Physician enthusiasm for artificial intelligence outpaces readiness in Jeddah hospitals
Physicians and medical interns in Jeddah are already looking at artificial intelligence as a tool for screening and diagnosis, but the knowledge needed to use it safely is lagging behind the optimism. In a multicenter cross-sectional study across governmental and private hospitals, most participants described a positive attitude toward AI integration, yet only about half demonstrated good knowledge of AI concepts and applications. The result is a familiar clinical mismatch: willingness is high, but the technical foundation for dependable bedside use is incomplete.
A total of 435 physicians participated between February and April 2025. The study used a validated, self-administered questionnaire to assess demographic characteristics, knowledge, attitudes, and perceived barriers to AI use in clinical screening and diagnosis. Descriptive statistics and multivariate logistic regression were applied, giving the findings enough structure to move beyond opinion alone. In practical terms, the data describe a workforce that is open to AI but still unsure how to judge what it can and cannot do.
Knowledge gaps remain the central obstacle
Only 50.3% of respondents showed good knowledge of AI concepts and applications, a proportion that sits uncomfortably beside the 83.7% who held positive attitudes toward AI in clinical practice. The gap matters because AI tools in screening and diagnosis are not neutral accessories; they can shape triage, influence prioritization, and affect whether subtle disease is recognized early or overlooked. A clinician who trusts the output without understanding model limitations, training data, or error patterns may place patients at avoidable risk, especially in high-volume services where diagnostic speed can overshadow reflection.
Insufficient training was the most frequently reported barrier, cited by 58.9% of participants. Lack of trust in AI systems followed closely at 53.8%. Those concerns are not abstract. In screening pathways, clinicians must decide whether an AI-generated signal deserves escalation, repetition, or skepticism. Without training, the tool becomes a black box. Without trust, it becomes a tool that is admired in theory and resisted in practice.
Radiology and pathology were seen as the most fertile ground
Radiology emerged as the clinical field with the greatest perceived potential for AI implementation, selected by 75.9% of respondents. Pathology followed at 43.7%. Both specialties rely heavily on pattern recognition, image interpretation, and repetitive analytic tasks, which makes them logical entry points for machine learning systems. The appeal is easy to understand: AI can process large image sets quickly, flag anomalies, and support prioritization in overloaded services where human attention is finite.
Still, perceived suitability does not automatically translate into safe deployment. In radiology, false reassurance can be dangerous when an algorithm misses an early lesion or undercalls a subtle abnormality. In pathology, misclassification can distort diagnostic pathways and delay treatment decisions. The study’s specialty preference therefore signals opportunity, but it also points to the need for rigorous local validation, workflow integration, and physician oversight before AI becomes routine.
Non-Saudi physicians showed more favorable attitudes
One of the clearest associations in the multivariate analysis involved nationality. Non-Saudi physicians were significantly more likely to exhibit positive attitudes toward AI adoption, with an adjusted odds ratio of 1.75 and a P value of .047. The analysis does not explain why this difference emerged, but the finding suggests that professional background, prior exposure, training environment, or comfort with digital tools may shape openness to AI in measurable ways.
This result has direct organizational relevance. If readiness differs across physician subgroups, a single educational campaign may not be enough. Hospitals planning AI implementation need to anticipate uneven adoption patterns and design training that addresses varying levels of familiarity, confidence, and skepticism. A uniform rollout may produce uneven uptake, even when the technology itself is sound.
Vision 2030 gives the study its policy weight
Saudi Arabia’s healthcare digital transformation under Vision 2030 gives this study a wider frame. AI is not being evaluated in isolation; it is being considered as part of a national push toward modernized, data-driven care. In that setting, physician readiness becomes a core implementation variable rather than a secondary concern. Technology can be purchased quickly. Competence, trust, and institutional norms take longer to build.
The study makes clear that enthusiasm alone will not guarantee safe integration. Physicians who understand AI are better positioned to identify error, challenge implausible outputs, and use the technology as an adjunct rather than a substitute for clinical judgment. The learning curve is not merely technical. It is also ethical and operational, because AI touches accountability, transparency, and the chain of responsibility in diagnosis.
Training and regulation will determine whether adoption is sustainable
Structured educational programs emerge as the most direct response to the readiness gap. Training needs to cover core AI concepts, common clinical use cases, performance limitations, and the practical implications of false positives and false negatives. Trust-building will require more than promotional language. Physicians need evidence of local validation, clear explanation of how algorithms are built, and safeguards that preserve clinician oversight.
Clear regulatory frameworks are equally important. Without them, physicians are left to navigate uncertainty about accountability, data governance, and appropriate use. A system that introduces AI into screening and diagnosis without clear rules risks creating confusion at the point of care. The study’s findings support a simple but consequential conclusion: implementation should advance alongside education, governance, and clinical supervision, not ahead of them.
The sample was confined to physicians and medical interns in Jeddah, so the findings should not be treated as a national census of readiness. Even so, the pattern is difficult to ignore. AI is already viewed favorably, especially in image-based specialties, but the clinical workforce still lacks the knowledge and training needed to use it with confidence and caution. In that gap, the success of AI in Saudi healthcare will be decided.