Patterns of artificial intelligence use among physicians in Mexico: Implications for oncology practice.

9002 Background: Artificial intelligence (AI) is increasingly incorporated into medical practice; however, physician-level factors associated with AI adoption and usage patterns, particularly among oncologists practicing in low- and middle-income countries, remain insufficiently characterized. This study aimed to describe patterns of AI use among physicians in Mexico and to explore physician-level factors associated with adoption. Methods: We conducted a cross-sectional, anonymous survey among physicians from multiple specialties in a tertiary care center in Mexico to assess AI use, professional activities (clinical care, research, and teaching), training interests, and perceptions regarding AI. The primary outcome was self-reported AI use. Associations between physician characteristics and AI use were evaluated using nonparametric statistics. Among oncologists, AI use patterns, training interests, and perceptions were summarized descriptively. Results: Between August and September 2025, 170 physicians completed the survey. Overall, 77.1% were < 40 years old, 50% were women, and 91.7% had ≤10 years of professional experience. Clinical specialties predominated (78.8%) over surgical specialties (21.2%). Most respondents were involved in clinical care (94.1%), followed by research (48.4%), and teaching (27.1%). AI use was highly prevalent (88.2%), with generative models being the most commonly used tools (80.0%). Physicians younger than 40 years reported significantly higher AI use than those aged ≥40 years (93.1% vs 71.8%, p<0.001) and were more likely to use AI in clinical practice (77.9% vs 43.6%, p<0.001). No significant age-related differences were observed for research or teaching activities. Use of AI-based data analysis tools varied by specialty and was more frequent among surgical compared with clinical specialties (63.9% vs 37.3%, p=0.004). Among oncologists (n=32), 87.5% reported AI use, primarily in clinical practice (75.0%) and research (50.0%), while use in teaching was less frequent (21.9%). Most oncologists expressed strong interest in formal AI training (84.4%) and ethical considerations (87.5%). The majority agreed that AI could improve quality of care and support clinical decision-making (95.2%), while emphasizing the need for continued physician oversight. Concerns regarding data security were common (84.7%), whereas concerns about loss of medical autonomy were generally moderate (36.4%). Conclusions: AI adoption among physicians in Mexico is strongly associated with age and professional activity. Oncologists demonstrate widespread AI use in clinical care and research, high interest in structured and ethical training, and a consistent emphasis on human oversight. These findings support the development of targeted educational and governance strategies to guide responsible AI integration in oncology.