Prompt engineering helps AI tailor pediatric cataract materials to lower reading levels
Large language models can significantly improve the readability of online patient education materials (PEMs) on pediatric cataracts, making them easier for patients and caregivers to understand across multiple languages, especially when combined with web browsing and prompt engineering techniques, according to a study.
Researchers analyzed 103 PEMs in multiple languages using 3 ChatGPT-4o, Gemini 2.0, and DeepSeek-R1 to evaluate the impact of multilingual adaptation, content retrieval, and prompt engineering on readability.
The study found that large language model-generated PEMs initially exceeded a 10th-grade reading level, but when existing materials from Google searches were processed through large language models, readability improved significantly. DeepSeek-R1 produced the largest reduction, lowering the average reading level from 10.59 to 7.01. Advanced prompt strategies, including Zero-shot-Cot, were able to bring materials below a sixth-grade reading level.
Reference
Qiu X, Luo C, Zhang Q, et al. Enhancing the Readability of Online Pediatric Cataract Education Materials: A Comparative Study of Large Language Models. Transl Vis Sci Technol. 2025;14(8):19. doi: 10.1167/tvst.14.8.19. PMID: 40824260; PMCID: PMC12366858.
Related Content
