This study examines how local governments use social media during crises and how citizens respond to their communication. Focusing on the COVID-19 pandemic, we analyse over 50,000 social media posts and 700,000 citizen reactions from 127 German local governments. Drawing on expectancy – disconfirmation theory and emotions as social information theory, we explore how different message framings – analytical vs. emotional – affect citizen responses: approval (likes), diffusion (retweets), and engagement (replies). Our findings show that analytical framing increases all three response types. Emotional framing, however, yields mixed effects: while positive emotions increase approval, they reduce diffusion and engagement. Negative emotions show the opposite pattern. These results reveal trade-offs in framing strategies and offer insights into how public administrations can communicate more effectively during crises. The study contributes to public administration and crisis communication research by linking specific communication strategies to distinct citizen behaviours.