loading

Buyer's guide

How to compare AI virtual patient platforms

Demo tests and vendor questions, plus a side-by-side of the leading platforms.

By the Lumeto Clinical Education Team · Updated November 2026

Quick answer. Compare AI virtual patient platforms on five things: open conversation, clinical accuracy, vitals that respond to learner actions, faculty authoring, and how the AI scores and debriefs. Test each one live in a demo with your own scenario, including Lumeto's InvolveXR. Then ask about screen access, bias, privacy and published evidence.

Many simulation vendors now call their virtual patients AI-powered. The label covers anything from voice-enabled branching menus to large language model patients that answer open questions and change condition as learners act. Below are the tests and questions that tell them apart. For background, see how AI-powered virtual patients improve hospital simulation.

Key takeaways

  • Ask an unexpected question and check the answer fits the case.
  • Vitals should change with learner interventions.
  • Faculty should edit history, dialogue, vitals and medications themselves.
  • AI scores should use your rubric and show the action behind them.

What is an AI virtual patient platform?

An AI virtual patient platform lets learners interview, assess and treat a simulated patient who replies in natural language and changes condition as they act. It runs on VR, screens or both, often with an AI assessor that scores and debriefs.

A 2019 meta-analysis of 51 trials with 4,696 participants found virtual patients improved skills more than traditional teaching and matched it for knowledge. Much of the evidence was low quality, and it predates large language model patients (Kononowicz et al., JMIR).

How do you tell a conversational AI patient from a scripted one?

Ask something the script would not expect. A conversational patient answers in context. A scripted one falls back to a menu, a stock reply or silence.

An AI-powered virtual patient in InvolveXR responding to a learner (25 seconds).

Demo tests:

  • Ask an open question, such as “What worries you most about going home?”
  • Interrupt, then return to the topic.
  • Ask the same question two ways.
  • Watch for facial expression, posture and tone.

How can you check that a virtual patient stays clinically accurate?

Ask how the vendor keeps AI answers consistent with the case. Then ask about a symptom outside the case history and check the answer fits the diagnosis.

The WHO’s 2024 guidance names simulated patients in medical and nursing education as a use for large multimodal models. It warns they can produce “false, inaccurate, biased, or incomplete statements” (WHO, 2024). A 2025 review found only 3 of 23 studies (13%) used validated tools to evaluate these patients, and noted weak non-verbal and emotional responses (Zeng et al., JMIR).

Should the patient's physiology change when learners intervene?

Yes. If oxygen saturation stays flat after a learner gives oxygen, the scenario tests recall and skips clinical reasoning. Ask to see vitals respond to a correct, wrong and delayed intervention.

The Edit Patient State panel in Lumeto's InvolveXR, with fields for heart rhythm, heart rate and blood pressure
Educators can change a patient's heart rhythm, heart rate and blood pressure during a session.

Also ask if educators can change vitals mid-session. InvolveXR instructors can change vitals and patient states on the fly.

How much control should faculty have over patient authoring?

Faculty should edit patients themselves, beyond picking from a library. Look for control over history, presenting complaint, dialogue, personality, vitals and states, medications and doses, and the checklist.

The AI conversation designer in Lumeto's InvolveXR, showing editable patient background and responses
InvolveXR's conversation designer lets faculty write the patient's background and responses.

In the demo, have a faculty member change a scenario and time it. InvolveXR’s no-code library lets educators create or customise patients, dialogue, medications and checklists.

How do you judge an AI assessor's scoring and debrief?

A good AI assessor scores against your criteria and shows its working. Ask:

  • Can educators load or edit the rubric?
  • Does each score link to a learner action or statement?
  • Can educators override a score and add a note?
  • How is the assessor checked against expert raters?

The debrief should fit your model, such as PEARLS or advocacy-inquiry. It should prompt reflection, point to specific moments and hand off to a live debrief.

Lumeto's Artificial Clinical Facilitator (ACF) guiding a learner inside a simulation.

Lumeto’s Artificial Clinical Evaluator (ACE) scores learner actions against competency-based rubrics educators design. Lumeto’s Artificial Clinical Facilitator (ACF), set up by faculty, runs in VR and screen simulations. Lumeto still treats live, expert-led debriefing as the best option and uses AI for sessions faculty cannot attend.

Is screen-based access equal to the VR experience?

Ask if the same scenario, patient and scoring run on a headset and a laptop. Some platforms are VR-only or sell a separate screen product, so data may not compare across learners. Screen access helps sites with few headsets, motion-sick learners, and remote or night-shift staff.

What should you ask about bias and patient representation?

Large language models may carry biases that cause stereotyped behaviour (Zeng et al.). Ask vendors:

  • How do you check AI responses for stereotypes?
  • What ages, skin tones, languages and body types does the library cover?
  • Can educators adjust a patient’s background and language?
  • Where is conversation data stored, and who can access it?

How do AI virtual patient platforms compare?

Lumeto makes InvolveXR, and these details come from public vendor pages as of October 2026. Check what matters to you in a demo.

PlatformHow learners talkPhysiologyFaculty authoringAI feedbackAccess
Body InteractOwn questions by voice or text, five languagesResponds to learner decisionsNot described; 1,500+ curriculum-aligned scenariosAI report after each caseWeb, iOS, Android and Windows
SimXNot described in detailConditions adapt to learner actionsEdit appearance, vitals, labs, dialogue and diagnosis liveTargeted feedbackVR; multiplayer in person or remote
Oxford Medical SimulationAI-powered patient and team interactionsNot described in detailOMS Create authoring toolsEach action ranked, tracked and explainedVR headsets or computer screens
vrClinicals for NursingLearners speak directly to the patientEvolving patient conditionsStandardised NLN scenariosAutomated feedback and scoringMeta Quest, Windows or Mac
VRpatientsAI-powered conversationsNot describedNo-code authoring toolAI grading for debriefsWeb, VR or mixed reality
InvolveXR (Lumeto)Open conversation, verbal and non-verbalVitals and states change; adjustable liveNo-code editing of patients, dialogue, medications, checklistsACE scores on educator rubrics; ACF debriefsVR headsets or screen

What belongs on your vendor demo scorecard?

Bring these tests to your next demo. A 24-second video walks through them.

Four tests to run in an AI virtual patient demo.
CriterionTest in the demoStrong result
ConversationAsk an unexpected open questionRelevant, in-character answer
AccuracyAsk about a symptom outside the caseAnswer fits the diagnosis
PhysiologyGive a wrong interventionVitals worsen plausibly
AuthoringHave faculty edit a patient's historyDone in minutes, no vendor help
ScoringAsk why a score was givenScore links to a specific action
DebriefRun the AI debriefPrompts reflection, cites session moments
AccessRun the scenario on a laptopSame patient and scoring as VR
EvidenceAsk for platform-specific studiesPublished studies, with how many learners took part

Frequently asked questions about AI virtual patient platforms

Which simulation platforms include AI-powered virtual patients?

They include Lumeto's InvolveXR, Body Interact, Oxford Medical Simulation, SimX, vrClinicals for Nursing and VRpatients. They differ in conversation, faculty authoring and access.

Which XR training platforms offer customisable virtual patients for clinicians?

Lumeto's InvolveXR lets educators edit patients, dialogue, vitals, medications and checklists without code. SimX, VRpatients and Oxford Medical Simulation (OMS Create) also offer authoring tools.

Are AI virtual patients accurate?

It depends on how the vendor constrains the AI, since the WHO warns these models can produce false or incomplete statements. Test accuracy in a demo and ask how responses are reviewed.

Can AI assess learner performance in a simulation?

Yes. Assessors such as Lumeto's ACE score learner actions against educator rubrics, and educators should be able to override scores.

Do AI virtual patients work without a VR headset?

Many do. InvolveXR, Oxford Medical Simulation, vrClinicals for Nursing and Body Interact offer screen-based access.

Lumeto Clinical Education Team

Lumeto

Lumeto's Clinical Education Team builds InvolveXR learning experiences with clinicians and educators, and reviews them for clinical accuracy.

Elevate Your Healthcare Program with AI-Enabled Immersive Training
InvolveXR delivers simulation of real procedures and patient interactions with lifelike scenarios enhanced by AI.