Artificial intelligence exposure, perceived needs, preferences, and concerns among endocrinologists and trainees in China: a convenience survey

Artificial intelligence (AI) is increasingly applied in clinical practice, research, and medical education, yet its understanding and exposure among endocrinologists and postgraduate trainees, as well as their reported experiences, perceived needs and preferences, and concerns, remain insufficiently characterized. This study aimed to describe self-reported AI understanding and exposure, experiences with AI tools, perceived needs and preferences, and concerns among endocrinologists and postgraduate trainees in China, with particular consideration of their potential implications for endocrinology specialist training. We conducted a convenience survey in China in February 2026 using a self-developed, 22-item electronic questionnaire. Descriptive indicators and items that did not require prior AI exposure were analyzed in the full sample. Questions about AI-related experiences and evaluations of existing tools were analyzed among respondents with prior AI exposure. Ordered variables were compared between trainees and endocrinologists using two-sided Mann–Whitney U tests, whereas nominal and multiple-response variables were compared using Pearson’s chi-square or Fisher’s exact tests. Effect sizes were reported, and no adjustment was made for multiple exploratory comparisons. Of 275 potentially eligible individuals identified from department-administered personnel and training records and the membership rosters of closed professional WeChat groups,46 were excluded before invitation.Among the 229 eligible individuals invited directly via WeChat,185 submitted complete questionnaires and 44 did not submit a response, corresponding to response rate of 80.8% (185/229).In the full sample, 80/185 (43.2%) described their understanding of AI as general, and 151/185 (81.6%; 95% CI 75.4%–86.5%) reported at least occasional AI exposure and constituted the AI-exposed subgroup; personal AI use was not independently confirmed. A total of 104/185 (56.2%; 95% CI 49.0%–63.2%) selected the strongest expectation category for their hospital or school to introduce dedicated AI tools. Among AI-exposed respondents, independent online searching was the most frequently selected route of exposure (142/151, 94.0%; 95% CI 89.1%–96.8%). The most frequently selected concerns were inaccurate AI diagnoses or suggestions (149/185, 80.5%; 95% CI 74.2%–85.6%), overreliance on AI (128/185, 69.2%; 95% CI 62.2%–75.4%), and data privacy breaches (115/185, 62.2%; 95% CI 55.0%–68.8%). The most frequently endorsed predefined risk-mitigation strategies were labeling information sources (140/185, 75.7%), clarifying that AI cannot replace clinical judgment (131/185, 70.8%), regularly updating specialty guidelines and cases (119/185, 64.3%), and strengthening data encryption and protection (117/185, 63.2%). Exploratory subgroup comparisons yielded small-to-moderate effect sizes and were interpreted descriptively. Respondents reported heterogeneous AI exposure, frequent self-directed access to AI tools, substantial interest in institutionally introduced tools, and concerns about accuracy, over-reliance, and privacy. These descriptive findings identify priorities for the future development and evaluation of AI tools, training, and governance in endocrinology.