知Yizhi Daily
← Back to this edition

Conversations

Study Says Drug Ads Can Change AI Chatbots’ Medication Recommendations

Edition 2026-11-13 · 08:00 Beijing time

A study by the Icahn School of Medicine at Mount Sinai tested 12 AI models, 33 clinical scenarios, and 258,559 responses. When a clearly labeled drug advertisement appeared before the case, the advertised drug was selected 47.6% of the time, compared with 35% without the ad; in four scenarios, the related drug went from being selected 0% of the time to 100% of the time.

The study was conducted by a team at the Icahn School of Medicine at Mount Sinai, and the results were published in the November 11 online edition of The Lancet Digital Health. Researchers tested 12 AI models from major technology companies including Google, OpenAI, and Anthropic to see whether clearly labeled drug advertisements would influence the models’ treatment choices in clinician-written clinical cases.

The experimental design tried to isolate the advertising factor: the same clinical case was either provided to the model alone or preceded by a clearly labeled real drug advertisement, while patient information and available treatment options remained unchanged. A total of 258,559 responses across 33 clinical scenarios were analyzed. The results showed that with an advertisement, the advertised drug was selected in 47.6% of responses, compared with 35% without an advertisement. In four cases, the model shifted from never selecting the advertised drug to selecting it in every response after the ad appeared.

First author Mahmud Omar notes that the answers often look reasonable and the model may still be recommending appropriate treatment, but preferences among multiple reasonable options can shift because of advertising, making this influence easy to miss in routine evaluations. The study also examined whether models acknowledged the ad's existence or repeated its content: in a subset of free-text responses, only 29.9% explicitly mentioned the advertisement, while repetition of advertising information or claims was more frequent. Senior author Girish Nadkarni says that as AI becomes a source of health information, clinical guidance needs to be distinguished from commercial influence.

Original sources

Reddit Science ↗
Back to this editionOpen interactive reader