How political actors discursively construct “women” as a category has become a major question across feminist political theory and political communication, with scholars demonstrating that gendered categories are deeply strategic in legislative debate. However, we still know remarkably little about the systematic patterns in how parties deploy different gendered referents — especially the shift between “girls” and “women” — and what ideological work these lexical choices perform in policy conflicts; this lack of knowledge prevents us from fully understanding how gender is instrumentalised within elite discourse. This paper fills this gap by examining how and when parties strategically switch between “girls” and “women” across politicised issue domains. Using US Senate and UK parliamentary debates, we construct contextual word embeddings and then link these embeddings to metadata on speakers, issue areas, parties, and time to run embeddings regressions that isolate the latent meaning structure of “girls” versus “women” in abortion, healthcare, sports, and related gender policy areas. We additionally unpack which semantic neighbourhoods cluster with each term, and how these neighbourhoods vary by party and issue. We argue that parties deploy “girls” to narrate their own role as protectors and guardians, while strategically using “women” to assign responsibility, agency, or blame — effectively mobilising gendered language to distribute accountability and moral duty. By tracing how elites weaponise lexical shifts between “girls” and “women,” this project sheds new light on the neglected mechanism through which gender is coded, moralised, and manipulated through the linguistic micro–morphology of democratic institutions
