ChatGPT response accuracy for amblyopia and childhood myopia questions
Reviewed by Fiona Rowe

With the increasing use of ChatGPT by the public to obtain information, the authors of this study aimed to critically evaluate the scientific quality of the responses to frequently asked questions about amblyopia and childhood myopia. They submitted 27 and 28 questions respectively for the two topics (110 questions in total) on two consecutive days. The answers were reviewed by two independent paediatric ophthalmologists with high agreement of 96.4% on their assessment of the response. The responses obtained on day two were almost identical to day one for 87.3% of ChatGPT responses. Acceptable responses were obtained for 87.3% overall: 81.5% for amblyopia and 87.5% for childhood myopia. For amblyopia and childhood myopia responses, answers were incomplete for 12.9% and 7.1%, or inappropriate for 5.6% and 5.4% respectively, with inaccurate treatment facts. There was a poorer performance of ChatGPT when questions were at a higher level and more subspecialised. This study shows the positive potential of ChatGPT but also urges caution, particularly regarding specifics of treatment options.

Assessment of the responses of the artificial intelligence-based Chatbot ChatGPT-4 to frequently asked questions about amblyopia and childhood myopia.
Nikdel M, Ghadimi H, Tavakoli M, Suh DW.
JOURNAL OF PEDIATRIC OPHTHALMOLOGY AND STRABISMUS
2024;61(2):86–9.