OpenAI Vision is capable of detecting retinal pathologies, including diabetic retinopathy (DR) and macular degeneration (MD). This paper examines the vision disorders detectable by OpenAI and compares its effectiveness against traditional convolutional neural networks (CNNs). The proposed method prioritizes the most serious and common vision disorders, such as DR, MD, and glaucoma. Publicly available datasets were utilized for performance comparison. The findings indicate that conventional CNNs exhibit higher accuracy than OpenAI Vision, although both show comparable specificity for DR and MD. Unlike the CNNs tested, OpenAI Vision is pre-trained on a large number of datasets, making it independent of the image acquisition device used, which enhances its suitability for clinical applications. OpenAI Vision demonstrates high specificity, resulting in few false positives. Therefore, when a negative diagnosis is issued, it is highly likely that the patient does not have the targeted disorder. However, its low sensitivity (the ability to detect true positives) currently makes it unsuitable for clinical screening of DR and MD. This study lays the groundwork for future research aimed at further assessing the usability of OpenAI Vision and similar large language models (LLMs) for developing clinically relevant classifiers in the field of ophthalmology.