Large language models and generative artificial intelligence in endodontics: a scoping review

The aim of this study is to comprehensively examine the evolution of artificial intelligence (AI), specifically large language models (LLMs), in the field of endodontics under the headings of clinical decision support systems, trauma management, regenerative treatments, patient education, and academic performance. This scoping review was conducted in accordance with the PRISMA-ScR guidelines. A search was performed on March 9, 2026, across the PubMed and Scopus databases, screening original research articles published in English between January 1, 2020, and March 9, 2026. A total of 54 studies involving the applications of LLMs and generative AI (GenAI) in the field of endodontics were included in the qualitative synthesis. The studies demonstrated that LLMs (specifically ChatGPT-4o, ChatGPT-5.1, ChatGPT-5 Plus, Gemini 2.5 Pro/Flash, Claude 3.5/3.7, and DeepSeek-R1) compete with specialist clinicians in standard theoretical and clinical scenarios, even achieving equivalent alignment in certain cases; however, their performance declines in complex cases. It has been determined that the Retrieval-Augmented Generation (RAG) architecture significantly reduces the risk of 'hallucination' by integrating current clinical guidelines to make the system document-grounded. However, critical medico-legal challenges and ethical constraints persist, such as data confidentiality, algorithmic biases, malpractice accountability, and the necessity of informed consent. While artificial intelligence holds transformative potential in endodontics, it must not be regarded as an autonomous decision-maker. To mitigate risks like 'automation bias' and potential deskilling, AI should be positioned as an 'assistive clinical adjutant' that operates in seamless synergy with human proficiency, ultimately augmenting the clinician's cognitive filtration and legal responsibility.