New research out from Mount Sinai has found that a dozen AI models it tested from major tech companies were more likely to recommend an advertised drug when it was promoted alongside a clinician’s question about treatment.
Researchers from the Icahn School of Medicine at Mount Sinai tested 12 AI models from the likes of Google, OpenAI, Anthropic and others to see what happens when advertising for meds appears alongside a clinical question.
In their report, published Nov. 11 on The Lancet Digital Health, the team found that when the AI models were given a doctor-written clinical case, preceded by an ad for a drug, “the ad could shift which treatment the AI recommended, making it more likely to select the advertised drug,” according to a release on the study.
Across 33 tests, the promoted drug was selected in just under half (47.6%) of responses when an ad was present, compared with 35% without one.
“In four cases, models shifted from selecting the advertised drug zero percent of the time to selecting it 100 percent of the time when the ad was introduced,” the researchers found.
“What I find quite interesting, and even surprising, was how reasonable the answers often appeared,” said physician-scientist and first author Mahmud Omar, M.D., a lecturer in the Windreich Department of Artificial Intelligence and Human Health at the Icahn School of Medicine, in a statement.
“The model could recommend an appropriate treatment, but its preference among several reasonable options changed after seeing the ad," Omar continued. "Because the final answer could still appear medically sound, that influence could be easy to miss in typical research evaluations and in other patient and physician facing personal use cases.”
The researchers stressed that their findings do not mean pharma advertising “will necessarily cause AI models to recommend unsafe or inappropriate treatments.”
This is because the ads did not clearly increase recommendations for unsupported supplements or clearly suboptimal drugs. The researchers also did not check whether changes in AI recommendations ended up impacting real world elements such as prescriptions, patient decisions, costs or health outcomes.
However, the researchers added that their findings “may have implications for regulators” when it comes to pharma advertising. This could specifically apply to how promotional content is added into AI-generated health information.