Google DeepMind's "AI co-clinician" beats GPT-5.4 in blind doctor tests but still trails experienced physicians

Google DeepMind tested an AI co-clinician in blind doctor evaluations, medication benchmarks, and telemedicine simulations.

The system beat GPT-5.4 on several measures and was often preferred by doctors, especially in simulated patient interactions.

Even so, experienced physicians still performed better overall, and the AI recorded at least one critical safety error.

The findings point to real promise for clinical support, but not enough reliability to replace physician judgment in care.