Pharmaceutical ads may influence which drug AI recommends, new research finds

A study published this week in The Lancet Digital Health found that 12 consumer-facing large language models, including versions of Anthropic's Claude, Google's Gemini and OpenAI's GPT, recommended a treatment an average 47.6% of the time when shown an advertisement for it, versus 35% with no ad present, across 33 clinical scenarios. The models disclosed the ad's presence only about 30% of the time, and study author Dr. Girish Nadkarni of Mount Sinai's Icahn School of Medicine said, "An answer can be clinically reasonable and still raise an important question about why the AI arrived at that answer." The study did not examine whether the shifts affect real-world patient requests or prescriptions, and it noted that AI-mediated promotion "might not fit existing categories" of FDA advertising oversight.

Next time you ask AI https://www.fastcompany.com/section/artificial-intelligence for medical advice, you may want to consider whether it’s been influenced by Big Pharma. Prescription drug advertisements are everywhere, and may affect AI’s recommendations, according to a study <https://www.thelancet.com/journals/landig/article/PIIS2589-7500 26 00137-8/fulltext published in The Lancet Digital Health this week. The study authors conducted experiments with a dozen consumer-facing LLMs, including versions of Claude, Gemini, and GPT—made by Anthropic, Google, and OpenAI, respectively. The models were tested across 33 scenarios that each included a clinical case description and multiple treatment options. In some cases, the AI models were also given an ad for one of the treatment options. The ad’s content included information about the drug and its uses, but did not specifically direct the AI model to recommend the drug. When AI models were exposed to an ad, they recommended the treatment in question an average 47.6% of the time, compared to 35% when no ad was present, though the shifts varied between the 12 models. Although the models often changed their recommendation when an ad was present, in most cases, they still selected https://www.thelancet.com/cms/10.1016/j.landig.2026.101114/attachment/00195571-2984-424d-928d-e5e3c8c617ae/mmc1.pdf one of the appropriate options. The outputs only disclosed the ad’s presence about 30% of the time. “An answer can be clinically reasonable and still raise an important question about why the AI arrived at that answer,” study author Dr. Girish Nadkarni https://profiles.mountsinai.org/girish-n-nadkarni , a professor of medicine and system chair of the Windreich Department of Artificial Intelligence and Human Health at Mount Sinai’s Icahn School of Medicine, said in a statement https://www.mountsinai.org/about/newsroom/2026/drug-ads-can-influence-ai-treatment-recommendations-study-finds . “As AI becomes part of how people get health information, we need to be able to distinguish clinical guidance from commercial influence and be transparent about when advertising is part of the interaction.” Pharmaceutical companies spent more than $6 billion on TV advertising alone in 2024, according to https://www.iqvia.com/locations/united-states/blogs/2025/10/life-science-and-direct-to-consumer-television-advertising an IQVIA report, which said that direct-to-consumer advertising “directly influences patient requests and prescription rates, driving real action in healthcare.” All high-income countries, save for the U.S. and New Zealand, ban direct-to-consumer ads https://ebm.bmj.com/content/29/6/423 for prescription drugs. In the U.S., prescription drugs consume more than 30% of commercial time during major networks’ evening news programs, according to https://www.help.senate.gov/dem/newsroom/press/news-sanders-king-introduce-bill-to-ban-prescription-drug-ads a U.S. Senate Committee on Health Education, Labor and Pensions release from 2025. Last year, Ranking Member Senator Bernie Sanders introduced https://www.congress.gov/bill/119th-congress/senate-bill/2068/all-info the End Prescription Drug Ads Now Act, co-signed by six other senators, which would ban such advertising, including on social media. In practice, the Lancet study said, AI models could be exposed to drug ads “through user prompts, application-added context, retrieved online material, or inadvertent or malicious insertion,” and that, “unlike television or print advertising, which is subject to U.S. Food and Drug Administration oversight, AI-mediated promotion might not fit existing categories.” The study did not examine whether the effects measured reflect real-world impacts on patients’ requests or prescriptions. In a recent Doximity survey https://www.doximity.com/reports/doctor-driven-ai-report/2026 trust-and-caution of 3,379 U.S. physicians, 54% said they had moderate or high levels of trust in AI’s ability to recommend treatment or management. Three-quarters of respondents also said patients had become more likely to propose a diagnosis or treatment plan suggested by AI over the past six months. “Clinicians might defer to algorithmic recommendations even when they conflict with clinical judgement,” the study’s authors said; and an appropriate but ad-influenced suggestion “might provide little reason to question it.”