Podcasts have become an important medium for long-form public communication, providing a unique setting in which opinions, narratives, arguments, and social identities are developed through extended conversation. This project aims to develop and evaluate computational methods for understanding persuasive and potentially manipulative communication in podcasts, with particular attention to how linguistic, conversational, and acoustic signals contribute to processes such as framing, opinion formation, mobilisation, and collective action. Our previous work has used podcast transcripts to study how collective action is articulated in discussions surrounding the Black Lives Matter movement. This project is a continuation of NAISS 2025/22-1090.