Leveraging Large Language Models to Generate Multiple-Choice Questions for Ophthalmology Education

This study evaluated whether OpenAI's GPT-4 could reliably generate high-quality, novel multiple-choice questions (MCQs) for ophthalmology education comparable to those produced by experienced human experts. The findings demonstrated that LLM-generated MCQs matched human experts with identical median scores in appropriateness, clarity, and relevance. Notably, 95% of the AI-generated questions were novel, showing minimal similarity to existing databases while maintaining professional readability standards. These results suggest that LLMs offer an efficient way to expand examination resources and support ophthalmology residency training.