AMIE Tests Real-Time Video Medical Consultations

Google Research is testing AMIE, a medical AI system capable of conducting clinical video consultations and analyzing what the patient says and shows in real time. The system can listen to symptoms, observe visible signs and guide a virtual physical examination.

The demonstration aims to bring medical AI closer to the kind of consultation you have with a healthcare professional. Instead of simply reading a message, AMIE can detect a cough, notice the way you walk or identify visible signs of discomfort while talking with you.

How AMIE Works

AMIE is built on Gemini and Project Astra, Google's platform for systems capable of interpreting information from their surroundings in real time. Its architecture uses multiple AI agents, meaning specialized components that work together to collect data, reason about it and respond during the consultation.

In practice, the system can:

• Ask questions about medical history and symptoms.
• Interpret visual and auditory signals.
• Guide certain physical examinations remotely.
• Suggest diagnostic reasoning and management options in real time.
• Explain its responses and communicate with the patient.

What the Study Showed

Google evaluated AMIE in simulated consultations with actors portraying patients and compared its performance with that of a group of primary care physicians. Clinical evaluators assessed the system on aspects such as thoroughness in gathering information, diagnostic accuracy, appropriateness of management and quality of communication.

In those evaluations, AMIE received favorable ratings across the clinical competencies assessed. The actors portraying patients also preferred the video experience over a conversation conducted only by text.

That does not mean AMIE can already replace a doctor or that its results are ready for use in hospitals or real-world consultations. The study took place in a simulated environment, and Google itself says more research is needed before responsible clinical deployment can be considered.

What This Changes for You

The main difference from a conventional medical chatbot is that video adds information text cannot capture. An AI could ask you to walk a few steps, observe a visible area or listen to certain sounds while deciding which questions to ask next.

This could be useful for initial medical guidance, especially when you do not have immediate access to a professional. But it also increases the risks: misinterpreting an image, gesture or symptom could lead to an inappropriate recommendation. That is why medical supervision, data privacy and validation with real patients will be just as important as the system's technical capabilities.

AMIE shows where AI in healthcare is heading: from answering written questions to observing, listening and following a complete clinical conversation. What remains to be established is whether it can maintain its accuracy and safety outside a controlled demonstration, with real patients and much more unpredictable medical situations.