LARGE language models in glaucoma show promise for education and decision support, yet performance varies widely.
Large Language Models in Glaucoma: What Was Reviewed
A scoping review assessed how large language models in glaucoma and related vision language approaches are being used across research and early clinical scenarios. The authors systematically searched five databases from 2014 through July 2025 and screened studies using a structured methodology. Of 316 records identified, 27 studies met inclusion criteria, capturing original research and research letters that applied these models to glaucoma related tasks.
Where the Evidence Is Strongest
Most included studies clustered into patient facing communication and education, diagnosis and risk prediction, and surgical management support. Patient education represented the largest group, suggesting that large language models in glaucoma may be especially useful for explaining concepts, translating clinical language into more understandable terms, and supporting consistent messaging. Across several studies, text based clinical decision support also appeared promising, particularly when models were asked to summarize information, organize differential considerations, or respond to structured prompts.
Key Constraints for Clinical Use
Despite encouraging applications, the review emphasizes that current systems are best positioned as assistive tools rather than autonomous decision makers in glaucoma care. Limitations reported across the included studies centered on variable accuracy, gaps in multimodal integration when image interpretation is required, and insufficient domain specific fine tuning for ophthalmology. The authors also highlight the need for approaches that better ground outputs in reliable sources, improve text image fusion for real world glaucoma workflows, and adapt readability for patients without losing clinical nuance.
What Comes Next
Future work, the authors argue, should prioritize domain trained and retrieval augmented models, stronger evaluation standards, and clearer ethical and regulatory frameworks to support safe implementation. Taken together, the current evidence suggests large language models in glaucoma are advancing quickly, but clinical deployment should be guided by validation, transparency, and careful oversight.
Reference: Rubegni G et al. Applications of Large Language Models in Glaucoma: A Scoping Review. Vision. 2026;10(1):9.
