GenAI-Supported Virtual Patients in Health Care Education: Systematic Review

In plain English

A systematic review of 15 studies covering 645 learners in nursing, medicine, pharmacy, radiography and first-responder training found that generative-AI virtual patients are feasible and well received, and generally beat control conditions on decision-making, history-taking and communication. Only six of the 15 used 3D avatars; most were text-based, single-session and brief, and 13 ran on OpenAI GPT models. The authors rate most studies as moderate risk of bias — small samples, few control groups, unvalidated self-report measures — and warn that novelty enthusiasm may be inflating the positive results.