Artificial intelligence chatbots may encourage higher medication spending by favoring branded drugs, according to a study from Osaka Metropolitan University published in Nature Health. The research examined responses from five advanced models across thousands of simulated clinical scenarios.
The analysis covered Claude Sonnet 4.6, GPT-5.4, Gemini 3.1 Flash, Llama 4 and MegGemma. In 97.5% of cases, a single commercial brand dominated the algorithms’ top recommendation, leaving out cheaper generic alternatives described as identical. The AI suggestions matched doctors’ real-world prescriptions in 29.7% of responses.
According to lead researcher Daiju Ueda, the recommendations mainly targeted medicines with strong online visibility, including products that appeared prominently in Google searches or had dedicated Wikipedia pages. This suggests that the systems’ choices were influenced by commercial exposure rather than clinical efficiency alone.
The study also found that AI selected the most expensive options available in 74.5% of responses when compared with typical professional recommendations. If those suggestions were transferred directly into medical prescriptions, annual acquisition costs would rise by approximately $55.6 million, or close to €50 million, representing a 1,019% increase in prescription purchasing costs.
The researchers said the findings underline the need to assess healthcare AI for bias in product selection and in its handling of cost and regulatory information. Ueda also emphasized that combining AI systems with trusted medication databases and human supervision could support more appropriate treatment choices.
Sources: La Vanguardia.