Pharmaceutical advertisements can influence which medications AI models recommend, even when multiple treatments are clinically appropriate, according to a new study from researchers at the Icahn School of Medicine at Mount Sinai in New York City.
The findings, published Nov. 11 in The Lancet Digital Health, raise questions about how commercial advertising could influence AI-generated health information as more patients and clinicians turn to the technology for medical guidance.
Researchers tested 12 AI models from major technology companies, including Google, OpenAI and Anthropic, using 33 clinical scenarios. Each model received physician-written clinical cases, either alone or preceded by a clearly labeled pharmaceutical advertisement.
Here are six things to know from the study:
- Across 258,559 AI responses, the models selected the advertised drug 47.6% of the time when an advertisement was present, compared with 35% when no advertisement was included.
- In four scenarios, the models shifted from never selecting the advertised medication to recommending it in every response after exposure to the advertisement.
- Researchers also examined whether AI models acknowledged advertisements or incorporated their claims into responses. In a subset of free-text responses, models explicitly mentioned the advertisements only 29.9% of the time, despite more frequently repeating information or claims from the promotional materials.
- Researchers cautioned that the findings do not establish that pharmaceutical advertisements cause AI models to recommend unsafe or inappropriate treatments. The study did not find a clear increase in recommendations for unsupported supplements or clearly inferior medications, nor did it assess whether advertising-influenced recommendations affected prescriptions, patient decisions, costs or health outcomes.
- The researchers recommended that AI developers and health systems test whether exposure to promotional content changes treatment recommendations, disclose when advertising is included in AI interactions and separate promotional materials from clinical advice.
- The findings could also inform regulatory discussions about how pharmaceutical advertising rules apply to AI-generated health information.
