Peroral cholangioscopy (POCS) enables direct intraductal visualization and targeted biopsies for indeterminate biliary strictures. However, its diagnostic performance remains operator-dependent and is limited by sampling error. As artificial intelligence (AI) has rapidly advanced in gastrointestinal endoscopy, AI-assisted cholangioscopy has progressed from proof-of-concept image classification to multicenter validation, feature detection, and real-time application. Recent studies using digital single-operator cholangioscopy have reported diagnostic accuracy comparable to that of experts in differentiating malignant from benign strictures, and feature-based approaches can automatically detect tumor vessels and other key visual characteristics. Beyond classification, generative models may enhance mucosal and margin visualization through virtual chromoendoscopy, potentially improving lesion-margin assessment. Furthermore, new models compatible with multiple systems are emerging to address device heterogeneity, which is a key barrier to routine practice. This review summarizes the current evidence on AI-assisted POCS, discusses practical considerations for clinical translation, and highlights future directions, including real-time integration and multimodal decision support.