Changes in the acceptance of artificial intelligence in medicine among physicians and the general public in Japan: A repeated cross-sectional comparison of 2018 and 2025

Medical artificial intelligence (AI) has advanced rapidly, yet whether growing societal familiarity with AI is accompanied by greater willingness to receive relatively autonomous AI-led care remains unclear. We conducted repeated cross-sectional online surveys in Japan in 2018 and 2025 using the same questionnaire and compared self-identified physicians and members of the general public (2018: 399 physicians and 600 members of the public; 2025: 300 physicians and 300 members of the public). Mean responses to the item assessing intention to use medical AI did not differ significantly between survey years in either group, despite consistently favorable evaluations of its usefulness and future necessity. Among physicians, the unadjusted mean response regarding the potential of medical AI to improve clinical efficiency was higher in 2025, but this difference was not significant after adjustment for age and sex. Among members of the public, self-rated knowledge of medical AI was higher in 2025, while expectations that medical AI would increase healthcare costs were also greater. These findings suggest that increased self-rated knowledge may not necessarily be accompanied by greater intention to use relatively autonomous medical AI. Implementation strategies should therefore address concerns regarding responsibility, safety, data protection, and healthcare costs alongside efforts to improve public understanding.Author summary: Artificial intelligence (AI) is increasingly used in healthcare, where it may support diagnosis, treatment decision-making, and clinical operations. As AI technologies have advanced and become more familiar in society, attitudes toward medical AI may also have changed. However, greater familiarity with AI does not necessarily mean that physicians or members of the public are more willing to receive care led by AI. To examine these changes, we compared two online surveys conducted in Japan in 2018 and 2025 using the same questionnaire. Both physicians and members of the general public continued to recognize the usefulness and future necessity of medical AI, but their intention to use it did not significantly differ between the two survey years. Among physicians, expectations that AI could improve clinical efficiency were higher in the unadjusted comparison, but this difference was no longer significant after accounting for age and sex. Among members of the public, self-rated knowledge of medical AI was higher in 2025, while expectations that AI would increase healthcare costs were also greater. These findings suggest that greater knowledge or exposure to AI may not automatically lead to greater willingness to use relatively autonomous medical AI. Building informed trust will require improving understanding while also addressing persistent concerns surrounding the use of medical AI.