AI mock viva practice only helps if it trains the behaviour the exam rewards. For Final FRCA SOE prep, use it to answer aloud, commit to a safe first plan, tolerate interruption, and then repeat the same stem until your structure is tighter and faster.
The current Final FRCA Structured Oral Examination (SOE) is still a two-part oral for current candidates: SOE 1 contains four short cases with linked clinical science across two 26-minute parts, and SOE 2 contains a two-section long case plus two short cases. RCoA has also published the move to new FRCA formats from July 2027, so this workflow is aimed at candidates preparing for the legacy SOE in academic year 2026–27.
If you're using Viva Examiner or any similar AI tool, treat it as a drill partner, not an oracle.
Why AI mock viva practice works for this exam
RCoA guidance is very clear about what strong candidates sound like. Examiners are impressed by answers that are ordered, systematic, coherent, and structured, and the common failure pattern is not practising spoken answers enough. That is exactly where AI mock viva work can help: repetition, pressure, interruption, and fast feedback.
RCoA also provides sample SOE questions and videos showing pass and borderline performances, and explicitly says those clips are not model answers. Use those official examples to calibrate standard, then use AI to multiply your rehearsal volume between human mocks.
The practical point is simple: let the AI train delivery and structure, while you keep humans involved for nuance, judgement, and fact-checking.
Match your AI mock viva to the exam you actually sit
Do not ask for random anaesthesia viva questions and hope for transfer. Build sessions around the real structure. The current SOE samples general duties, essential curriculum units, and sometimes an optional unit, while the January 2026 syllabus version update appears to have been mainly editorial and clarificatory rather than a wholesale blueprint change. That means your prep should stay focused on broad Stage 2-level clinical anaesthesia and linked science, not rumours of a brand-new topic map.
Build separate workflows for SOE 1 and SOE 2
For SOE 1, ask the AI to run short cases with a linked science question.
- Use 13-minute case blocks.
- Make the AI switch between clinical reasoning and applied science.
- Alternate anatomy and physiology topics with pharmacology and physics or clinical measurement.
- Ask it to challenge your choice of technique, drug, or monitor with at least two follow-up questions.
For SOE 2, build a different rhythm.
- Start with a long case that forces prioritisation.
- If your platform allows clinical material, use ECGs, blood gases, imaging, or theatre charts. If not, use a text description and summarise it out loud before answering.
- After the long case, force an abrupt switch to two short cases so you practise resetting quickly.
- Keep the emphasis on safe decision-making, not polished speeches.
A good AI session feels slightly awkward. That is the point.
Set up an AI mock viva that feels real
Use voice if you can. Typing tidy paragraphs is useful for notes, but it is poor preparation for an oral exam. Ask the AI to give one question at a time, avoid hints during the station, interrupt if you drift, and only debrief once the case is over.
Prompt template for SOE 1
Act as an examiner for the Final FRCA Structured Oral Examination (SOE).
Run one SOE 1 case for me.
Rules:
- Ask one question at a time.
- Make this a clinical short case with a linked clinical science question.
- Do not teach or give hints during the station.
- If I speak for more than 90 seconds without structure, interrupt and ask me to prioritise.
- Challenge unsafe or vague answers.
- After the station, give feedback under: safety, prioritisation, structure, applied science, communication, and omissions.
- Then ask me to repeat my opening answer in 30 seconds.
Prompt template for SOE 2
Act as an examiner for the Final FRCA Structured Oral Examination (SOE).
Run one SOE 2 rehearsal for me.
Rules:
- First give me clinical material or a text description to review.
- Then ask me a two-part long case.
- Follow this with two unrelated short cases.
- Do not rescue me early.
- If my answer lacks a plan, ask: What would you do next?
- After the station, mark my performance as pass, borderline, or fail for each area: safety, judgement, prioritisation, communication, and depth.
- Finish by asking me for a corrected version of my first answer.
You can ask the AI to use pass, borderline, and fail labels, but treat those labels as coaching shorthand rather than prediction. The real exam is marked by two independent examiners across 12 questions, with 0, 1, and 2 marks and a standard-set pass mark.
The best prompt is the one you can reuse five times a week without editing.
Answer like an examiner can follow you
Most weak AI practice fails for a boring reason: the candidate knows plenty but says it in the wrong order. In this exam, your first 20 to 40 seconds matter. They tell the examiner whether you can identify the problem, prioritise risk, and build a sensible plan.
A workable structure for short cases
For a clinical short case, try this sequence.
- Start with a one-line frame: what is the case, and why is it high risk?
- Clarify the missing facts that would change management.
- Give a practical plan in sequence: preoperative, induction, maintenance, emergence, postoperative care.
- Name the main complications you are trying to avoid.
- Finish with what would make you escalate, postpone, or change course.
A clean opening sounds like this: a 67-year-old with severe aortic stenosis for urgent laparotomy is high risk because fixed outflow obstruction and emergency surgery make hypotension, ischaemia, and postoperative organ failure more likely. I would first clarify symptoms, valve severity, urgency of surgery, current haemodynamics, and optimisation already achieved.
That answer is not fancy. It is useful. Examiners can follow it.
A workable structure for linked clinical science
Do not turn the linked science question into a detached textbook recital. Keep it clinical.
Use this sequence.
- Name the principle.
- Explain the mechanism briefly.
- Apply it to this patient.
- State the clinical consequence.
- Link it back to your anaesthetic plan.
For the same severe aortic stenosis stem, a linked science answer might be: coronary perfusion depends on diastolic pressure and is threatened when systemic vascular resistance falls or tachycardia shortens diastole. That is why I would avoid sudden vasodilation, maintain sinus rhythm, and have vasopressor support ready.
A worked AI rehearsal example
Ask the AI for a stem like this: a 32-year-old with septic shock needs source control laparotomy and has worsening acidosis on noradrenaline.
A stronger opening answer would be:
- This is a time-critical, high-risk case.
- My priorities are resuscitation, source control, haemodynamic stability, and postoperative critical care.
- I would clarify airway status, current vasopressor dose, recent blood gas, lactate trend, renal function, coagulation, antibiotics given, and aspiration risk.
- I would prepare invasive monitoring early, optimise vascular access, use a haemodynamically stable induction plan, anticipate major vasopressor requirements, and plan postoperative ICU.
Then force the AI to ask linked questions such as:
- Why does severe metabolic acidosis alter the response to catecholamines?
- What does positive pressure ventilation do to venous return in this patient?
- Why might ketamine still produce hypotension in advanced septic shock?
That is high-value AI mock viva practice. It pushes you from recital into explanation.
Turn each AI session into marks
Raw repetition is not enough. The gain comes from the debrief.
After each case, ask for a very short post-viva review.
- What was my best opening sentence?
- What were the three biggest omissions?
- Where did I lose prioritisation?
- Did I give any unsafe or overconfident statements?
- Which science point did I mention without applying it clinically?
- What is the one sentence I should use next time?
Then close the transcript and answer the same case again from memory. That second attempt is where learning sticks.
If you have already had an unsuccessful sitting, use official feedback to sharpen the loop. RCoA states that unsuccessful SOE candidates may request examiner comments, and candidates who have failed more than once may request a guidance interview; by the sixth attempt, guidance is mandatory. Those comments are far more useful if you convert them into a checklist for your next AI sessions.
Common mistakes
- Using AI as a fact source instead of a rehearsal tool.
- Typing answers instead of speaking them.
- Giving long essays before stating the main risk.
- Forgetting that the science question must still feel clinically relevant.
- Accepting flattering AI feedback without checking whether the answer was actually safe.
- Practising favourite topics and avoiding pain, transfer, obstetrics, paediatrics, or ICU crossover.
- Never rehearsing interruption, redirection, or uncertainty.
- Memorising polished model answers that collapse as soon as the case changes.
Practice workflow
For most trainees, the sweet spot is short, frequent, slightly uncomfortable practice. Three focused AI sessions a week will usually beat one heroic four-hour session on a Sunday.
A workable week looks like this.
- Session 1: one SOE 1 short case plus linked science.
- Session 2: one SOE 2 long case.
- Session 3: two mixed short cases with deliberate interruption.
- Session 4: one human mock or peer viva every 1 to 2 weeks.
- Debrief: keep a running error log under structure, safety, science, and communication.
Once a month, compare your answers with the official RCoA example questions and performance videos so your standard does not drift.
If the AI gives you an answer that sounds neat but unfamiliar, do not memorise it. Check the syllabus, current guidance, and your trainers.
Summary
- Use AI mock viva practice to train spoken structure, interruption handling, and repetition.
- Split your rehearsal into SOE 1-style and SOE 2-style sessions.
- Ask for short post-case debriefs, then repeat the same stem from memory.
- Keep linked science clinical: principle, mechanism, patient relevance, consequence.
- Let AI increase volume, but keep humans involved for judgement, calibration, and fact-checking.
References
- https://www.rcoa.ac.uk/examinations/final-frca-examination/final-frca-soe
- https://www.rcoa.ac.uk/examinations/final-frca-examination
- https://www.rcoa.ac.uk/examinations/final-frca-examinations/final-frca-soe/final-frca-soe-resources
- https://www.rcoa.ac.uk/sites/default/files/documents/2026-01/Final-FRCA-Syllabus.pdf
- https://www.rcoa.ac.uk/examinations/2027-launch-new-frca-exams
- https://www.rcoa.ac.uk/examinations/2027-launch-new-frca-exams/final-frca-exam-pathways
- https://www.rcoa.ac.uk/examinations/2027-launch-new-frca-exams/planning-new-examinations
- https://www.rcoa.ac.uk/examinations/2027-launch-new-frca-exams/new-exam-format-faqs
- https://www.rcoa.ac.uk/examinations/frca-examination-information/examination-calendars
- https://www.rcoa.ac.uk/sites/default/files/documents/2026-05/P005_Candidate%20Examination%20Feedback%20Policy_April2026.pdf
- https://www.rcoa.ac.uk/media/49226
- https://rcoa.ac.uk/examinations/final-frca-examinations/final-frca-soe/final-frca-soe-resources
- https://rcoa.ac.uk/examinations/final-frca-examination/final-frca-soe