Objective:

To evaluate the quality, readability, and actionability of hypertension patient education materials generated by various large language models (LLMs).

Approach:

• Study Design: A cross-sectional comparative study was conducted using a curated set of 20 frequently asked questions about hypertension, validated by cardiovascular experts.
• Evaluation Metrics: Quality, readability, and actionability of AI-generated materials were assessed using various tools including the Patient Education Materials Assessment Tool (PEMAT-P) and seven readability formulas.
• AI Platforms: Six LLM platforms were selected for analysis: ChatGPT 5.2-Thinking, DeepSeek-R1, Doubao, ERNIE Bot 4.5 Turbo, Qwen3-Max-Thinking-Preview, and Kimi K2.

Key Findings:

• Many existing patient education materials exceed the recommended sixth-grade reading level, limiting patient comprehension.
• Variability in output quality exists among different LLM platforms.

Interpretation:

The study highlights the need for standardized evaluation of generative AI outputs to ensure they meet patient education needs effectively.

Limitations:

• The study focused on a limited number of LLM platforms.
• Evaluation metrics may not capture all aspects of patient engagement and understanding.

Conclusion:

The findings suggest that while generative AI can enhance patient education materials, careful evaluation is necessary to ensure their effectiveness.