Answers are grounded in this paper and cite their sources.
Large Language Models (LLMs) are rapidly reshaping the landscape of conversational agents for health behavior change, enabling more human-like interactions than earlier rule-based systems. In the domain of physical activity promotion, however, the understanding of the outcomes of these technologies remains fragmented. This review examines the current state of LLM-based conversational agents designed to support physical activity, drawing on 13 studies. The analysis identifies three cross-cutting themes. First, it emphasizes the experiential, motivational, and behavioral outcomes of the reviewed studies, highlighting positive effects on user engagement, while stressing that evidence for a direct, sustained impact on objectively measured physical activity remains limited. Second, it shows that LLMs may assume a variety of social roles, entailing different relational dynamics. Third, it points out that people anthropomorphize LLM-based conversational agents, which can enhance emotional investment and strengthen the user-agent “relationship”, but may also foster over-reliance and misplaced expectations. Building on these findings, we critically discuss ethical concerns raised by the growing persuasive capacities of LLMs in this domain, including the redistribution of agency between users, technologies, and third parties and the risks tied to users’ tendency to ascribe humanness to artificial agents.
No takes yet. Share an insight, caveat, or question.
Silacci et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: