Congenital nasolacrimal duct obstruction (CNLDO) is the most common lacrimal drainage disorder in infants, affecting 6–20% of newborns [1]. The condition is primarily caused by a membranous blockage at the distal end of the nasolacrimal duct. Although spontaneous resolution occurs in approximately 96% of cases by the age of one, persistent tearing and crusting can cause significant concern for parents [2]. To better understand the condition and its management, many parents seek health-related information online — either before or after consulting a physician.

Given the increasing reliance on the Internet for medical guidance, especially among caregivers, access to accurate, timely, and readable health information is essential. A cross-sectional study reported that 84.7% of respondents used the Internet to search for health information for themselves or their relatives [3]. However, the accuracy and clarity of online medical content can vary widely, and misleading or overly complex information may contribute to unnecessary parental anxiety or even delayed medical intervention.

Recent advances in natural language processing (NLP) have led to the development of large language models (LLMs) such as OpenAI's GPT-4o, Microsoft Copilot, Google's Gemini, and DeepSeek, which can generate fluent and context-aware answers to user queries [4], [5], [6], [7]. While these systems are increasingly used in health communication, their ability to deliver clinically accurate and readable responses remains under investigation [8], [9], [10].

This exploratory comparative study aims to evaluate the performance of four state-of-the-art LLMs in generating responses to frequently asked questions (FAQs) about CNLDO. Specifically, the study assesses (1) the appropriateness of responses — measured by clinical relevance, accuracy, and coherence — and (2) their readability, or how easily the information can be understood by laypersons. By benchmarking these models across both dimensions, the study seeks to identify their strengths and limitations and provide insight into how AI-generated content might support or mislead parental understanding in real-world pediatric ophthalmology contexts.