A research team at Tokyo University of Science, in collaboration with researchers from Juntendo University and the Gunsei Okinawa Clinical Training Center, revealed that approximately 4% of clinical trainees nationwide who responded to the survey use generated AI as a search engine in their clinical practice, while only about 1% of users use it as a primary source of information for differential diagnosis.
As generative AI is rapidly spreading in clinical settings, the research team conducted a cross-sectional survey in January 2025 targeting Japanese clinical trainees (1st to 2nd year postgraduates) to investigate the actual use of generative AI as a search engine in clinical settings and its relationship to AI literacy (the ability to appropriately utilize AI while understanding its characteristics and limitations) and behavior in infectious disease treatment.
The results showed that of the 2850 people who responded about whether they use generative AI in their clinical work, 1124 (approximately 39%) used generative AI as a search engine, and of these users, 144 (13.0%) used generative AI as a primary source of information for differential diagnosis. Furthermore, users of generative AI were more likely than non-users to understand the limitations of generative AI, such as "confabulation" which can produce false information, and to be more aware of ethical considerations such as fairness and transparency.
However, overall awareness of ethical considerations was not always sufficient, with only about 47% of respondents being conscious of fairness and about 50% being conscious of transparency. On the other hand, the use of generative AI was independently associated with desirable clinical behaviors, such as emphasis on physical examination and appropriate use of antibiotics.
At present, the use of generative AI has not been associated with clinical behaviors indicating a decline in knowledge or skills, but rather with desirable clinical behaviors such as emphasis on physical examination and appropriate use of antibiotics. However, both users and non-users do not have a sufficient understanding of the limitations and ethical challenges of generative AI, highlighting the importance of medical education in the age of AI.
Paper information: [JMIR AI] Current landscape of generative artificial intelligence use as asearch engine among resident physicians: Cross-sectional study
