Evaluating the Accuracy, Completeness, and Readability of Chatbot Responses to Refractive Surgery-Related Patient Questions: A Comparative Analysis of ChatGPT and Google Gemini

Purpose This study evaluates the performance of ChatGPT and Google Gemini in addressing refractive surgery-related patient questions by analysing the accuracy, completeness, and readability of their responses. Methods A total of 40 refractive surgery-related questions were compiled and categorized into three levels of difficulty: easy, medium, and hard. Responses from ChatGPT and Google Gemini were blinded and evaluated by two experienced ophthalmologists using standardized criteria. Accuracy was scored on a six-point Likert scale, completeness on a three-point scale, and readability using Flesch-Kincaid Grade Level, Gunning Fog Index, Simple Measure of Gobbledygook (SMOG) Index, and word count. Intra- and inter-rater reliability were assessed using intra-class correlation coefficients (ICC). Results Both chatbots demonstrated high intra-rater (ICC>0.75) and inter-rater reliability. Accuracy scores were similar for most questions; however, statistically significant differences were observed for harder questions, where Gemini showed slightly reduced performance compared to ChatGPT. Readability metrics revealed no significant differences between the two tools, although ChatGPT responses tended to be more detailed, while Gemini generated more concise answers. Harder questions resulted in longer and more complex responses, as indicated by higher Gunning Fog and SMOG Index scores. Conclusions ChatGPT and Google Gemini exhibit strong potential in patient education, with complementary strengths in accuracy, readability, and response detail. The influence of question complexity on chatbot performance highlights the need for ongoing optimization to enhance both clarity and accessibility. These findings underscore the value of integrating artificial intelligence (AI) tools into healthcare to support patient education and engagement.