Content compensation design for older adults’ perceived health information comprehension based on large language models: a random experiment in China

Abstract As Chinese older adults increasingly rely on online health resources, ensuring the accessibility and understandability of such information is critical. In the Chinese context,this study investigates the effectiveness of an AI-enabled content compensation design for improving older adults’ perceived comprehension of health information. A randomized experiment with Latin square design was conducted in China with 264 older adult participants, who were randomly assigned to read six versions of health information materials, including the original version and modified versions with improved simplicity, positive framing, metaphor framing, narrative framing, and increased cohesion, generated by ChatGPT 4o. The results revealed that, compared with the original version, all the modified versions significantly improved the perceived comprehension of health information, with metaphor framing showing the most substantial improvement. Additionally, the study revealed that health literacy and age were significant predictors of perceived comprehension, whereas education level and subjective health were not. This research expands the application of message framing theory in online health communication for older adults and provides practical guidelines for creating accessible health information. The study also demonstrates the potential of LLMs in automatically modifying health information to suit the needs of older adults and proposes a two-stage model that combines linguistic characteristics with message framing and manual review.