Google Research and Google DeepMind have demonstrated real-time clinical video consultation capabilities in AMIE, their research medical AI system. The project marks a first-of-its-kind demonstration of expert-level artificial intelligence capabilities within a clinical video environment. Google announced the research findings on August 11, 2026.
Developers constructed AMIE on foundation models Gemini and Project Astra, using a multi-agent architecture. In traditional clinical practice, a doctor's examination relies heavily on non-verbal observation alongside conversation. A physician regularly notices a cough, observes a patient's gait, or registers visible signs of discomfort during an office visit. AMIE replicates this multimodal process in real time by interpreting visual and auditory cues, guiding virtual physical exams, and reasoning diagnostically throughout the video consultation.
The research team evaluated the software using a randomized study design featuring simulated consultations. This trial involved patient actors working alongside a group of primary care physicians to benchmark performance. Clinical evaluators judged AMIE favorably across core clinical competencies tested during the trial. These competencies encompassed history-taking thoroughness, diagnostic accuracy, management appropriateness, and communication quality. Patient actors who participated in the study also reported preferring the video consultation experience over text chat interactions.
Clinical Deployment Requirements
Google emphasized that AMIE remains an experimental research system rather than a commercial product. Google did not specify a release date or clinical deployment timeline, maintaining that substantial additional research is needed before responsible real-world clinical deployment can take place.
