Real-world trial assesses safety of patient-facing AI in primary care
Researchers at Beth Israel Deaconess Medical Center tested a conversational AI system called AMIE with patients who had scheduled urgent primary care appointments. The study, published in The Lancet, is described as the first prospective real-world evaluation of patient-facing AI in primary care. It examined safety, acceptability, and potential effects on patient and physician experience under physician oversight.
Researchers partnered with Google to test AMIE, a medical chatbot patients reached from home via secure text chat after booking urgent primary care visits. It gathered symptoms and histories, offered possible diagnoses for patients to review with their doctor, and generated a clinician summary before the appointment.
Between April and November 2025, 114 patients enrolled; 98 completed both the chatbot exchange and their visit. A board-certified internist monitored every conversation live, able to intervene for safety or distress concerns. No safety stops occurred; supervisors logged one hallucination and added clinical clarification in five cases.
If such tools prove safe and acceptable, they could reshape how primary care absorbs rising demand and workforce strain, potentially extending clinician reach. Patients might gain earlier symptom review and better-prepared visits, while physicians could face altered workflows and new oversight duties. Trust concerns and the need for live human monitoring suggest adoption may stay cautious, and any benefits could depend on how reliably these systems perform beyond controlled trials.