Google's AMIE gets its first prospective real-clinic study, published in The Lancet: 98 patients chatted with it before urgent visits under live physician supervision, no conversation was stopped, and doctors found the summaries helpful in 75% of visits

Why it matters

【Key points】

On Oct 8 (US time), Google announced that its AMIE study with Beth Israel Deaconess Medical Center (BIDMC) was published in The Lancet. AMIE is Google's research conversational diagnostic AI, and this is its first prospective study in a real clinic: 98 patients at BIDMC's ambulatory primary care clinic chatted with AMIE before urgent care visits; AMIE took their history and wrote a summary for the doctor.

Google's reported results: supervising physicians monitored every conversation live and none had to be interrupted under predefined safety criteria; clinicians said the AI summaries helped them prepare in 75% of visits and influenced their approach in more than half; AMIE's differential diagnoses matched the doctors' final diagnoses 90% of the time. Google says larger trials are still needed.

【Verifiable】

Google blog: https://blog.google/innovation-and-ai/technology/health/amie-clinical-study-lancet/

The Lancet: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(26)01535-7/abstract

Sundar Pichai's post: https://x.com/sundarpichai/status/2108329198955958339 (about 970 likes, 150k views)

Heat: about 180 related posts on X in 24 hours, steady at 10-25 an hour since release.

【Our read】

1. The point is not the 90% figure but the change in evaluation. Most medical LLM results come from retrospective data or simulated patients; this one tests real patients in a real workflow.

2. Mind the limits: 98 patients, one clinic, live physician supervision throughout, and AMIE only did pre-visit history-taking and summaries, no diagnosis or prescribing. "No stops" shows safety with a human watching, not without one. A 90% match on differentials is not the same as getting it right first time.

3. Next, watch for controlled multi-site trials, supervision relaxed from live monitoring to spot checks, and whether AMIE moves from research system to a Google product. If all three happen, patient-facing AI intake stops being a demo.