The study examines the use of language on social media by Ukrainian citizens before and during the Russian invasion of Ukraine, analyzing 4,453,341 geo-tagged tweets from 62,712 users from January 2020 to October 2022. The research uses statistical models to disentangle sample effects (user flux) from behavioral changes (user behavior). Key findings include: 1. **Language Shift Before the War**: A steady shift from Russian to Ukrainian language use was observed before the war, with the probability of tweeting in Ukrainian rising from 33% to 48%. 2. **Behavioral Changes During the War**: The shift accelerated with the outbreak of the war in February 2022. The probability of tweeting in Ukrainian increased to 76%, driven by both Russian users leaving and Ukrainian users joining, as well as a significant behavioral change (+249% in odds to tweet in Ukrainian). 3. **User Turnover**: User turnover played a role in the shift, with a decline in Russian tweets due to users leaving and an increase in Ukrainian tweets due to new users joining. 4. **Language Proportions**: A substantial number of users switched from Russian to Ukrainian, with over half of Russian-tweeting users switching to Ukrainian after the war began. 5. **User Characteristics**: Users who switched from Russian to Ukrainian had more followers, higher tweet and like frequencies, and published more geo-tagged tweets about Ukraine. 6. **Topic Modelling**: Multilingual topic modelling revealed that users who switched languages were more likely to discuss war-related topics, but this difference diminished when controlling for total tweet volume. The findings suggest that the use of Ukrainian language on social media reflects a conscious choice by users to express a more Ukrainian identity and self-definition, particularly in response to the Russian invasion.The study examines the use of language on social media by Ukrainian citizens before and during the Russian invasion of Ukraine, analyzing 4,453,341 geo-tagged tweets from 62,712 users from January 2020 to October 2022. The research uses statistical models to disentangle sample effects (user flux) from behavioral changes (user behavior). Key findings include: 1. **Language Shift Before the War**: A steady shift from Russian to Ukrainian language use was observed before the war, with the probability of tweeting in Ukrainian rising from 33% to 48%. 2. **Behavioral Changes During the War**: The shift accelerated with the outbreak of the war in February 2022. The probability of tweeting in Ukrainian increased to 76%, driven by both Russian users leaving and Ukrainian users joining, as well as a significant behavioral change (+249% in odds to tweet in Ukrainian). 3. **User Turnover**: User turnover played a role in the shift, with a decline in Russian tweets due to users leaving and an increase in Ukrainian tweets due to new users joining. 4. **Language Proportions**: A substantial number of users switched from Russian to Ukrainian, with over half of Russian-tweeting users switching to Ukrainian after the war began. 5. **User Characteristics**: Users who switched from Russian to Ukrainian had more followers, higher tweet and like frequencies, and published more geo-tagged tweets about Ukraine. 6. **Topic Modelling**: Multilingual topic modelling revealed that users who switched languages were more likely to discuss war-related topics, but this difference diminished when controlling for total tweet volume. The findings suggest that the use of Ukrainian language on social media reflects a conscious choice by users to express a more Ukrainian identity and self-definition, particularly in response to the Russian invasion.