AI models skew drug recommendations toward products with strong online presence
Start here
Researchers from Osaka Metropolitan University and MD Anderson Cancer Center tested five large language models on drug selection. For each active ingredient they examined, an average of 5.5 FDA-approved products were available, yet a single product accounted for 97.5% of first-choice recommendations on average — favoring drugs with a prominent online presence. In simulations, following these AI recommendations could raise annual acquisition costs by $55.6 million. The finding raises concerns for clinicians and health systems that may integrate LLM-based prescribing or formulary support, since brand concentration in model outputs could distort procurement and spending. The work has been reported as of late October 2026; further validation, peer-review status, and any policy or vendor response remain open questions.
Written by a model from the reports · updated 3 hours ago
Coverage timeline
Follow the coverage to see the different angles.
- medicalxpress.comPickAI's preference for drugs with a strong online presence may cost millions
A team from Osaka Metropolitan University and MD Anderson Cancer Center tested five LLMs. Although an average of 5.5 FDA-approved products were available for each specified active ingredient, a single product accounted for an average of 97.5% of first-choice recommendations. Under simulated prescribing, annual acquisition costs could rise by $55.6 million.
Heat trend for this story
Heat now 5·Peak in the comparable set 10 (Oct 27 19:00)·Change over 24h in the comparable set –·Participants now 1
Participants counted above, heat below. The trend compares only the participants observed in full throughout, so its set can be smaller than the heat figure itself. Move the pointer or click the chart for an hour's reading; the left and right arrow keys work too.