Accuracy, completeness and readability of large language model chatbots answering dental patient questions: a systematic review

Introduction: Large language model chatbots are increasingly used by patients to seek answers about oral health, but the quality of the information they provide remains uncertain. Methods: A systematic review was conducted following PRISMA 2020. Scopus, PubMed and Crossref were searched up to 2026 for studies that evaluated chatbots answering patient questions about oral health against a reference standard, with no language restriction. Screening was assisted by artificial intelligence and verified by the authors. Results: Twenty-six studies published in 2025 and 2026 were included. ChatGPT was the most frequently evaluated model (21 studies), followed by Gemini (16) and DeepSeek (12). Accuracy was generally moderate to high but varied by model and clinical domain, with no model consistently superior. Readability was a recurrent weakness, with most responses exceeding the recommended sixth-grade reading level. Completeness was higher for chatbot responses than for clinician responses in one comparative study. Conclusions: Large language model chatbots provide largely accurate but frequently hard-to-read answers to dental patient questions. Their use should be adjunctive to professional care, and readability of outputs must be improved before they can safely support patient education.