Use AI mock vivas for FRCEM OSCE prep to rehearse short, spoken, examiner-facing answers under real time pressure. RCEM’s current FRCEM OSCE is a 16-station exam with one minute of reading time and eight minutes in each station, and the blueprint spans complex stable patient, EPIC, supervision and teaching, three resuscitation stations, injured patient, paediatric emergency medicine, management, and management or research.
That means your AI practice needs to look like stations, not essays. RCEM uses domain-based marking, with one to four domains tested per station and a 0 to 10 score for each station; across the exam, the available domains include clinical reasoning, communication, organisation and prioritisation, leadership and management, and research and data management.
If the model is letting you ramble for three pages, you’re not rehearsing this exam. You’re just revising badly.
Why AI practice can work for this exam
The FRCEM OSCE is not a classic viva, but many stations are won or lost on spoken performance: how quickly you frame risk, how clearly you lead, and whether your plan sounds safe when the examiner interrupts you. AI is useful because it can give you repeated station reps on a tired Tuesday night, when organising a full group mock is unrealistic.
The pass rule raises the stakes in resus. RCEM states that the FRCEM OSCE includes three resuscitation stations and that you must pass at least one of them as well as achieve the overall pass mark; from the November 2025 diet, RCEM moved resus pass-fail decisions to Borderline Regression and said candidates do not need to prepare differently.
Used well, AI helps you practise three things fast:
- opening the station with a safe priority statement
- keeping structure when new data arrive
- closing with disposition, escalation, and risk
The practical takeaway is simple: use AI for repetition and verbal fluency, then use humans for realism, nuance, and bedside manner.
How to build AI mock vivas for FRCEM OSCE prep
The safest workflow is to force the model to behave like the current RCEM blueprint. Build separate prompt templates for complex stable patient, resus, injured patient, PEM, supervision and teaching, management, and management or research. For the critical appraisal element, RCEM says candidates receive the journal article four weeks before the OSCE starts and the station becomes a guided conversation about that paper, so your AI drills should interrogate a real article rather than invent a fake abstract.
Also, do not rebuild your whole prompt bank around rumours of a new exam. RCEM’s curriculum updates, approved in July 2025 and effective from 6 August 2025, did not change the underlying clinical syllabus or the overall curriculum structure.
A prompt that is actually useful
Paste a template like this into Viva Examiner and reuse it with different cases:
Act as an FRCEM OSCE examiner.
Choose one station type from:
- complex stable patient
- resus
- injured patient
- PEM
- supervision and teaching
- management
- management or research
Give me:
- a one-minute reading stem
- the task
- any opening observations or results that would be visible at the start
Then stop and wait.
After I answer aloud, mark me on up to 4 relevant domains only.
Use domain-style feedback on:
- clinical reasoning and decision making
- communication
- history or information gathering if relevant
- examination or practical skills if relevant
- teaching if relevant
- organisation and prioritisation
- leadership and management if relevant
- research and data management if relevant
After marking, give:
- my best line
- my first unsafe omission
- the point where I lost structure
- a better 30-second opening
- one follow-up question for an immediate redo
Do not coach me during the station. Interrupt only if an examiner would.
Two rules make this work. Speak your answer out loud, and do not read from notes until the debrief.
Make the timing unforgiving
Set one minute to read, then answer. If you want extra pressure, ask the AI to interrupt at minute 5 or 6 with a new observation, a distressed relative, or a nurse asking what you want done now.
That matters because examiners are not listening for a perfect textbook chapter. They are listening for whether you can impose order on a messy emergency department problem.
Use micro-frameworks, not scripts
For most adult clinical stations, a short internal scaffold is enough:
- first line: sick or not sick, immediate priorities, call for help if needed
- middle: focused history, examination, investigations, and first treatment steps
- final line: senior escalation, disposition, reassessment, safety net
A 67-year-old with chest pain who becomes clammy and hypotensive does not need a cinematic monologue. A better opening is brief: unstable adult, monitored bed, ABCDE approach, ECG, IV access, analgesia, likely acute coronary syndrome until proved otherwise, and early senior help. Then you build from there.
Match the drill to the station type
Resuscitation stations
In resus, AI is best used to drill your first 60 to 90 seconds. Ask for cardiac arrest, peri-arrest arrhythmia, shocked sepsis, severe asthma, toxicology, or the crashing patient whose physiology is deteriorating in front of you.
Your target is not elegance. It is safe sequence:
- recognise instability
- call for the right help
- start structured resuscitation
- request immediate actions from the team
- state your reassessment loop
- name the decision point that changes the next step
If the AI feedback says your answer was knowledgeable but late to declare priorities, believe it.
EPIC, management, and supervision stations
These stations punish generic clerking language. You need to sound like a registrar who can manage flow, conflict, supervision, and risk in a crowded department.
Use AI to rehearse phrases such as:
- what I am worried about right now
- what I need the team to do in the next 10 minutes
- who I would escalate to and why
- how I would explain a difficult decision to a colleague or relative
- what I would document after the conversation
If your answer could have been said by a medical student on a ward round, it is not senior enough.
Management or research stations
RCEM’s current blueprint includes management stations and management or research stations, and the FRCEM domain set includes leadership and management plus research and data management.
For these, get AI to probe method rather than trivia. A good critical appraisal loop is:
- what was the question
- was the design appropriate
- how were patients selected
- what biases matter most here
- what is the size and precision of effect
- do the outcomes matter to emergency practice
- would this change what I do on my next shift
When the pre-released paper arrives, load that actual paper into your sessions. Do not waste the last month practising on invented studies.
Debrief like an examiner, not a fan
RCEM feedback letters break OSCE performance down by station and SLO area. Your AI debrief should mimic that level of specificity rather than telling you that you were good overall.
After every station, write down five things:
- station type
- first missed safety point
- first vague sentence
- best phrase worth keeping
- one sentence you will use differently next time
Then redo the same station within 24 hours. That second attempt is where most of the gain lives.
A simple debrief grid
Ask the AI to answer these questions after each station:
- What would worry an examiner in my first 30 seconds?
- Did I commit to a diagnosis or management plan too early?
- Did I sound senior, or just busy?
- Which single sentence would have lifted my score most?
- What should I say if interrupted here?
This is how you turn AI from a novelty into deliberate practice.
Common mistakes
- Using AI to generate model answers before you have attempted the station yourself.
- Typing polished answers instead of speaking under time pressure.
- Letting the model praise you for being broad when the station needed prioritisation.
- Rehearsing outdated exam formats. RCEM’s current Fellowship by examination is FRCEM SBA plus FRCEM OSCE; older standalone components such as the FRCEM Final Critical Appraisal, FRCEM Final QIP, Management Viva, and CTR Viva are historical, not current live exam targets.
- Using the same prompt for every station type.
- Ignoring physical examination and actor interaction practice because AI feels more convenient.
- Trusting invented references, guidelines, or statistics without checking them.
A realistic practice workflow
You do not need marathon sessions. You need short, repeatable reps that survive a run of nights.
A workable weekly loop looks like this:
- 2 evenings: two timed AI stations back to back, then 10 minutes of debrief notes
- 1 short session: openings only, where you practise the first 30 seconds of 6 to 8 different stems
- 1 human session each week or fortnight: colleague, senior, or study group for realism and interruption handling
- 1 review block at the weekend: revisit your error log and rerun the three weakest stations
If you already have an RCEM feedback letter from a previous sitting, use the low-scoring station or SLO areas to build your next two weeks of prompts. RCEM says the feedback letter includes your score, the pass mark, cohort average, and a station or SLO breakdown of performance.
When to stop using AI and find a human
Switch to human practice when the problem is not structure but presence. If your issue is rapport, eye contact, tone under pressure, examination fluency, or how you handle interruption from an actor or examiner, AI can help less than a blunt colleague with a stopwatch.
Use the right tool for the right weakness.
Summary
- Build AI mock vivas around the current FRCEM OSCE blueprint, not a generic viva format.
- Rehearse aloud with one-minute reading time and eight-minute stations.
- Mark yourself on structure, prioritisation, and safe escalation, not on how encyclopaedic you sound.
- Use the pre-released journal article for research practice and your official feedback letter for targeted redo sessions.
- Keep AI in its lane: repetition and debrief are excellent; human realism is still essential.
References
- https://rcem.ac.uk/frcem-exams/
- https://rcem.ac.uk/osce-exams-faqs/
- https://rcem.ac.uk/wp-content/uploads/2023/11/FRCEM_Regulations_and_Information_Pack.pdf
- https://rcem.ac.uk/wp-content/uploads/2025/07/Domain_Based_Marking.pdf
- https://rcem.ac.uk/wp-content/uploads/2024/11/FRCEM_Exam_Regulations_2025_V1.pdf
- https://rcem.ac.uk/results-feedback-and-awarding/
- https://rcem.ac.uk/college-news/change-to-frcem-osce-passing-criteria-for-resus-stations/
- https://rcem.ac.uk/college-news/gmc-approves-rcems-training-curriculum-updates/