There is a new study by Google DeepMind and Google Research that was published in Natura today that shows how an AI system can sometimes outperform board-certified primary care physicians in medical practice. There is no need to sound the alarm bells and start talking about jobs being at risk because I think this points to a future where doctors can focus a lot better on what matters the most – the patient in front of them.
Also read: India at G7 2026: Access to frontier AI models key to fight cyber threats
The basic goal of the research was to test the upgraded version of Google’s AMIE Model (Articulate Medical Intelligence Explorer) against professional primary care physicians, 21 to be exact. This was done across 100 different clinical scenarios that went over five medical specialties. AMIE was given the same clinical guidelines that the doctors had to follow. Once this was done, both were evaluated under the same parameters by specialist physicians and patient actors. This is where it got interesting because AMIE wasn’t just competitive, it even ended up pulling ahead in some areas.
The first is preciseness. Where a human doctor might write “prescribe an antibiotic” and leave the specifics to be sorted later, AMIE consistently named the exact drug, the dose, the duration, and the administration route – across every single visit. Specialists rated AMIE’s treatment recommendations as sufficiently precise in 96% of cases in the first visit, compared to 62% for human physicians. That gap didn’t close over subsequent visits. It widened.
The second is guideline alignment. Clinical practice guidelines exist for a reason – they encode the best available evidence for treating a condition. Yet following them consistently is genuinely hard for a doctor managing fifteen patients in a morning, relying on memory and instinct. AMIE cited relevant clinical guidelines explicitly in 100% of cases. Human physicians managed between 86 and 92% depending on the visit. Again, AMIE improved or held steady over time. The human doctors did not.
Also read: Snap Specs vs Meta AI Glasses vs Android XR: Which wearable Is the best
Here is where the narrative shifts from concerning to quietly optimistic. Precision in prescriptions and strict guideline adherence are not the parts of medicine that patients remember or value most. Nobody walks out of a consultation moved by the fact that their doctor correctly cited NICE guidance NG28. What patients remember is whether their doctor listened. Whether they felt understood. Whether their concerns were taken seriously and addressed with empathy and clarity. These are the things that human doctors do and in this study, when patient actors rated both AMIE and physicians on relationship-building, communication, and empathy, the results were notably closer.
Medicine has always had a finite resource problem. A good doctor’s most valuable asset is time, and the current reality of primary care – globally stretched, increasingly fragmented, burdened by administrative load and rising patient complexity – is that there is never enough of it. When an AI can handle the cognitively demanding but systematisable work of pulling the right guideline, cross-referencing drug formularies, and generating a precise, citation-backed management plan, it frees up the clinician to do the work that cannot be automated: evaluating how a patient is actually responding to treatment, reassessing their overall health trajectory, and building the kind of trust that makes patients follow through on care plans in the first place.
The researchers themselves are careful not to overstate what this means. AMIE is not ready for deployment. The study used patient actors, compressed timelines, and a text-only interface. Significant further research – including real-world prospective trials – is needed before any of this reaches a clinic.
But the direction of travel is becoming harder to ignore. AI handling the procedural precision of medicine, while doctors reclaim the human core of it, is not a dystopia. It might be exactly the division of labour that an overstretched global health system needs.
Also read: Android 17 is bringing foldable specific updates to the OS, here’s why it matters