WGO AI Guidelines

Artificial intelligence is increasingly transforming gastrointestinal endoscopy, particularly in endoscopic ultrasound (EUS), biliary disease, and upper gastrointestinal (GI) endoscopy. EUS plays a key role in evaluating subepithelial lesions (SELs) and solid and cystic pancreatic neoplasms, incorporating techniques such as elastography and contrast enhancement. Nonetheless, its diagnostic performance remains strongly operator-dependent, with notable interobserver variability and only moderate accuracy even among experienced endosonographers. Computational models applied to EUS images and videos have shown encouraging results in lesion detection, characterization, and risk stratification, in some cases reaching or exceeding expert-level performance. However, most available data are retrospective and derived from single-center studies. Cholangiocarcinoma continues to be associated with poor outcomes, largely due to delayed diagnosis and the difficulty of distinguishing malignant from benign biliary strictures. Although advances in cholangioscopy, imaging, and tissue acquisition have improved diagnostic yield, indeterminate strictures remain frequent. The incorporation of artificial intelligence into imaging analysis, radiologic assessment, and tissue evaluation may improve diagnostic precision, decrease unnecessary procedures, and provide real-time decision support. In upper GI endoscopy, detection and characterization of esophageal and gastric neoplasia remain inconsistent in daily practice and are highly dependent on operator expertise. Emerging systems can enhance lesion recognition, reduce variability among observers, and assist real-time assessment. Overall, these technologies represent a promising adjunct across endoscopic modalities, although prospective multicenter validation and real-time clinical integration are still required for widespread adoption.