Cardiovascular disease (CVD) is the leading cause of death worldwide, and AI-powered natural language processing (NLP) models, such as ChatGPT and MediSearch, have the potential to improve patient education by providing accessible health information. This study evaluated the readability and similarity of CVD-related responses from ChatGPT Free (GPT-4o mini), ChatGPT Premium (GPT-4o), and MediSearch across three education levels: Lower Secondary, Higher Secondary, and College Graduate. FAQs on CVD were reformatted into user-friendly prompts and entered into each model. Readability was assessed using the Flesch-Kincaid Ease Score and Grade Level, while response similarity was analyzed using BERT embeddings with Cosine Similarity. Results showed that readability significantly decreased as education level increased (p<0.001), with ChatGPT Free providing the most accessible responses and MediSearch delivering more complex outputs. ChatGPT models demonstrated greater similarity to each other than to MediSearch. These findings emphasize the need for improved AI-driven medical communication tools.
