Ninety-eight patients completed conversations with AMIE, Google’s research diagnostic AI chatbot, ahead of urgent primary care appointments, and zero conversation safety stops were required under prespecified criteria, according to a study published online in The Lancet on 8 October 2026.
The prospective, single-centre feasibility study was run with Beth Israel Deaconess Medical Center and funded by Alphabet. Google says it is its first-ever publication in the main journal of The Lancet.
Business Pill · A SAMPLE IS NOT EVERYONE
A short explainer of panel data: a sample is not everyone. It teaches the general idea only and says nothing about any company or study in this story.
The key insight: As we read it, the study’s headline numbers sit on small denominators. The zero-stops result covers 98 supervised conversations, and the 75% figure Google cites is 33 of the 44 cases in which doctors reviewed the transcript.
What the Study Measured
According to the abstract, English-speaking adults interacted with AMIE up to five days before a single-complaint urgent primary care appointment, while physician safety supervisors monitored every interaction and were trained to intervene on predefined safety criteria. Transcripts and summaries were shared with the patients’ doctors before the visit.
From April to November 2025, 114 patients were enrolled and 98 completed both the AMIE interaction and the appointment, the abstract says. Safety supervisors noted one hallucination and added clinical information in five interactions.
Clinical evaluators rated AMIE’s conversations favourably in 87–100% of cases across 17 criteria, and patients in 48–96% across 16 criteria, per the abstract. Patient attitudes towards AI improved after the interaction and remained elevated after the doctor’s visit.


What Doctors Reported
Primary care physicians completed post-surveys in 60 of 98 cases, including 44 in which they reviewed the AMIE transcript before the visit. They found AMIE helpful for visit preparation in 33 of those 44 cases and reported that it might have changed their behaviour in 25 of 44, the abstract says.
Google’s post summarises the first figure as helping clinicians prepare in 75% of cases; on our arithmetic, 33 of 44 is 75%. Google also says AMIE’s differential diagnoses matched the doctors’ final diagnoses 90% of the time, a figure that does not appear in the abstract we read.
What the Authors Conclude
“Although further research is needed, this study shows the initial feasibility of conversational AI in a real-world setting,” the authors write in the abstract’s interpretation.
Google’s post says larger clinical trials are needed to assess patient-facing AI at scale. It describes the work as part of its efforts in AI for science and health.
The Structural Read
Supervision was part of the design. Physician safety supervisors monitored every interaction and could intervene on predefined criteria, the abstract says.
Doctor feedback came from a subset. Doctors completed post-surveys in 60 of 98 cases, and reviewed the transcript beforehand in 44.
Patients moved too. The abstract reports that patient attitudes towards AI improved after the interaction and remained elevated after the visit.
Google, 8 October 2026
“While larger clinical trials are needed to assess patient-facing AI at scale, the findings suggest the potential for AI to enhance the patient-physician relationship and ease the strain on healthcare workers.”
Three Implications
A FEASIBILITY STEP The authors say further research is needed, and Google says larger clinical trials are needed.
TRANSCRIPTS BEFORE THE VISIT AMIE transcripts and summaries were shared with doctors before appointments, per the abstract.
FUNDED BY ALPHABET The abstract lists Alphabet as the study’s funder.
The Business Engineer Lens
This story maps onto the Business Engineer framework The Product Overhang Doctrine.
The framework’s starting point: “In any product cycle, there are two curves to follow. The first is what your underlying technology can do. The second is what your users can actually do with it.”
As we read it, this study measures the second curve: what patients and doctors did with the system in a real clinic, under supervision, rather than what the model scores on a benchmark.
What Is Not Established
We read the Lancet abstract and Google’s post; the full paper is behind a paywall and we did not read it, so we could not check the 90% diagnosis figure or the methods in detail. This was a single-centre, single-arm study without a comparison group. We did not contact Google, BIDMC or the authors.
The Bottom Line
In a Lancet feasibility study funded by Alphabet, 98 patients completed supervised conversations with Google’s AMIE before urgent primary care visits with zero safety stops required, while supervisors noted one hallucination; doctors who reviewed the transcripts found it helpful for preparation in 33 of 44 cases.
94,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. We read the Lancet abstract and Google’s post of 8 October 2026; the full paper is paywalled and we did not read it. The 75% check is our arithmetic. This is not medical advice. We did not contact Google, BIDMC or the authors. Nothing here is a forecast, and nothing here is financial or investment advice.
Sources: The Lancet: Conversational diagnostic artificial intelligence in ambulatory primary care: a prospective feasibility study (8 Oct 2026, abstract) · Google: Study in The Lancet suggests AI could improve patient-physician relationships (8 Oct 2026)








