Twitter has emerged as an important forum for discussion among academic librarians. In this research, we take a mixed‐methods approach to study the thematic content and sentiment of tweets authored by academic librarians in the United States, Canada, and the United Kingdom. We found differences in the semantic content and themes present in the data from each country that point to differences in how librarians in each country engage on Twitter. While more work remains to be done, we cast new light on how members of professional communities use social media. Our qualitative analysis identified 11 thematic categories in academic librarians' Twitter discussions, focusing on professional topics. UK librarians exhibited a higher frequency of labor‐ and employment‐related terms compared to their US and Canadian counterparts. Sentiment ratios for US and Canadian tweets were similar, while the UK displayed nearly double the positive‐to‐negative tweet ratio. We also present a methodological intervention comparing two different sentiment analysis methods, VADER, and Zero‐Shot Learning (ZSL), to classify posts by academic librarians. ZSL significantly outperformed the off‐the‐shelf classifier, highlighting how accurate prediction is possible without annotated training data.