Google DeepMind's AI Co-Clinician Matches Doctors in Medical Consultations

Could AI become a doctor's most trusted ally? Google DeepMind's new system rivals human performance in medical consultations—with surprising results. Here's how it works.

, and Administrator

2026 May 2 . 8:19 PM

1 min read

In this picture we can see the doctor in white uniform explaining about the CT scan machine to 2... — In this picture we can see the doctor in white uniform explaining about the CT scan machine to 2 people wearing black coats. These two people are listening to the doctor.

Google DeepMind's AI Co-Clinician Matches Doctors in Medical Consultations

Google DeepMind is developing an AI system designed to work alongside doctors in patient care. Called the 'AI co-clinician', the tool aims to support medical teams rather than replace them. Early tests show it can match or even surpass human performance in certain areas of consultation quality. The AI co-clinician underwent rigorous evaluation across 140 aspects of medical consultation, spanning seven key domains. While experienced physicians still outperformed the system overall, the AI matched or exceeded general practitioners in 68 of the tested areas. In open-ended questions, it achieved a quality score of 95.0%, higher than OpenAI’s GPT-5.4-thinking-with-search, which scored 90.9%.

Researchers used the adapted NOHARM framework to check for errors of both commission and omission. The system also outperformed doctors—even those using reference materials—on the RxQA benchmark, scoring 73.3% accuracy. Unlike text-only models, Google DeepMind’s tool incorporates multimodal telemedicine, expanding its capabilities beyond simple chat interfaces. In a blind comparison, doctors consistently preferred the AI co-clinician’s responses over those from leading evidence-synthesis tools and an existing clinical AI system. DeepMind researcher Alan Karthikesalingam noted that while the technology shows clear potential, it remains in the early stages of development.

The AI co-clinician is designed to function as a supervised member of the medical team, assisting rather than replacing human judgement. Its performance in specific benchmarks suggests it could become a valuable tool in clinical settings. Further development and real-world testing will determine its long-term role in healthcare.

Latest

The image shows a whiteboard with the words "Students: What, Why, How, and How" written on it,...

All about learning.

How B1 Sentence Combining Worksheets Elevate Learners' Writing Skills

Struggling with choppy sentences? These targeted exercises transform basic ideas into polished, cohesive writing—one clause at a time. Teachers swear by them for building confidence at the B1 level.

, and Administrator

2026 May 2