Did AI search recommendations lead to illegal online pharmacies?

Details

The online pharmacy market is growing, and legitimate online pharmacies can offer convenience and accessibility. But growing demand has also attracted malicious actors: illegal vendors use deceptive techniques to rank higher in search results and pose public health risks by dispensing substandard or falsified medicines.

Search engine providers have begun integrating generative AI into their interfaces, potentially offering more personalized results through a user-friendly experience. If integrated improperly, however, these technologies could inadvertently direct users to illegal vendors and worsen existing risks. The role of generative AI integration in reshaping search results, particularly for online pharmacies, had not yet been studied. The researchers aimed to identify, measure the prevalence of, and characterize illegal online pharmacy recommendations in AI-generated search results and recommendations.

They conducted a comparative assessment of recommendations from Google’s Search Generative Experience (SGE) and Microsoft Bing Chat, focusing on popular, well-known medicines across multiple therapeutic categories, including controlled substances. They examined websites individually for legitimacy and cross-referenced the National Association of Boards of Pharmacy and LegitScript databases to identify known illegal vendors.

Of 262 recommended websites, 47.33% (124/262) belonged to active online pharmacies, and 31.29% (82/262) led to legitimate ones. Illegal vendors accounted for 19.04% (24/126) of Bing Chat recommendations and 13.23% (18/136) of Google SGE recommendations, including recommendations for controlled substances. The proportion of illegal pharmacies varied by drug and search engine. The distribution of illegal websites differed significantly between the search engines.

For prescription medicines, links to illegal online pharmacies were significantly more prevalent in Bing Chat (21/86, 24%) than in Google SGE (6/92, 6%; P=.001). For controlled substances, Google suggestions led to a significantly higher number of rogue sellers (12/44, 27%; P=.02) than Bing suggestions (3/40, 7%).

The researchers concluded that generative AI integration into search engines offers promise but also poses significant risks. They described this as the first study to shed light on vulnerabilities in these platforms and the potential public health implications of inadvertently promoting illegal pharmacies. The concerning proportion of recommendations leading to illegal pharmacies could potentially increase traffic to those sites and worsen existing public health risks. The researchers called for urgent oversight and safeguards in generative search, including actively guiding users to verified pharmacies, prioritizing legitimate sources, and excluding illegal vendors from recommendations.