Toward Scalable Content Generation for Gamified mHealth Interventions: The Evaluation of LLM-Generated Goals on User Engagement

Gamified mHealth applications are increasingly used to promote healthy behavior change, yet low participant engagement remains a challenge. Personalized content has shown promise in increasing engagement, however, creating personalized content is a time-consuming task. Generative Artificial Intelligence models, particularly Large Language Models, offer a potential solution with their ability to quickly generate relevant content. In this study, we aim to evaluate the impact of LLM-generated content on participant engagement in health interventions using gamified mHealth applications. A total of 73 students and staff members of a university participated in a health intervention and were assigned into groups receiving either LLM-generated goals or predetermined goals. Engagement and perceived intrinsic motivation were measured and compared between the two groups. The results show no significant difference in engagement or perceived intrinsic motivation between participants receiving LLM-generated goals and those receiving predetermined goals. LLM-generated goals did not significantly negatively impact participant engagement and could therefore possibly offer a time-efficient approach to scalable content generation for mHealth applications.