Key Points:
- Dr. Sara Grossman, PharmD, an associate professor of pharmacy practice at Long Island University (LIU), and her team evaluated ChatGPT’s accuracy in answering medication-related questions.
- The researchers presented ChatGPT with 39 medication-related questions, the answers to which were definitively known from published professional literature.
- Out of the 39 questions, ChatGPT provided accurate, complete, and relevant responses for only 10. For 11 questions, ChatGPT did not provide any response but gave general background information instead.
- When asked to provide references for its responses, ChatGPT did so for only eight out of the 28 questions it responded to. However, all these references were found to be fabricated, with seemingly real PubMed PMID numbers and citation formats, but the sources did not actually exist.