When evidence meets artificial intelligence

Evidence-based medicine (EBM) integrates clinical expertise, patient values, and best available research evidence, yet classical trial-based approaches struggle to accommodate the scale, heterogeneity, and structural constraints of contemporary health systems. The rapid expansion of biomedical data is positioning artificial intelligence (AI) as an extension of evidence generation rather than a replacement for epidemiologic reasoning. This state-of-the-art Review examines how AI is beginning to reshape the methodological and ethical foundations of EBM through multimodal data integration, target trial emulation, and simulation-based modelling, while emphasizing that predictive performance does not ensure causal validity or clinical benefit. Implementation depends on rigorous epidemiologic design, integration within clinical informatics systems, and effective human-machine collaboration. In the Americas, fragmented health systems and uneven digital infrastructures constrain representativeness and equitable deployment. AI will not transform EBM on its own; its constructive role requires governance reform, transparent validation, and structural change to prevent the reinforcement of existing inequities.