This study investigates the determinants of mobile app usage patterns in a 4G network across 20 French cities, focusing on the relative influence of socioeconomic characteristics of temporary visitors and locals. Using the NetMob23 dataset and a range of spatial and socio-economic data sources, we apply random forest models to predict mobile traffic. Our results indicate that accessibility based variables, which reflect potential visitors, provide a better explanation of 4G data usage patterns compared to demographic characteristics of residents and the composition of land use. This suggests that mobile data traffic reflects temporary visitor behaviour rather than the socio-economic profile of residents, challenging conventional interpretations of 4G usage data in studies on urban socio-economic segregation and inequalities.