Viva Examiner
Start now
Viva Examiner
Toggle sidebar
How to Use AI Mock Vivas for MRCGP SCA Prep

Specialty Exam Guides Family Medicine

How to Use AI Mock Vivas for MRCGP SCA Prep

UK MRCGP SCA guide to AI mock viva practice: use Viva Examiner to rehearse exam-shaped consultations, sharpen feedback, and avoid scripted answers.

Used properly, AI mock vivas can make your MRCGP SCA prep far more exam-shaped. The key is simple: practise the real task, not a loose chatbot conversation. That means a short reading phase, a focused 12-minute consultation, and a blunt debrief against the three domains examiners actually use: Data gathering and diagnosis, Clinical management and medical complexity, and Relating to others.

Many trainees use AI as a question bank or a source of model phrases. That is not enough. The Simulated Consultation Assessment is built around 12 simulated consultations, some video and some audio only, drawn from a large case bank mapped to the curriculum, and RCGP is explicit that there is no single preferred consultation model or script. If your AI practice is vague, overlong, or rehearsed-sounding, you may be training the wrong habits.

The good news is that the exam shape remains stable after the August 2025 curriculum update: RCGP states that the update did not change the structure or format of the SCA. So an AI workflow that mirrors the current consultation task is still the right way to rehearse.

Match the exam, not the chatbot

Your first job is to force the AI session to resemble the real consultation. The SCA gives you 3 minutes of reading time and 12 minutes for the case. Physical examination is not assessed, some cases are audio only, and the exam is delivered remotely from a local GP surgery through the Osler platform. Build your rehearsal around those realities.

A useful rule: each AI session should do one of three jobs.

  • Case generation: produce a realistic GP brief with enough ambiguity to think.
  • Live rehearsal: let you speak through the consultation under time pressure.
  • Debrief: analyse what helped or hurt you in each domain.

If you try to do all three at once, the session often turns messy. Keep the task narrow.

Set up the case properly

Ask the AI to give you only what you would have at the start of the station: age, setting, a one-line presenting problem, consultation type, and any pre-supplied information. RCGP notes that the purpose of a case may not always be obvious from the reading material, so your prompts should leave some uncertainty rather than telegraph the diagnosis.

A prompt that works well is:

Create one RCGP SCA-style GP consultation brief.
Give me only:
- patient age and role
- consultation type: video or audio only
- presenting problem
- any pre-supplied note or investigation result
- one hidden concern or agenda for the role player
- one clinical risk issue if relevant
Do not reveal the diagnosis or ideal management.
Wait for me to start the consultation.

That last line matters. You want to think, not recognise a pattern and recite a memorised answer.

Rehearse on the clock

Use a timer. Three minutes to read. Twelve minutes to consult. If you have voice mode available, use it. The SCA is about what you can say clearly, safely, and naturally under pressure, not what you can type after a pause.

In the 3-minute read, train yourself to spot four things:

  • What is the consultation task? Acute problem, explanation, review, negotiation, risk, follow-up?
  • What must I not miss? Red flags, safeguarding, deterioration, suicide risk, sepsis, cancer concern.
  • What is the likely decision point? Reassure, investigate, treat, review, refer, escalate.
  • What is the person-level issue? Fear, agenda, preference, work impact, family pressure, cultural context.

That is usually enough. More than that, and you risk burning time before you even start.

Build an AI mock viva workflow for MRCGP SCA prep

The best AI mock viva workflow for MRCGP SCA prep is not glamorous. It is repeatable.

Phase 1: consult as if an examiner is listening

Aim for a consultation that is selective, not exhaustive. RCGP feedback repeatedly warns against overlong data gathering, vague decision-making, and rushed endings. Examiners want to hear that you can gather enough information to be safe, reach a working diagnosis or analysis, and move into a realistic plan within the time available.

A simple internal structure helps:

  • open and clarify the problem
  • gather targeted clinical and contextual data
  • say your working diagnosis or analysis out loud
  • explain options and agree a plan
  • safety-net and arrange follow-up if needed

Notice the middle step: say your reasoning out loud. RCGP guidance specifically points out that sharing your diagnostic thinking helps the examiner understand how you are making decisions.

Phase 2: get domain-based feedback

Do not ask the AI, “Did I pass?” That is not a useful question.

Instead, ask for feedback under the three official domains:

  • Data gathering and diagnosis: What did I miss, over-ask, or fail to prioritise?
  • Clinical management and medical complexity: Was my plan current, safe, realistic, and suited to UK general practice?
  • Relating to others: Did I respond to cues, explore agenda, adapt explanations, and share decisions well?

A debrief prompt might be:

Debrief my consultation using the three official SCA domains.
For each domain, tell me:
1. one thing I did that helped,
2. one thing that would worry an examiner,
3. one sentence I could have said better,
4. one priority for my next attempt.
Do not give me generic praise.

That last sentence saves time. Generic praise feels nice and teaches very little.

Phase 3: compare the AI debrief with the RCGP toolkit

The RCGP consultation toolkit includes an overview of the consultation and a RAG self-assessment tool. Use that as your anchor. AI can help you generate cases and notice patterns, but your final frame for improvement should stay aligned to the official domains and the toolkit language.

A good habit is to finish every case with two scores of your own before you read the AI feedback:

  • What cost me marks?
  • What would I keep next time?

That stops you becoming passive.

What examiners are really listening for

The exam is not rewarding a polished monologue. It is rewarding safe, flexible consulting.

In Data gathering and diagnosis, examiners want to hear targeted questions, sensible prioritisation, and a coherent working diagnosis or explanation. They do not want an all-inclusive review of systems or a candidate who sits on the fence when the likely diagnosis is already clear.

In Clinical management and medical complexity, they are listening for a management plan that fits current UK practice, makes use of the wider team when appropriate, handles uncertainty, and ends with realistic follow-up or safety-netting. RCGP feedback also flags ineffective time management as a common issue, often because the candidate spends too long on history and rushes the plan.

In Relating to others, they are listening for something more subtle: whether you actually respond to the person in front of you. RCGP warns against formulaic consulting, stock empathy phrases, and missing verbal or non-verbal cues. In practice, that means your AI rehearsal should include hidden worries, awkward pauses, and preferences that force you to adapt.

A worked case: same-day telephone chest pain

Imagine an audio-only case: a 67-year-old man calls with central chest discomfort when walking to the shops.

A weak AI rehearsal becomes a checklist recital. A stronger one sounds more like this: you clarify the immediate risk, establish exertional features and associated symptoms, test for red flags, assess current stability, and then say your concern out loud: “Given the way this comes on with exertion, I’m worried this could be cardiac rather than simple indigestion, so I don’t think this is one to sit on.”

Then you move. You explain the urgency, arrange the next step clearly, check he understands what to do if pain recurs, and avoid pretending that a physical examination has happened. That matters because physical examination is not assessed in the SCA; what matters is your decision-making and communication about what needs to happen next.

A worked case: PR bleeding with cancer anxiety

Now take a video case: a 54-year-old with rectal bleeding who is frightened because a neighbour has recently been diagnosed with cancer.

Here, AI practice is useful if it forces you to do two jobs at once. First, make the consultation clinically safe: duration, change in bowel habit, weight loss, anaemia symptoms, medication, family history, and relevant background. Second, deal with the person’s fear rather than parking it until the last 30 seconds. RCGP feedback specifically values exploration of agenda and a plan that reflects current practice.

Your debrief question after this case should be sharp: Did I sound like someone who noticed the cancer worry early, or someone who discovered it accidentally at the end? That one question often separates a flat pass from a much better performance.

Use AI safely

AI is a rehearsal tool, not your guideline source. The RCGP toolkit highlights outdated or poorly evidenced management as a frequent reason candidates struggle, especially around prescribing, referral, investigations, uncertainty, and safety-netting. If the AI suggests a management plan you would not happily defend in a real GP surgery, do not rehearse it into your muscle memory.

A few rules keep AI useful:

  • Never paste identifiable patient data into a mock case.
  • Use synthetic or heavily de-identified scenarios.
  • Ask for UK primary care context.
  • Challenge management that sounds overconfident or oddly specific.
  • Practise audio-only cases as well as video-style ones, because both formats appear in the exam.

Also, do not let AI train you into a script. The moment every consultation starts with the same empathy line, the same ICE phrasing, and the same closing sentence, you are drifting towards the “formulaic” consulting RCGP feedback warns about.

Common mistakes

The same errors crop up again and again in AI sessions.

  • Taking an exhaustive history and leaving the management plan until the end.
  • Failing to say your working diagnosis or analysis out loud.
  • Giving a management plan that sounds generic rather than NHS-realistic.
  • Missing the patient’s hidden agenda, fear, or preference.
  • Using canned empathy that does not fit the moment.
  • Safety-netting mechanically rather than tailoring it to the case.
  • Forgetting continuity, follow-up, or the role of the wider team.
  • Letting the AI “teach” you an answer you have not checked against current UK practice.

If one of those themes keeps repeating, narrow your next week of practice around it. Improvement is usually domain-specific before it becomes global.

Practice workflow

You do not need endless mocks. You need a sustainable loop.

A practical weekly rhythm looks like this:

  • Two solo AI cases focused on one weak domain.
  • One audio-only case to sharpen listening and verbal clarity.
  • One trainer or peer review using the RCGP toolkit language.
  • One short redo of the same case opening, explanation, or closing after feedback.

Keep a very small error log. Three columns are enough:

  • repeated missed cue
  • repeated management gap
  • phrase or behaviour to try next time

That log becomes gold after 10 to 15 cases. You will usually find that your problem is not “communication” in general. It is something narrower: over-questioning, timid decision-making, poor signposting, weak safety-netting, or missing emotional cues.

If you are getting close to your sitting, add the official RCGP preparation resources to the loop. RCGP provides SCA preparation pages, a consultation toolkit, and webinars for trainees preparing to take the exam. AI works best as an amplifier for that work, not a replacement for it.

Summary

  • Use AI mock vivas to rehearse the real SCA task: timed reading, a focused consultation, then domain-based feedback.
  • Ask the AI to debrief by Data gathering and diagnosis, Clinical management and medical complexity, and Relating to others.
  • Practise saying your reasoning out loud. Examiners cannot reward thinking they never hear.
  • Avoid scripted empathy, generic plans, and unchecked management advice.
  • Pair AI sessions with the RCGP toolkit, official preparation material, and human debrief.

References

Related guides

Practice Family Medicine with a live AI examiner

You have the framework — now run realistic oral-exam cases, answer follow-up questions aloud, and get structured feedback when you finish.

Try it free — start practicing

Instant access No credit card required Cancel anytime

We use cookies as described in our Cookie Policy. By clicking Accept, you agree to our use of cookies. See our Terms and Conditions, Privacy Policy and Cookie Policy.