ABSTRACT Introduction The use of large language models (LLMs) such as ChatGPT to generate multiple‐choice questions (MCQs) for medical and dental education is rapidly increasing. However, the educational validity, cognitive depth, and practical usability of AI‐generated questions remain underexplored in dental education. This study aimed to evaluate the performance of ChatGPT‐4o in generating MCQs for pre‐doctoral oral and maxillofacial radiology curricula. Methods ChatGPT‐4o was prompted to generate 100 multiple‐choice questions based on lecture materials from a pre‐doctoral oral and maxillofacial radiology course. A panel of expert oral radiologists independently evaluated the quality of the questions, answers, and explanations. Additionally, a randomised subset of 64 MCQs was assessed by human experts and an AI‐detection tool to determine the accuracy of source identification. Results Experts rated 43% of AI‐generated questions and 65% of their corresponding answers as correct and usable with minor adjustments. Most questions focussed on knowledge recall, with few assessing higher order thinking skills. Both human experts and AI‐detection tools struggled to accurately differentiate between AI‐generated and human‐created questions. Conclusion While ChatGPT‐4o can generate MCQs, its output often requires refinement. Future research should explore ways to improve the quality and cognitive level of AI‐generated questions for dental education.
Elkersh et al. (Thu,) studied this question.