating practice stems, explaining a mechanism, drilling a fact, or giving fast first-pass feedback on a written reasoning step. It is unreliable on anything jurisdiction-specific or recently changed — it will confidently state an overseas threshold as if it were Australian — so every guidance claim needs verification against a primary source, exactly as described in calibrating automated feedback before you trust the score. And it cannot score a consultation: reading a patient, judging rapport and safety-netting, and applying the official CCE criteria require a human examiner. The method for pressure-testing any AI tutor's reliability is in the audit-an-AI-tutor pillar. Use AI for volume and first-pass drilling; use a clinician for anything that carries marks in a performance exam.
A balanced case and task matrix
Candidates drift toward familiar scenarios and arrive over-practised on chest pain and under-practised on the awkward, low-frequency, high-stakes cases the exam favours. Build a matrix and fill every cell before you repeat any.
| Dimension | Vary across |
|---|---|
| Patient group | Child, adolescent, adult, older person, pregnant patient |
| Acuity | Routine, urgent, emergency, uncertainty and undifferentiated presentation |
| Domain | Physical, psychological, sexual and reproductive, preventive, multimorbidity |
| Context | Aboriginal and Torres Strait Islander health, rural and remote, aged care, mental health |
| Task type | Diagnosis, management, communication, ethics and professionalism, safety-netting |
Sequence the matrix deliberately: begin each week by filling the cells you have never practised, not the ones you enjoy. A quick audit — tick every cell you have genuinely rehearsed under time — usually reveals a lopsided grid clustered around common, comfortable presentations, while the empty cells are disproportionately the high-stakes, low-frequency, cross-cultural and ethically complex cases the exam favours precisely because they discriminate between candidates. Rotate so that no dimension is neglected for more than a week, and retire a cell only once you have handled an unfamiliar case in it, under time, with human feedback. If a cell is empty, that is your next case — not another repeat of a scenario you already handle well.
Red flags that your preparation is not working
- Memorised scripts: you can only perform rehearsed consultations and fall apart on an unfamiliar opening.
- Repeated cases: your practice set is small enough that you now recognise cases instead of reasoning through them.
- Generic feedback: the feedback you receive would fit any candidate and names no specific behaviour to change.
- Uncalibrated scoring: your scores swing with the assessor rather than tracking your competence, because no one is marking against the official criteria.
- No official-rubric check: you have never mapped your performance against the RACGP's own assessment areas, so you are optimising against a rubric you invented.
Any two of these together mean the modality is wrong, not that you need more questions.
Reading your readiness across the three assessments
Do not collapse three different readiness signals into one number. For the AKT, readiness is unseen, timed accuracy by curriculum context, at or above the pass standard. For the KFP, it is precision at the key steps — the right, minimal set of actions, without over-calling or omission — on unfamiliar cases. For the CCE, it is the consistent achievement of the official assessment criteria across varied consultations, judged by a human, within the consultation time. A candidate can be ready on one and not the others, and the commonest trap is letting a strong written signal stand in for a clinical one it cannot represent. Track the three separately, and let the weakest set the agenda for your remaining weeks.
Bottom line
MCQ banks build the knowledge that the AKT and KFP reward, and that is necessary but not sufficient. Key-feature reasoning needs option-free practice against the official marking approach; Australian primary-care guidance needs source-and-date discipline, not overseas recall; and consultation skills need timed, human-scored practice against the CCE criteria. Train each up a modality ladder, verify every AI-generated guidance claim, keep a human in the loop for the performance rung, and measure the knowledge layer on unseen items — while remembering plainly that no bank, and no chatbot, is a CCE simulator.
Frequently asked questions
How do I know whether I have covered the full RACGP Fellowship blueprint? Not by finishing a question bank, but by mapping your recent unseen performance against the RACGP curriculum's contextual units and your consultation practice against the CCE's assessment areas, then confirming no cell in your case matrix is empty. Coverage is a map you build deliberately, as described in completion is not coverage — completion of a bank is not the same thing.
Can one question bank be enough for RACGP Fellowship? For the AKT it can carry much of the knowledge load, and a KFP-style bank helps with the format, but one bank cannot be enough for the fellowship as a whole because it cannot assess consultation performance at all. The CCE is a separate performance exam that requires timed, human-scored practice; a bank is a component of preparation, not the whole of it.
What should I measure instead of my overall Q-bank percentage for RACGP Fellowship? Measure unseen, timed accuracy on AKT-style items by curriculum context; your precision on KFP-style key steps (over-calling and under-calling, not just right or wrong); and, for the CCE, your consistency against the official assessment criteria across unfamiliar cases judged by a human. An aggregate percentage hides all three. The wider caution is in why your Q-bank percentage is not your exam score.
When should I stop doing new RACGP Fellowship questions? For the written exams, stop adding new questions once your unseen scores have plateaued and your remaining errors are about exam technique and Australian-context specifics rather than missing knowledge; then move to timed mixed papers and KFP precision drills. For the CCE, the question of stopping does not apply — that preparation is consultation practice, not question volume.
Which RACGP Fellowship resource should I use for my weakest component? For weak key-feature reasoning, use option-hidden KFP practice marked against the official guidance. For weak Australian-guidance recall, drill source-and-date against Therapeutic Guidelines and RACGP material. For weak consultation skills, book timed role-play with a supervisor or trained peer scoring against the CCE criteria — an AI tool cannot fill that gap. Match the modality to the deficit rather than defaulting to more MCQs.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 20 July 2026. Exam-format details — AKT and KFP question counts, delivery mode and duration, and the CCE structure — should be confirmed on racgp.org.au, as they are refined between diets; any AI-generated guidance claim must be verified against current Australian primary sources. Disclosure: iatroX operates a competing question bank covering AKT and KFP-style knowledge; its role here is confined to the unseen written-knowledge and reasoning layer, and it is explicitly not a Clinical Competency Exam simulator — the consultation stage requires human-scored practice. Corrections are welcome via the feedback route on iatrox.com.
References: Royal Australian College of General Practitioners — AKT, KFP and CCE exam pages and curriculum (racgp.org.au); Therapeutic Guidelines and RACGP guidelines for Australian primary-care content; iatroX internal resources — calibrating automated feedback, auditing an AI medical exam tutor and the comparison hub.