Generative AI in Endodontics: A Scoping Review of Applications, Performance, and Limitations

Generative artificial intelligence (AI), particularly large language models (LLMs), has been increasingly studied and applied in endodontics, offering accessible tools for clinical decision support, education, and patient communication without requiring advanced technical expertise. This scoping review aimed to map current applications of LLM-based generative AI in endodontics and to synthesize evidence on their performance, limitations, and key evidence gaps.

A comprehensive literature search using a predefined Boolean search strategy was conducted in PubMed, Scopus, and Web of Science Core Collection for studies published between January 1, 2023 and November 19, 2025. English-language studies evaluating LLM-based generative AI applications in endodontics were included. Data were independently extracted by two reviewers and synthesized to identify applications, key outcomes, and limitations.

A total of 50 studies were included. Applications were categorized into education and training (n = 20), diagnosis and clinical support (n = 21), and patient communication (n = 9). GPT-series models via ChatGPT were the most frequently assessed. In education, LLMs demonstrated high accuracy in structured multiple-choice assessments, but showed lower performance in open-ended and case-based tasks; in the four studies that included human comparators, LLMs outperformed students in two and fell below human examinees in two. In diagnostic and clinical decision-making contexts, performance was promising, but inconsistent, with improvements being observed when structured clinical input was provided, yet remaining below expert-level performance. In patient communication, LLMs were associated with improved patient ratings and engagement, but revealed a trade-off between accuracy, readability, and completeness.

This review shows the potential of LLM-based generative AI in endodontics; however, its performance is strongly affected by task type and both model- and implementation-related factors. Therefore, the development of standardized, prospectively validated evaluation frameworks is important for supporting a reliable assessment and is expected to facilitate the safe and effective integration of generative AI into clinical and educational practice.