AI simulated patients for medical education

Clinical conversation stations with AI simulated patients, in chat and real-time voice. History taking, breaking bad news, and OSCE preparation, with competency-based scoring and documented evidence per student.

Request Demo See plans

Risk to real patients while students practice

24/7

History-taking practice without booking an actor or a tutor

100%

Sessions documented with transcript and per-criterion score

Minutes

To build or adjust a station with the AI assistant

50+

Gamification badges to keep cohorts engaged

Languages, for international programs and exchange students

Real challenges

What blocks training today

What we hear from training leaders in real conversations. No fluff.

Standardized patients are expensive and do not scale

Recruiting, training and scheduling actors consumes budget and limits practice to a handful of sessions per term. The result: each student talks to a simulated patient only a few times across the whole program. In Roleplays, the same station runs for the entire cohort, on the same day, with no per-session cost.

Students reach clinical rotations without enough conversations

Years of theory and multiple-choice exams, and then the first real history taking happens with a real patient in front of a preceptor. Clinical conversation is a motor skill like any other: without repetition, it never becomes reflex. With AI patients, students accumulate dozens of interviews before touching a real patient.

Communication assessment is subjective

The same performance gets different marks depending on which examiner sits in the station. In Roleplays, criteria are locked before the session starts and every score carries the transcript quote that justifies it. Faculty review evidence instead of relying on hallway memory.

Feedback arrives late or never

No tutor has the bandwidth to observe and debrief dozens of individual conversations every week. On the platform, every session returns per-criterion scores and specific feedback instantly, and faculty spend their time on the cases the AI flagged as critical.

Running an OSCE is a logistics operation

Rooms, actors, examiners, rotations, printed checklists. The exam runs once or twice a year because running it more often is unaffordable. The AI-simulated version does not replace the in-person exam, but it lets students practice the station format as many times as they want before the official day.

Breaking bad news is taught on slides and practiced on real families

Frameworks like SPIKES are taught in lectures and rarely rehearsed, because the scenario is emotionally hard to roleplay between classmates. The AI persona sustains the patient's emotion (fear, tears, denial, anger) without peer embarrassment, and students repeat the conversation until they can lead it with confidence.

Pedagogical documentation is manual

Accreditation reviews and curriculum committees ask for evidence that communication competencies were taught and assessed. Attendance sheets and paper checklists do not tell that story. Every Roleplays session generates a timestamped record with transcript, criteria and scores, exportable per student, cohort or term.

In large cohorts, practice is unequal

In group dynamics, the same volunteer student always performs while everyone else watches. With individual AI stations, every student leads their own conversation at their own pace, and the dashboard shows faculty exactly who practiced and who is avoiding it.

How Roleplays solves it

For every pain, a concrete answer

Segment-specific features mapped to each pain point above.

Configurable simulated patients

Age, chief complaint, medical history, personality, health literacy and emotional state. The AI patient improvises within the case defined by faculty: it withholds information until asked, shifts mood based on how the interview is conducted, and uses lay language that forces students to translate jargon.

OSCE-style stations with your own rubric

Your institution's checklist becomes a weighted rubric on the platform. Criteria are locked when the station starts, each one scored 0-100 with a transcript quote as evidence, and criteria marked as critical can fail the station on their own.

Communication frameworks as criteria

SPIKES for bad news, Calgary-Cambridge for the consultation, or your department's own protocol. Each step becomes a scoreable criterion: students see exactly which step of the conversation cost them points and why.

Real-time voice, not just text

Real history taking happens out loud. In voice stations, tone, pace and hesitation are also scored, and the persona reacts to how the student speaks, not just to content. Filler words and hedging language are measured per session.

Adaptive remediation per competency

A student scored low on active listening? The platform automatically generates a new station focused exactly on that gap. Remediation is individual, without making the whole cohort repeat what they already master.

Dashboards per cohort and per student

Program directors and faculty track time series per competency, compare cohorts and terms, and identify who needs attention. Every number is clickable down to the transcript that produced it.

Programs per term

Combine theory content, practice stations and preceptor validation into a reusable journey per academic term. Build it once, apply it to every following cohort with the same criteria.

Gamification across cohorts

Badges, levels and challenges configurable by the institution. Clinical conversation practice becomes a weekly habit instead of an exam-eve cram.

Multi-language for international programs

The same station runs in English, Portuguese, Spanish and 20 more languages, with comparable metrics. Useful for medical English courses, exchange programs and double-degree institutions.

Data isolation per institution

Isolated database, corporate SSO (SAML 2.0 and OIDC) to integrate with institutional login, and xAPI integration with the LMS or VLE the institution already uses.

See a real session

One conversation.
One rubric.

An example roleplay in this context. Each turn is scored against your tenant competency framework.

Scenario

Breaking bad news station. The clerkship student must communicate a malignant biopsy result following the SPIKES protocol. The patient came alone, is anxious, and has low health literacy.

Rubric criteria

Rapport and settingPatient's perception (SPIKES)Warning shot and clarity of the newsLay language, no jargonHandling emotion and silencePlan and next stepsFiller words and hesitation

Thomas Carter

Patient, 62 years old, follow-up visit for biopsy results

in session

Doctor, I will be honest with you: I have not slept properly since the receptionist called. She said the biopsy results were in and that I should come in person. When it is good news, they just tell you on the phone, don't they?

Mr. Carter, thank you for coming in. I can only imagine how hard this wait has been, and we are going to go through the results together, calmly and with as much time as you need. First, could you tell me what you understand so far about why we did this biopsy?

The other doctor said there was a spot on my bowel exam and they needed to take a small piece to analyze. I looked it up on the internet, and my daughter told me to stop looking things up... Doctor, is it cancer? You can tell me the truth.

Mr. Carter, you asked me for the truth and that is what you will get. I am afraid the results are not what we were hoping for. The analysis shows that the lesion is a malignant tumor. I know this is very difficult news to receive, and I want you to know that there are treatment options and that you will not go through this alone.

Type your reply...

Frameworks & assessment

Aligned with how medical education already assesses

From OSCE checklists to communication frameworks: the platform adapts to your institution's assessment model, not the other way around.

OSCE methodology

The Objective Structured Clinical Examination organizes assessment into stations with checklists. In Roleplays, each checklist becomes a weighted rubric, locked when the session starts, with a score per criterion and a transcript quote as evidence. Students run the station as many times as they need before the in-person exam.

Communication frameworks

Calgary-Cambridge for the consultation, SPIKES for breaking bad news, or your faculty's own protocol. Each step of the framework becomes a scoreable criterion, so students see exactly where in the conversation they lost points and why.

Licensing & residency exams

Medical licensing and residency selection processes worldwide use simulated-patient stations. Institutions and prep programs build station banks in Roleplays so candidates can practice the exact exam format with immediate feedback on every attempt.

GDPR & LGPD

Student data lives in an isolated database per institution, encrypted in transit and at rest, with configurable retention. Simulations use fictional AI-generated patient personas: no real patient data ever enters the platform.

FAQ

Frequently asked questions

Questions that come up in almost every first conversation.

Can we turn our OSCE checklist into platform criteria?

Yes, that is the standard implementation path. Each checklist item becomes a weighted criterion in the station rubric. Criteria are locked the moment the session starts, each one scored 0-100 with a transcript quote as evidence, and eliminatory items can be marked as critical, failing the station on their own. Faculty can revise the rubric at any time; sessions in progress are not affected.

Does the AI simulated patient replace standardized patient actors?

No, and that is not the goal. Trained actors and in-person exams remain the high-fidelity standard, especially for physical examination. What AI solves is the volume of deliberate practice actors cannot cover: students arrive at the in-person OSCE having led dozens of scored conversations instead of two or three. Each complements the other.

Is AI assessment reliable for communication skills?

Assessment runs against objective criteria defined by faculty and locked before the session, never against a generic rubric. Every score comes with the exact transcript quote that justifies it, so faculty can audit any score in seconds. This removes the variability of different examiners applying the same checklist, one of the most common criticisms of the traditional format.

Which courses and program stages does it serve?

Clinical skills and communication courses in early years, practice stations during clerkships, preparation for licensing and residency exams, and continuing education for residents and preceptors. Beyond medicine, the same model serves nursing, psychology, dentistry and clinical pharmacy, any program where talking to patients is part of the competency.

Can students practice off campus?

Yes. The platform runs in the browser and has native iOS and Android apps. Students practice stations at home, on rotation or between classes, by chat or by voice. Every attempt is recorded with score and transcript, and faculty see the full history.

How does the program office track usage and results?

Dashboards per cohort, term and student, with time series per competency. Coordinators see who practiced, how many times, on which stations and with what score evolution. Reports are exportable for accreditation reviews and curriculum committees, with a timestamp on every session.

What about student data privacy?

Each institution gets an isolated database, with encryption in transit and at rest and configurable retention. We are GDPR and LGPD aligned. Patients in the simulations are fictional AI-generated personas: no real patient data ever enters the platform. Access can be integrated with institutional login via SSO (SAML 2.0 and OIDC).

How long does implementation take for a medical school?

Typical implementation takes 2 to 4 weeks: first we build the initial stations with faculty and convert checklists into rubrics; then we run a pilot with one cohort to calibrate personas and criteria; in the following weeks we expand to the remaining cohorts with faculty training on the dashboard. There is a dedicated implementation team, not chatbot onboarding.

What is the commercial model for educational institutions?

The model is per active user license, with specific conditions for student volume. The best path is to book a demo with a real case from your institution: we build a station with your checklist and you evaluate the scored transcript before any decision.

Back to overview

Ready to transform how your team trains?

For organizations with 50+ employees. Book 45 minutes and we'll think the setup through with you.

Request Demo Talk to Sales