AI mock vivas for MRCP PACES work best when you use them as a timed examiner, not as a magic tutor. Use them to rehearse the parts of the MRCP(UK) Part 2 Clinical Examination (PACES) that you must say out loud: opening a consultation, explaining risk in plain English, ranking differentials, and handling examiner questions. Then take the same case to the bedside, because the exam still awards marks for physical examination, identifying physical signs, and maintaining patient welfare across five stations and eight encounters.
Why this matters in PACES
PACES23 is built around five 20-minute stations. Stations 2 and 5 are consultations; Stations 1 and 4 each combine a communication encounter with a physical examination encounter; Station 3 combines cardiovascular and neurological examination. In the consultation stations, you get 15 minutes for the patient interaction and 5 minutes for examiner discussion. In the system examination encounters, you have up to 6 minutes to examine and at least 4 minutes for questioning. That is exactly why AI rehearsal is useful: it can reliably drill the verbal, timed, decision-making parts of the station.
Examiners are not just listening for knowledge. MRCP(UK) marks seven skills: physical examination, identifying physical signs, clinical communication, differential diagnosis, clinical judgement, managing patients' concerns, and maintaining patient welfare. Two consultation encounters assess all seven skills, while communication encounters emphasise communication, judgement, concerns, and welfare. If your AI practice never maps feedback onto those skill domains, it is too vague to help.
Build an AI mock viva workflow for MRCP PACES
Start with the official frame, then make the AI copy it. The MRCP(UK) preparation material points candidates to sample scenarios, marksheets, and targeted advice after failure. That gives you a clean workflow on Viva Examiner: run a case, score it against the official skill headings, then repeat the same case after feedback.
Set the brief before each case
Before you start, tell the AI exactly what role it is playing. A loose prompt gives you loose practice.
Use a short brief with these elements:
- station type: Communication, Consultation, or post-examination questioning
- clinical setting: acute, chronic, inpatient, outpatient, or ward review
- candidate level: registrar preparing for PACES
- timing rule: 10 minutes, 15 plus 5, or 6 plus 4 depending on the station
- marking lens: score me against Skills C, D, E, F, and G, or all seven if appropriate
- feedback style: first give examiner-style questions, then a short debrief, then one repeat case
For consultation stations, ask the AI to run a full 15-minute patient phase followed by a 5-minute examiner phase. For system examination discussions, ask it to stop you after 6 minutes, then challenge your findings, differential, investigations, and initial management. Matching the timer to the real station matters; official guidance repeatedly tells candidates to practise with a timer and to be structured and focused within the allowed time.
Match the drill to the station
The quickest way to waste AI practice is to rehearse generic viva chat that does not resemble PACES. Match the prompt to the station you are training.
-
Stations 1 and 4 communication Ask for a patient or relative who needs explanation, shared decision-making, or a difficult conversation. Example brief: act as a PACES surrogate. Give me a case about explaining suspected lung cancer investigations to a worried patient. Mark me on clarity, clinical judgement, managing concerns, and patient welfare.
-
Stations 2 and 5 consultation Ask for a whole-case rehearsal. Example brief: run a 20-minute PACES consultation with a 58-year-old with pleuritic chest pain. Let me take the history, ask for focused examination findings when I request them, then question me as the examiner for 5 minutes.
-
System examination follow-up Use AI after a real bedside examination, not instead of one. Example brief: I have just examined a patient with a pansystolic murmur radiating to the axilla. Ask me to present findings, give the most likely diagnosis, relevant differentials, key investigations, and immediate management.
That last format is especially useful because PACES physical examination encounters do not just test hand movements; they test what you make of the findings and whether you can discuss them coherently under pressure.
What good PACES answers sound like
Official preparation advice is very clear on the failure patterns: unclear communication, non-systematic examination, missed or invented signs, and textbook differentials that do not fit the patient in front of you. Good AI rehearsal should therefore train you to sound structured, relevant, and patient-centred rather than encyclopaedic.
A workable framework for communication encounters
In Stations 1 and 4, keep a simple spine:
- open and set the agenda
- find out what the patient or relative already knows
- explain the issue in plain English
- give the plan and the reason for it
- check understanding, concerns, and questions
- close with a clear next step
A good answer feels calm and deliberate. Suppose a 67-year-old with iron deficiency anaemia asks whether this could be bowel cancer. A weak answer jumps straight to colonoscopy details. A stronger PACES answer names the concern early, explains that cancer is one important possibility but not the only one, sets out the need for urgent investigation, checks what the patient has understood, and avoids false reassurance.
A workable framework for consultation stations
For Stations 2 and 5, think in three passes:
- first pass: define the problem and the patient's agenda
- second pass: test the dangerous diagnoses and do a focused examination
- third pass: summarise, rank your differential, propose investigations, and agree an immediate plan
When the examiner takes over, do not restart the whole case. Switch gears. Give a one-line problem representation first, then your top differential, then the evidence for and against, then the immediate management priorities.
A useful template is:
- summary sentence
- likely diagnosis
- important alternatives not to miss
- focused investigations with purpose
- immediate management
- what you told the patient and what you still need to clarify
A workable framework for post-examination discussion
After a bedside examination, your answer should sound like a physician thinking aloud, not a student reciting a list.
Try this order:
- examination findings in one clean summary
- localisation or syndrome if relevant
- most likely diagnosis
- two or three sensible alternatives
- one or two key investigations
- first-line management and urgency
If the AI keeps telling you that your answer is too long, believe it. PACES rewards relevant discrimination.
What AI can and cannot judge
AI is excellent for structure, transitions, wording, and examiner pressure. It can tell you when your explanation is too technical, when your differential is unfocused, when you failed to answer the actual question, and when you forgot to ask about the patient's concerns. That fits well with the official communication, judgement, and concern-management skills.
AI is poor at the parts that depend on real examination technique and real signs. It cannot tell whether your apex beat localisation was accurate, whether you elicited ankle clonus correctly, or whether you were physically gentle with a tired patient. MRCP(UK) preparation advice explicitly tells candidates to examine many real patients, be observed by senior colleagues, and build confidence in recognising genuine findings.
Use that difference to divide your practice:
- Use AI for: opening structure, question prioritisation, plain-English explanations, differential diagnosis, management summaries, and timed examiner challenge.
- Use real patients and supervisors for: examination technique, sign recognition, pacing around the bed space, professionalism, and patient comfort.
Also, do not let an AI score become a fantasy pass prediction. MRCP(UK) currently states that PACES requires a minimum overall score of 126 and minimum standards across all seven skills. An AI can help you organise feedback against those domains, but it cannot reproduce live examiner calibration or the feel of a real encounter.
Common mistakes
The usual errors are not a lack of knowledge. They are a lack of exam discipline.
- rehearsing generic cases that do not look like PACES encounters
- letting the AI spoon-feed hints instead of making you commit to a differential
- speaking in long paragraphs without a headline summary
- giving a textbook list instead of a ranked differential for this patient
- forgetting to address the patient's actual worry
- using AI to replace bedside examination instead of preparing for the discussion that follows it
- accepting flattering feedback that is not anchored to the official skill domains
These failure patterns closely mirror the official preparation advice: candidates lose marks for non-systematic examination, incorrect technique, missing important findings, unclear information-giving, and irrelevant textbook differentials.
Practice workflow
Keep the cycle short enough to sustain during a busy rota. One polished case is better than five shallow ones.
A realistic weekly loop looks like this:
- Twice a week: one 20-minute AI consultation case and one 10-minute communication case
- Twice a week: real bedside system examination with a colleague or senior watching at least part of it
- Once a week: one mixed mock where you do the bedside examination first, then immediately run the examiner discussion with AI
- At the end of the week: review your transcript or notes and write three lines only: what I missed, what I said well, what I will change next time
The debrief matters as much as the mock. After each case, force yourself to rewrite your answer in a tighter form. Then repeat the same station within 24 to 48 hours. That is how you turn feedback into a habit rather than a good intention.
If you use Viva Examiner, keep your prompts consistent for a fortnight before changing them. You want comparable reps. Otherwise you will feel busy without getting measurably better.
Summary
- Use AI mock vivas for the spoken parts of PACES: communication, consultation structure, differential diagnosis, management explanation, and examiner questioning.
- Keep your practice station-specific. PACES23 has five 20-minute stations, eight encounters, and different timing rules for consultation versus system examination work.
- Score yourself against the official skill domains, not vague impressions.
- Do not let AI replace bedside practice. Real patients are still where you learn technique, genuine signs, and patient welfare.
- Treat AI feedback as rehearsal data, not a pass prediction. The real exam still depends on calibrated examiners and minimum performance across all seven skills.
References
- https://www.thefederation.uk/examinations/paces
- https://www.thefederation.uk/examinations/paces/format
- https://www.thefederation.uk/examinations/paces/preparation
- https://www.thefederation.uk/examinations/paces/marksheets
- https://www.thefederation.uk/examinations/paces/sample-scenarios
- https://www.thefederation.uk/examinations/paces/exam-dates-and-fees
- https://www.thefederation.uk/examinations/paces/paces-faqs
- https://www.thefederation.uk/examinations/guidance-and-information/pass-marks-explained
- https://www.thefederation.uk/examinations/guidance-and-information/regulations
- https://www.thefederation.uk/examinations/guidance-and-information/regulations/limit-attempts
- https://www.thefederation.uk/news/mrcpuk-paces-pass-standard