Evaluating the efficacy of large language models in providing information for parental inquiries regarding primary care of pediatric oral and dental health

Background: The use of artificial intelligence (AI)-based large language models (LLMs) for accessing health information is rapidly increasing; however, limited research has evaluated their efficacy in pediatric dentistry, particularly regarding the accuracy of oral health information. This study aimed to assess the accuracy, readability, and similarity of four AI-based LLMs; ChatGPT 4.0, Google Gemini, Microsoft Copilot, and DeepSeek-R1 in responding to parental questions about children’s oral and dental health. Methods: Twenty frequently asked questions, developed by experienced pediatric dentists, were presented to each LLM over a seven-day period, from 27 January 2025 to 03 February 2025. Responses were independently assessed for accuracy, readability, and similarity. Statistical analyses used IBM SPSS 27.0. Normality was checked by Kolmogorov-Smirnov. T-tests/Analysis of Variance (ANOVA) with Tukey’s Honestly Significant Difference (HSD) were applied for normal data; Mann-Whitney U/Kruskal-Wallis with Dunn’s post hoc and Bonferroni correction for non-normal data. Fisher’s Exact test examined categorical variables; binary accuracy after Likert dichotomization. Inter/intra-rater reliability employed two-way random effects Intraclass Correlation Coefficients (ICC). p < 0.05 indicated significance. Results: The study demonstrated that ChatGPT 4.0 and DeepSeek-R1 achieved significantly higher accuracy scores compared to Google Gemini and Microsoft Copilot (p < 0.001). ChatGPT 4.0 also produced the most readable content, reflected by the lowest Average Reading Level Consensus (ARLC) score of 8.05, whereas DeepSeek-R1 generated the shortest responses (p < 0.001). Regarding originality, ChatGPT 4.0 and Google Gemini exhibited the lowest similarity indices, indicating a greater degree of response diversity (p = 0.02). Conclusions: These findings underscore the potential role of AI-based LLMs in facilitating parental access to evidence-based pediatric dental health information. Based on accuracy, readability, and similarity results, ChatGPT 4.0 appears the most reliable platform for delivering oral health information to parents. Furthermore, the newly introduced DeepSeek-R1 showed comparable accuracy and originality to ChatGPT 4.0, highlighting its promise as an efficient tool for user-friendly dental health guidance.