Rhetorical echoes: source-audience style alignment in youtube immigration debates

The positions users express during political discussions on social media function as acts of social identity work. While the ideological alignment between online sources and their audiences is well documented, whether this alignment extends to the rhetorical style they use to express their positions, remains largely unexplored. We conduct a large-scale analysis of immigration-related YouTube discussions spanning 2020–2024. We release a corpus of user comments and video transcripts span-annotated with divisive rhetorical techniques and develop token-level multi-label span detection models for both domains, enabling source–audience comparison via a unified rhetorical intensity measure. Commenting audiences are segmented using the combination of the channels where users choose to post their comments and their stance towards immigration. YouTube channels exhibit distinct rhetorical profiles shaped jointly by political leaning and source type. Audiences display cluster-specific rhetorical signatures related to the ideological composition of each group, while sharing a common macro-category hierarchy consistent with divisive rhetoric functioning as a domain-level communicative toolkit. We provide statistical evidence () of the rhetorical alignment between matched source and audience groups. Interpreted through the lens of rhetorical ecology, our findings are consistent with users and channel owners co-constitutively enacting, normalising and reproducing the rhetorical means through which identity work is performed, though the correlational design cannot adjudicate whether this reflects audience selection, source-driven influence, or both.