Artificial intelligence is increasingly being explored as a support tool in healthcare. A recent study evaluating Google’s conversational AI medical assistant, Articulate Medical Intelligence Explorer (AMIE) suggests that AI systems may assist physicians by collecting patient histories and suggesting possible diagnoses before consultations.
The research was published in March 2026 as a preprint on arXiv and reported by multiple medical news platforms. The study examined whether a large language model-based system could conduct structured medical interviews and generate diagnostic reasoning similar to clinicians in real clinical scenarios.
The findings indicate that AI could support clinical workflows by gathering patient information and helping physicians prepare for consultations.
The study was conducted in a real-world urgent care setting and involved 100 adult patients who had scheduled clinical visits. Before meeting their physicians, patients interacted with the AI assistant through a text-based chat interface.
During the conversation, the AI asked questions similar to those asked during a standard medical history interview. It collected details about symptoms, medical history, medications, and other relevant health information.
After each interaction, the system generated a structured clinical summary, which included a list of possible diagnoses and management suggestions.
Researchers then compared the AI’s diagnostic reasoning with that of primary care physicians who later examined the same patients.
Researchers reported that the AI system included the correct diagnosis in its list of possible conditions in about 90 percent of cases. When examining the top three diagnostic suggestions, the accuracy rate was approximately 75 percent.
Independent physicians reviewed both the AI-generated responses and the doctors’ clinical assessments without knowing which source produced them. Their analysis found no major difference in overall diagnostic reasoning quality between the AI system and physicians.
However, physicians performed better in certain areas, particularly when considering practical aspects of patient care such as treatment feasibility and cost considerations.
The AI assistant was not used as a replacement for doctors. Instead, it functioned as a pre-visit digital medical assistant.
Patients interacted with the system before their scheduled appointments, allowing the AI to collect medical information and prepare a summary for clinicians. Physicians could review these summaries prior to seeing the patient.
This process helped doctors gain an early understanding of patient concerns and symptoms before the consultation began.
Throughout the study, human safety monitors supervised all AI-patient interactions. Researchers reported that no conversation required intervention due to safety concerns.
Patient feedback collected during the study suggested that most participants were comfortable interacting with the AI assistant.
Many patients reported that the system was easy to use and helped them organize their symptoms before the consultation. The study also found that participants’ willingness to use AI tools in healthcare increased after the experience.
Researchers followed patient records for eight weeks after the clinic visit to confirm the final medical diagnosis and compare it with the AI’s earlier predictions.
The authors noted that the study represents an early feasibility assessment rather than evidence for routine clinical use. The research was conducted at a single clinical site with a limited number of participants, which restricts how widely the findings can be generalized.
Further studies involving larger patient populations and multiple healthcare centers will be necessary to evaluate safety, reliability, and long-term impact on patient outcomes.
Researchers also emphasized that AI systems should be used as clinical decision-support tools, rather than replacements for physicians, since medical diagnosis also relies on physical examination, clinical judgment, and contextual patient factors.
The findings suggest that conversational AI systems could assist healthcare providers by collecting medical histories, organizing patient information, and supporting diagnostic reasoning.
If validated through larger trials, such tools may help reduce administrative workload and improve preparation for clinical consultations.
However, researchers stress that careful evaluation, regulation, and clinical oversight will be essential before AI medical assistants can be widely integrated into patient care.
1. Brodeur, Peter, Jacob M. Koshy, Anil Palepu, Khaled Saab, Ava Homiar, Roma Ruparel, Charles Wu, et al. 2026. “A Prospective Clinical Feasibility Study of a Conversational Diagnostic AI in an Ambulatory Primary Care Clinic.” arXiv preprint arXiv:2603.08448. https://doi.org/10.48550/arXiv.2603.08448.
(Rh/SS)