Researchers at Boston-based Beth Israel Deaconess Medical Center have tested a patient-facing conversational AI system with people scheduled for urgent primary care appointments, in what they say is the first prospective real-world study of its kind, according to a study in The Lancet.
Here are six things to know:
- The researchers worked with Google to test AMIE, the Articulate Medical Intelligence Explorer, a chatbot patients used from home up to five days before a single-complaint urgent visit. It asked about symptoms and medical history, suggested possible diagnoses for patients to discuss with their physician and generated a summary for clinicians.
- From April to November 2025, 114 patients enrolled, and 98 completed both the AI interaction and their appointment. A physician monitored every conversation in real time and could step in if a patient appeared at risk, was distressed, asked to stop or raised a safety concern.
- None of the conversations required a safety stop under prespecified criteria. Supervisors noted one hallucination and added clinical information in five interactions.
- Clinical evaluators rated the conversations favorably in 87% to 100% of cases across 17 criteria, and patients did so in 48% to 96% across 16 criteria. Patients’ attitudes toward AI improved after using the tool and stayed higher after their visit, though they remained concerned about confidentiality and the chatbot’s honesty.
- Physicians completed surveys in 60 cases, including 44 in which they reviewed AMIE’s transcript or summary before the visit. They found it helpful for visit preparation in 33 of those 44 and said it might have changed their approach in 25. In one case, a physician reported the interaction as somewhat harmful because a patient may have felt anxious after AMIE listed lymphoma as a possible diagnosis.
- The authors said the single-site, single-arm study was designed to assess feasibility, not whether AI improves health outcomes.
