Artificial Intelligence in endodontics: from diagnosis to clinical decision-making

To provide a comprehensive overview of contemporary Artificial Intelligence (AI) applications in endodontics and critically evaluate their clinical relevance, diagnostic performance, and future potential. A narrative review was conducted using PubMed/MEDLINE, Scopus, and Web of Science databases. Peer-reviewed original studies, systematic reviews, and meta-analyses published between 2020 and 2026 on AI applications in endodontics were included. AI applications demonstrated promising performance in periapical lesion detection, root canal morphology assessment, second mesiobuccal canal (MB2) identification, pulpitis diagnosis, treatment outcome evaluation, postoperative pain prediction, and surgical planning. Large language models and retrieval-augmented systems further expanded the role of AI in clinical decision support. Several studies reported diagnostic performances comparable to experienced clinicians. However, challenges related to external validation, annotation quality, reproducibility, explainability, and hallucinations remain significant barriers to widespread clinical implementation. AI has the potential to substantially improve diagnostic accuracy, treatment planning, and clinical decision-making in endodontics. Nevertheless, current systems should be considered adjunctive tools rather than replacements for clinician expertise. Future research should focus on externally validated, transparent, and clinically applicable AI models to facilitate safe integration into routine endodontic practice.