Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
How to Live Better, Longer
A medical AI chatbot built by Google interviewed 98 patients about their symptoms before real doctor visits in Boston, and supervising physicians never had to stop a conversation for safety reasons. The results, published Oct. 8 in The Lancet, come from one of the first real-world tests of a patient-facing diagnostic AI inside a working clinic.
The study was led by researchers at Google and Beth Israel Deaconess Medical Center and took place at the hospital’s ambulatory primary care clinic. Patients booked for urgent visits chatted with the AI, called AMIE, from home before seeing their doctor.
Many people already type symptoms into chatbots before deciding whether to see a doctor. This study asked a narrower question: can such a system talk with real patients safely when doctors are watching?
Safety oversight was built into the design. According to a preprint version of the study, seven board-certified internal medicine physicians observed each text conversation in real time through a secure video link. Patients could not see the supervisors, and a supervisor was never the patient’s own primary care doctor.
The team set four rules in advance for stopping a conversation: immediate concern for harm to self or others, significant patient distress caused by the AI, potential clinical harm spotted by the supervisor, or a patient’s request to end the session. None of the 98 analyzed conversations met any of those criteria.
After each chat, the supervisor debriefed the patient and corrected any errors. The researchers reported one hallucination, meaning a statement the AI made up. Supervisors added clinical clarification in five conversations, and one physician rated an interaction as somewhat harmful because the patient may have become anxious after AMIE raised lymphoma as a possible diagnosis, according to Becker’s Hospital Review.
Dr. Adam Rodman, director of AI programs at BIDMC’s Carl J. Shapiro Institute for Research and Education, said that before they are judged as time-savers, “AI tools must first be shown to operate safely and acceptably in real care workflows.”
The study also measured how useful AMIE’s work was. The AI’s list of possible diagnoses included the doctor’s final diagnosis 90% of the time, according to Google’s summary of the study. In blinded comparisons, AMIE and primary care physicians performed similarly on diagnosis and management plans, though human doctors scored better on practical, cost-conscious care.
Physicians who reviewed AMIE’s summaries before a visit said the summaries helped them prepare in about 75% of those cases. In more than half, the doctors said the AI’s work may have changed their approach to care.
Patients were more divided. Their views of AI improved after the chat and stayed higher after the visit, and most said they would use AMIE again. Yet fewer than half felt the system would keep their information confidential or seemed honest and trustworthy.
Rodman said research suggests patients’ trust in medical AI comes from their relationships with their human providers. Dr. Marc Cohen, clinical chief of primary care at BIDMC and a co-senior author, said the team sees a future in which “the triad of patient, human clinician, and AI” work together to provide evidence-based care.
The study was designed to test feasibility and safety, not to prove that AI improves health. It had no comparison group, took place at a single Boston clinic, and included only patients who agreed to take part. Of 1,452 urgent visits scheduled during the study period, 140 eligible patients consented and 98 completed both the chat and their visit.
Several groups were excluded, including pregnant patients, people seeking mental health care, and anyone who needed emergency care. That means the study does not show how AMIE would handle a true emergency. The AI also had no access to medical records and could not perform a physical exam.
Participants skewed younger and were comfortable with technology, and people who know they are being observed may behave differently. The trial was registered on ClinicalTrials.gov as a prospective observational cohort study.
The study was funded by Alphabet, Google’s parent company, and Rodman was a visiting researcher at Google during part of the work, according to BIDMC. Google itself says larger clinical trials are needed to show whether tools like AMIE benefit patients.
AMIE is a research system and is not available to the public or approved for clinical use. The findings do not change how patients should seek care today.
People who use consumer chatbots to research symptoms should treat the answers as general information, not a diagnosis. Chest pain, trouble breathing, sudden weakness, confusion, or severe bleeding call for emergency care, not a chat window. When a clinic offers an AI intake tool, patients can ask who reviews the information, how it is stored, and whether a human clinician will check it before decisions are made.
For health systems facing long waits and physician shortages, AI interviews could eventually save time by gathering a history before the visit. Whether that helps patients, saves money, or simply adds another step has not been tested. MedicalDaily will follow any expansion of AMIE testing beyond this single clinic.
What is Google’s AMIE?
AMIE is a research AI chatbot from Google designed to interview patients about symptoms and suggest possible diagnoses. It is not available to the public.
What did the Lancet study test?
It tested whether AMIE could safely interview real patients before urgent primary care visits at Beth Israel Deaconess Medical Center in Boston while physicians watched live.
Were there any safety problems?
No conversation met the study’s predefined safety stop criteria. Researchers reported one hallucination and one interaction rated somewhat harmful because it may have caused anxiety.
How accurate were AMIE’s diagnoses?
Its list of possible diagnoses included the final diagnosis in 90% of cases, and its overall performance was similar to primary care doctors in blinded review.
Did patients trust the AI?
Views of AI improved, but fewer than half of patients felt AMIE would keep information confidential or seemed honest and trustworthy.
Who paid for the study?
The study was funded by Alphabet, Google’s parent company. A BIDMC study author was a visiting researcher at Google during part of the study.
Should patients rely on AI chatbots for diagnosis?
No. Chatbot answers are general information. Emergencies need immediate care, and any AI advice should be reviewed with a clinician.
Published by Medicaldaily.com
© 2026 Medical Daily Inc. All rights reserved
Medical Daily is for informational purposes and should not be considered medical advice, diagnosis or treatment recommendation. Read more.
