Abstract: Artificial intelligence, centered on technologies such as deep learning and computer vision, is permeating every stage of endoscopic retrograde cholangiopancreatography (ERCP). It integrates imaging and laboratory data to precisely assess procedural necessity and stratify risk; analyzes real-time cholangioscopy or fluoroscopy images to accurately identify malignant strictures, locate the ampulla, and warn of difficult cannulation, thereby significantly reducing ERCP procedural difficulty and radiation dose; and efficiently predicts postoperative complications such as pancreatitis and cholangitis, enabling individualized treatment. AI has already demonstrated diagnostic accuracy surpassing that of traditional methods, predictive performance superior to experiential judgment, and training advantages in shortening the learning curve. In the future, generative adversarial networks may more efficiently synthesize high-quality ERCP images, providing a "digital film library" for zero-risk training and data augmentation of rare cases.