Artificial intelligence (AI) is increasingly promoted as a tool for improving health care delivery, yet successful implementation depends on workforce and organizational readiness. This cross-sectional analytical study assessed perceived barriers to AI integration and associated factors among 480 health care professionals working in public facilities in the Marrakech-Safi region of Morocco. Participants were recruited through professional and institutional networks and completed an anonymous electronic questionnaire. The final 11-item instrument measured technological, organizational, ethical and personal, and interdisciplinary collaboration barriers on a five-point scale, with higher scores indicating greater barriers. Organizational barriers had the highest mean score (3.36 ± 0.84), followed by ethical and personal barriers (2.69 ± 0.99), technological barriers (2.61 ± 0.76), and interdisciplinary collaboration barriers (2.25 ± 0.90). The overall score was 2.76 ± 0.64, and 80.2% of participants reported no prior formal AI training. In multivariable general linear models, prior formal AI training was the only factor significantly associated with all four dimensions and the overall score, with the largest effect sizes for organizational barriers (η_p² = 0.326) and overall barriers (η_p² = 0.148); these two estimates were attenuated (η_p² = 0.088 and 0.078) in a sensitivity analysis excluding one scale item that overlapped conceptually with the training variable. Because of the cross-sectional, self-reported design, this association cannot be interpreted causally. Age, professional category, educational qualification, and department showed dimension-specific associations, whereas gender and professional experience were not independently associated after adjustment. These findings indicate that AI integration in Moroccan public health care is constrained primarily by organizational preparedness and limited training. Role-specific education, clear institutional policies, supportive leadership, and adequate digital infrastructure are therefore needed. Given the cross-sectional design and convenience sampling, causal inference and broad generalization should be approached with caution.