Governments often engage with diasporas to promote homeland loyalty and political interests. How do home governments try to influence diaspora attitudes toward host countries? We develop a theory of diaspora-targeted propaganda that uses wedge narratives – both identity-based and political – to divide diaspora from host countries. Home governments frame racial discrimination and violence as targeting the diaspora, and frame alternative political systems as unstable and inferior. We test our theory in the important case of China. Scraping content from the dominant Chinese social media platform, WeChat, we measure propaganda framing with an unsupervised machine learning methodology, word embeddings. We find evidence in support of wedge narratives. Relative to private accounts, government accounts amplify coverage of anti-Asian racism and hate crimes in the United States, and are more likely to portray democracies as chaotic and corrupt. These findings shed light on how diaspora propaganda strategies can exploit host country divisions.