AI Chatbots Favor One Brand-Name Drug in Study

Five major AI models recommended a single brand-name product first in nearly all cases, researchers report in Nature Health.

Updated on Nov. 10, 2026 in Artificial Intelligence

Isometric editorial illustration of identical pale-blue capsule shapes in a branching arrangement, with one oxblood capsule isolated at the front.
Five major AI chatbots recommended one brand-name drug first in 97.5% of cases, a bias researchers estimate could add $55.6 million in annual costs for 51 drug ingredients. AI Illustration. Upload story photo >

Live Poll

Would you trust an AI chatbot's recommendation when choosing a prescription medication?

A joint study by Osaka Public University and the University of Texas MD Anderson Cancer Center found that five AI chatbots recommended one brand-name drug first in 97.5% of cases. Researchers estimate that following the AI recommendations would raise annual costs for 51 drug ingredients by $55.6 million.

Why it matters

The findings suggest patients and clinicians relying on AI for medication guidance could face higher drug costs without clearly understood reasons for the recommendations. The researchers note the bias may be linked to pharmaceutical companies' online marketing strategies, including Generative Search Optimization (GEO).

The analysis tested GPT-5.4, Claude Sonnet 4.6, Gemini 3.1 Flash, Llama4, and MedGemma, finding the most-favored product among 37 FDA-approved ibuprofen-based medications. Estimated cost increases would raise prices to 11.2 times the original, though the mechanism behind the bias is not yet established.

The players

Osaka Public University

A public university in Osaka, Japan, whose research team led the analysis of AI medication recommendations.

University of Texas MD Anderson Cancer Center

A cancer research and treatment center in Houston, Texas, that co-authored the study with the Osaka team.

The details

Researchers specified diseases and active ingredients and asked each AI to recommend up to three FDA-approved medications. Products frequently appearing in Google search results or with dedicated Wikipedia pages were more likely to be recommended, while recently approved or commonly prescribed medications received fewer recommendations.

Timeline

  1. Study published in Nature Health on November 10, 2026.

Big Picture

The findings extend concerns about Generative Search Optimization (GEO), a marketing practice aimed at making brands more visible to AI systems. If the bias stems from such strategies, it raises questions about how commercial influence reaches AI-generated health advice.

Patients who ask AI tools about medications may be steered toward a single brand-name product rather than cheaper equivalents with the same active ingredient. Checking recommendations against the full range of FDA-approved options and asking a pharmacist about generic alternatives can help control costs.

The takeaway

AI medication suggestions can carry hidden commercial bias, so cross-check any recommendation against FDA-approved alternatives and ask a pharmacist before switching or buying.

Further reading

Explore more on Artificial Intelligence to see how these systems shape advice across fields.

Live Poll

Would you trust an AI chatbot's recommendation when choosing a prescription medication?