Google AMIE primary care feasibility study published in The Lancet – Unite.AI

A prospective clinical study conducted by researchers at Google and Beth Israel Deaconess Medical Center evaluating Google’s AMIE conversational diagnostic AI in a real-world urgent care clinic was published in The Lancet on October 8, 2026, according to Google’s announcement. In the study, 98 patients consulted the AMIE chatbot before urgent care visits while supervising physicians monitored the interactions in real time and none of the conversations had to be interrupted based on predefined safety criteria.

Google described the paper as its first-ever publication in The Lancet’s flagship journal. The company said that larger clinical trials are needed to evaluate patient-facing AI at scale, and that the findings suggest the potential of AI to improve the patient-doctor relationship and relieve strain on healthcare workers.

Study design and safety supervision

The system, formally called Articulate Medical Intelligence Explorer (AMIE), is an AI-powered medical chatbot that patients use from home after scheduling an urgent primary care appointment with a BIDMC doctor. According to reporting from Google’s research team, 100 adult patients completed a pre-visit interaction with AMIE and 98 showed up for their scheduled appointment. Participants, who were booked for new, non-urgent, episodic complaints, interacted with the system through a secure text chat interface up to five days before their visit. The system asked about symptoms, collected medical histories, presented potential diagnoses for patients to discuss with their doctor, and produced a summary that the doctor could review in advance. The prospective, single-arm, single-center study was pre-registered on ClinicalTrials.gov and conducted under institutional review board-approved protocols, with patients assured that their decision to participate would not affect their care.

Participants were younger than the clinic’s overall urgent care population: Of a total of 1,452 urgent care visits during the study period, more than half involved patients over the age of 60, while the sample’s female and white population trends were consistent with the clinic’s population, the Google Research account says.

BIDMC reported that the study ran from April to November 2025 and enrolled 114 patients, and that each conversation was monitored in real time by a board-certified internal medicine physician via live video call with screen sharing. Supervisors were trained to intervene based on four pre-specified criteria: immediate concern for harm to self or others, significant emotional distress related to interaction with the AI, potential for clinical harm identified by the supervisor, or an explicit request from the patient to end the session. In all 98 completed meetings, no conversations required a safety stop; According to the BIDMC release dated October 9, 2026, supervising physicians identified a hallucination and provided further clinical clarification in five cases. “This study helps establish the basic characteristics of such conversations in the real world,” co-first author Peter Brodeur, a cardiovascular medicine researcher at BIDMC, said in the release. BIDMC said it believes the study is the first prospective, real-world study of a patient-facing conversational AI system in primary care.

Results of diagnostic reasoning

The AMIE differential diagnosis included the final diagnosis, established through chart review eight weeks after each encounter, among the top seven possibilities in 90% of cases, with 75% accuracy among the top three, and identified the final diagnosis as the single most likely possibility in 56% of cases, the research team reported. Accuracy remained high for the subgroup of 46 patients whose final diagnosis was confirmed by a diagnostic test such as a laboratory, microbiologic, pathologic, or imaging result.

In a blinded, randomized comparison, groups of three clinical raters per case evaluated the differential diagnoses and management plans of the AMIE and primary care providers, finding similar overall quality for the differential diagnoses (p = 0.6) and for the appropriateness (p = 0.1) and safety (p = 1.0) of the management plans, according to the preprint study record. Clinicians outperformed AMIE in terms of convenience (p = 0.003) and cost-effectiveness (p = 0.004) of management plans. Google Research attributed this gap to the AMIE’s lack of access to the patient’s electronic medical record, its inability to perform a physical exam, and the lack of multimodal inputs such as the patient’s overall physical appearance.

Patient and doctor experience

Patients completed the General Attitudes Toward AI scale before the chat, after the chat, and after the provider visit; attitudes changed significantly more positively after interacting with AMIE (p < 0.001) and remained elevated after seeing the provider, the researchers reported. BIDMC said patients rated the system highly for listening, explaining information and helping them feel at ease, while concerns remained about trusting the confidentiality of information shared with the system and trusting the chatbot's honesty and reliability. Adam Rodman, director of AI programs at BIDMC's Carl J. Shapiro Institute for Research and Education, said future research will need to explore what interaction characteristics can build trust in patients and how AI interactions can improve the patient-doctor relationship.

Primary care providers reviewed an AI-generated transcript or summary before seeing patients in 44 cases; doctors said this helped them prepare for the visit in 75% of cases and may have influenced their clinical approach in 57%, BIDMC reported. In one case, a provider described the interaction as somewhat harmful, citing concern that a patient may have experienced anxiety after AMIE included lymphoma among his possible diagnoses. In qualitative interviews reported by Google Research, physicians said that AMIE shifted the dynamic of the visit from simply collecting data to verifying data, allowing for more collaborative conversations and shared decision-making.

Limitations, funding and prior disclosure

The research team describes the work as a single-center feasibility study without controlled comparisons that does not support claims of quantitative effectiveness compared to a basic workflow, and lists additional limitations including the text-only interface and the unexplored effects of health literacy, technology literacy, and familiarity with chatbots. Rodman and colleagues stressed that the study was designed to evaluate feasibility rather than to determine whether AI improves health outcomes.

BIDMC revealed that the study was funded by Alphabet and that Rodman worked as a visiting researcher at Google during part of the study. Brodeur and Jacob M. Koshy are co-first authors of the paper, Rodman is listed last among the authors, and Marc L. Cohen is the co-senior author.

The study’s findings were first disclosed publicly in March 2026: the preprint was submitted to arXiv on March 9, 2026, and the Google research team published its detailed report on March 11, 2026, updating that post on October 8, 2026 to note the Lancet publication. The research team said it plans to continue evaluating the utility and impact of such systems in future larger studies with controlled comparisons.