A critical review of technical progress, clinical heterogeneity, and implementation challenges of artificial intelligence in digestive endoscopy

Introduction: Digestive endoscopy is a critical modality for diagnosing and managing gastrointestinal diseases, yet it faces challenges including operator dependence, procedural complexity, and potential complications. Artificial intelligence (AI) has emerged as a promising adjunct to address these limitations by enhancing diagnostic accuracy and optimizing procedural workflows. Areas covered: This review comprehensively examines the application of AI across diverse clinical practices in digestive endoscopy. Key areas include lesion detection and diagnosis, lesion characterization and classification, quality control, workflow optimization, and therapeutic guidance. Additionally, it highlights the principal AI technologies that have received regulatory approval from the Food and Drug Administration (FDA) and European CE marking, underscoring their clinical readiness and integration potential. Expert opinion: AI demonstrates significant potential to improve endoscopic outcomes by augmenting lesion detection rates and diagnostic precision. However, the translation of AI innovations into routine clinical practice is tempered by challenges, such as variability in clinical effectiveness, dependency on procedural quality, domain generalizability, and cost-effectiveness considerations. Future advancements should focus on enhancing AI robustness, integrating multimodal data, and establishing sustainable implementation frameworks to maximize clinical benefit while maintaining patient safety and ethical standards.