Generating targeted and tailored health communication narratives with AI

Customized narratives are effective tools to promote risk prevention behaviors in populations. However, the development of such narratives is resource-intensive. Advances in generative artificial intelligence (AI) offer promising solutions to these constraints. This study explored the potential of generative AI, such as ChatGPT, to create health communication narratives (HCN) tailored to specific demographics and theoretical constructs to convey critical health risk information. We conducted a two-phased, preregistered experiment involving cervical cancer screening messaging (N = 272) to examine the effectiveness of AI-tailored health communication narratives. After prescreening, participants were randomly assigned to read a nonnarrative, a generic narrative, or an AI-generated targeted or tailored narrative message. The targeted narratives were customized by AI according to participants' preassessed demographics, such as age, race, and ethnicity, and the tailored narratives further incorporated preassessed cues, such as fear of the screening procedure, cancer worry, and perceived benefit of cancer screening. A qualitative review of the AI-generated messages attests to ChatGPT's ability to craft personalized health narratives. However, respondents perceived AI-tailored narratives as lower in quality and believability and were not more motivated for cervical cancer screening uptake than those exposed to the nonnarrative and generic narrative messages. Older participants and those who had higher cancer worry or believed that cervical cancer screening is beneficial were more persuaded by the AI-tailored messages. AI models, such as ChatGPT, can create targeted and tailored risk messages with high efficiency, but additional fine-tuning is needed to improve their applicability in motivating health behavioral change.