AI Summary & Key Takeaways
AI InsightHowever, they should not be used as standalone sources in clinical decision-making and must be supported by up-to-date scientific evidence and clinician supervision.
📌 3 Key Takeaways
Abstract
Aim: The aim of this study was to comparatively evaluate the performance of the artificial intelligence (AI) models ChatGPT-4o Plus, Google Gemini 3 Flash, Claude 3.5 Sonnet, and Grok 4 in answering pediatric dentistry questions included in the Turkish Dental Specialization Exam (DUS). Methods: A total of 79 pediatric dentistry questions from DUS examinations conducted between 2015 and 2021 were initially identified from the official website of the Student Selection and Placement Center (ÖSYM), and 69 questions were included in the final analysis. Each question was submitted to the ChatGPT-4o Plus, Google Gemini 3 Flash, Claude 3.5 Sonnet and Grok 4 models by a single researcher on the same day, using a standardized prompt format. Correct answers were scored as “1” and incorrect answers as “0,” and accuracy rates were statistically compared between the models. Results: Based on the generalized linear model (GLM) analysis using the repeated measures approach and Wald Chi-Square tests, no statistically significant effects were observed for year, model, or the year × model interaction (p > 0.05). However, post-hoc pairwise comparisons with Bonferroni correction revealed limited year-specific differences (p < 0.05), with certain models demonstrating higher accuracy rates in specific years. Conclusion: Artificial intelligence models demonstrated effective performance in answering pediatric dentistry questions on the DUS and have the potential to be used as supportive educational tools in exam preparation. However, they should not be used as standalone sources in clinical decision-making and must be supported by up-to-date scientific evidence and clinician supervision.