Objective: To evaluate the feasibility of using Large Language Model (LLM) to generate family health education materials for bronchopulmonary dysplasia (BPD). Method: Collect the 20 core issues that parents of children with BPD are most concerned about, and generate materials using GPT-4 based on the latest guidelines. Multidisciplinary experts evaluate the accuracy (Level 5 Likert) and reliability (DISCERN) of the content. 30 parents were randomly divided into two groups to read AI generated materials (intervention group) and traditional manuals (control group), and their readability (Flesch Kincaid), comprehension (PEMAT), and knowledge test scores were compared. Result: The accuracy score of AI material is 4.63 ± 0.42 (out of 5), and the total score of DISCRON is 70.2 ± 4.8. The readability level is 8.2 ± 1.3 grade, which is better than the traditional material's 10.5 ± 1.8 (P