Background
Simulated patient encounters are essential for developing communication and clinical reasoning skills in medical students, but logistical challenges like preceptor shortages limit access. Generative AI (GenAI), specifically Large-Language-Model-based AI Patient Actors, offers scalable and flexible solutions. This study evaluates the integration of an openly available AI Patient Actor app (https://ai.dartmouth.edu/patient-actor) into a first-year clinical skills course.

Methods
Ninety-four students in the "On Doctoring" course at Geisel School of Medicine at Dartmouth completed at least three virtual patient interviews over 12 weeks. The AI simulations provided opportunities to practice clinical interviewing and diagnostic reasoning with immediate, personalized feedback based on a standardized rubric. A mixed-methods approach included surveys assessing student comfort with GenAI, perceived utility, and overall experience; sentiment and thematic analyses of open-ended comments; and a comparison of assignment submission timelines with cohorts using traditional methods.

Findings
The AI Patient Actor reduced late submissions by 75.9%, improving efficiency and flexibility. Post-training, student comfort with GenAI increased significantly (76% vs. 61%, p < 0.001). Survey results showed 77% rated their experience as “Good” or better. Positive feedback highlighted opportunities for low-pressure practice, iterative feedback, and skill development, while concerns focused on the realism of AI interactions and limitations in replacing human encounters.

Discussion
AI Patient Actors address logistical barriers in medical education by offering flexible, accessible training with immediate feedback. However, concerns about realism highlight that AI is no substitute for human-standardized patients and should complement, not replace, traditional training. Future research should explore strategies to enhance the authenticity and long-term educational value of AI-mediated simulations.