Quick Answer
AMIE Video is a Google Research system for real-time audio-visual clinical conversations. The August 2026 preliminary study tested it in a randomized, simulated Objective Structured Clinical Examination (OSCE), not in routine patient care. It is an important research result, but it does not establish public availability, regulatory clearance, safe autonomous diagnosis, or clinical utility with real patients.
For a health concern, use an appropriate licensed clinician or emergency service. This page interprets a research study; it is not medical advice.
What Google Announced
Google describes AMIE Video as a Gemini- and Project Astra-based system that can hold a spoken video conversation while processing clinical audio-visual cues and maintaining a diagnostic and management plan. Instead of asking one model to do every task in sequence, the system separates work among three asynchronously cooperating agents:
- the Talker maintains the patient-facing conversation;
- the Planner updates clinical goals, the differential diagnosis, and management reasoning;
- the Perception agent reviews the audio and video stream for relevant cues.
Read the AMIE multi-agent architecture guide for the component boundaries and latency tradeoffs.
What the Study Actually Tested
| Study element | Reported design |
|---|---|
| Setting | Simulated real-time video or text consultations |
| Scenarios | 100 across five body-system groups |
| Consultations | 300 across three study arms |
| Patient participants | 15 trained professional patient actors |
| Comparator clinicians | 10 board-certified primary care physicians using video |
| Independent evaluators | 20 experienced primary care physicians |
| Arms | AMIE Video, AMIE Text, and PCP Video |
The clinical evaluators used general competency scales and case-specific criteria. Patient actors also rated their experience. These are meaningful evaluation perspectives, but they are not the same as patient outcomes, diagnostic safety in an uncontrolled population, or longitudinal care.
What the Findings Support
Within this simulated OSCE, clinical evaluators rated AMIE Video on par with PCP comparators on several core competencies and found advantages on some physical-observation and guided-examination measures. The paper also reports that the video interface was preferred over the text interface on several experience measures.
The result should be read at the level tested: one research configuration, one scenario set, professional actors, defined rubrics, and a controlled interface. “Expert-level” in the paper title is not a general authorization to treat the system as an expert clinician in uncontrolled care.
What the Findings Do Not Support
The study does not establish:
- effectiveness or safety for real patients with unpredictable presentations;
- performance for conditions that professional actors could not authentically portray;
- reliable detection of every subtle, anatomical, fast-moving, or intermittent sign;
- public product access, supported countries, pricing, or a patient-facing care service;
- replacement of clinician judgment, examination, testing, escalation, or follow-up;
- regulatory approval or a cleared medical-device use.
Google reports occasional perception and reasoning errors plus intermittent technical issues. The paper also identifies limits involving fine anatomical precision, subtle affective cues, and high-frequency movements. See the AMIE Video vs Text OSCE guide for a comparison that keeps these limitations attached to the results.
How to Read Future AMIE Claims
Before treating a later announcement as evidence of deployment readiness, check five separate questions:
- Was the evaluation simulated, prospective, retrospective, or conducted in real care?
- Were participants professional actors, selected volunteers, or representative patients?
- Was the outcome a rubric score, diagnostic accuracy measure, workflow measure, safety endpoint, or patient outcome?
- Did an independent party evaluate the system, and was the complete protocol available?
- Does the exact product have a documented intended use, human-oversight model, supported setting, and regulatory status?
A favorable controlled study can justify more research. It cannot answer those deployment questions by itself.
Frequently Asked Questions
What is AMIE Video?
AMIE Video is Google Research’s experimental audio-visual configuration of the Articulate Medical Intelligence Explorer. It coordinates Talker, Planner, and Perception agents during a real-time simulated clinical video consultation.
Can patients use AMIE Video?
The cited sources describe a research system and do not establish public patient access, a clinical product launch, or regulatory clearance. Do not use the study page as a care pathway.
Did AMIE Video prove clinical utility with real patients?
No. The randomized OSCE used professional patient actors in simulated scenarios. Google and the paper say real-patient studies are needed before conclusions about real-world clinical utility can be drawn.
Primary Sources
- Google Research: Advancing AMIE towards expert-level audio-visual clinical consultations, published August 11, 2026.
- Nagda et al.: Towards Expert-level Medical AI for Real-time Video Consultations, arXiv v1 dated August 10, 2026.