Evaluation of artificial intelligence generated responses of commonly asked ophthalmic queries

Purpose: To evaluate artificial intelligence generated responses of commonly asked ophthalmic queries. Methods: In this study, 10 common questions related to ophthalmic conditions were prepared by experienced ophthalmologists based on common queries by patients in India. The responses to these questions were obtained from ChatGPT. The responses were scored for accuracy, completeness, and understandability by two experienced unacquainted ophthalmologists. Results: In this study, mean values of overall scores for accuracy, completeness, understandability of ChatGPT responses graded fair (mean scores of 3.25 ± 0.75, 1.95 ± 1.5 and 2.3 ± 0.7, respectively). Understandability of responses to surgery related queries were significantly low compared to non-surgical ones ( P = 0.041). Conclusions: Study findings indicate ChatGPT’s responses to overall accuracy, completeness and understandability were just fair, suggesting ChatGPT offers information to some extent, but it falls short of expert ophthalmologist’s level.