Live transcript
How it works
Each session mirrors the real OSCE format — from candidate brief to graded feedback.
Read the brief
You receive a candidate brief just like the real exam — presenting complaint, patient details, and your task.
Start the conversation
Speak to the AI patient using your microphone. They respond with natural speech, emotional cues, and clinically accurate answers.
Get graded instantly
Your full transcript is analysed against an examiner-calibrated rubric across the three OSCE domains.
Review and improve
See exactly where you gained and lost marks, with specific advice on what to change next time.
Why speaking out loud changes everything
Most OSCE preparation is silent. You read a scenario, think about what you'd say, maybe jot down some notes, and move on. But the actual exam isn't silent — it's a live, spoken conversation with a real person. The gap between thinking about what you'd say and actually saying it under pressure is enormous, and it's where most candidates lose marks.
When you practise with AI voice patients, you build the muscle memory of actually speaking through a consultation. You learn to structure your opening, transition smoothly between sections, and respond to unexpected answers — all while managing an 8-minute timer. This is fundamentally different from reading model answers.
Our AI patients don't follow a rigid script. They respond dynamically based on what you ask. If you ask about the pain, they'll describe it. If you skip social history, they won't volunteer it. If you show empathy, they open up. This means every session is different, and you're genuinely practising clinical reasoning — not memorising a flowchart.
The voice model has been built with input from a team with over a decade of experience in medical licensing exams. Every patient persona includes realistic emotional responses, appropriate levels of anxiety or confusion, and clinically accurate medical histories that match the GMC MLA Content Map 2026.
After each station, your entire conversation is transcribed and graded against a structured rubric. You get domain-specific scores (Data Gathering, Clinical Management, Interpersonal Skills) and line-by-line feedback showing exactly which questions earned marks and which were missed. This is the feedback loop that turns practice into improvement.
What you get in every session
Natural voice interaction
Real-time speech — not text chat. The patient responds with natural pacing, pauses, and emotional tone.
Emotional realism
Patients express anxiety, frustration, or confusion based on the clinical context. Build your empathy skills.
Dynamic responses
No two sessions are identical. The AI adapts its answers based on your questions, just like a real patient.
All 6 station types
History taking, counselling, acute/emergency, ethics, teaching, and procedure stations — all supported.
Safety netting practice
Learn to deliver red flags, follow-up instructions, and when-to-return advice naturally in conversation.
Exam-format timing
Built-in countdown timer matches your exam format — 8 minutes for PLAB 2, 10 minutes for MLA OSCE.
1,800+ scenarios
Every scenario from the GMC MLA Content Map 2026 — 25 specialty areas, 217 presentations, 430 conditions.
Full transcript
Complete word-for-word record of your consultation, with timestamps and speaker labels.
Structured grading
3-domain scoring (DG, CM, IPS) with rubric-level detail on every session.
Frequently asked questions
Very. The AI uses a voice model built specifically for medical consultations. Patients respond with natural speech patterns, appropriate emotional reactions (anxiety, confusion, relief), and clinically accurate histories. Most students tell us they forget they're talking to AI within the first minute.
Just a device with a microphone and internet connection. A laptop or desktop with a headset works best, but you can also use a tablet or phone. No downloads or special software required — everything runs in your browser.
Each scenario has a detailed patient persona including medical history, social history, presenting complaint, ideas/concerns/expectations, and clinical findings. The AI stays in character throughout the conversation, responding naturally to whatever you ask while remaining clinically accurate.
Yes, and you should. Each attempt generates slightly different conversation dynamics, so you'll face new variations even on repeated scenarios. This is great for refining your approach and building consistency.
It complements peer practice. The advantages are: available 24/7, consistent quality, no scheduling needed, objective grading every time, and 1,800+ scenarios versus the handful a friend can simulate. Many students use both — AI for daily volume practice, friends for exam-day simulation.
All six: history taking, counselling, acute/emergency, ethics and professional dilemmas, teaching/explanation, and procedure stations. Each type has specific AI behaviours — for example, acute stations add time pressure, and ethics stations present genuine dilemmas.