The way in which information is linguistically presented can impact audience attention, emotion, and cognitive responses, even if the content remains unchanged. The present study aims to examine the effects of rhetorical devices on audience responses by introducing a new theoretical framework, the augmented elaboration likelihood model (A-ELM), which integrates elements of the Elaboration Likelihood Model and narrative theory. The results show that the mediation effects of attention on the relationships between rhetorical devices and affective and cognitive elaborations are moderated by involvement. Nonnarrative evidence, combined narrative and numerical evidence, source credibility, and tropes versus the lack of figures of speech, elicit better audience responses in low-involvement situations, whereas numerical evidence outperforms narratives in high-involvement situations. This study not only offers a novel theoretical framework in the form of A-ELM, but also has important implications for advancing methodologies and practical applications.