Cities generate gains from interaction, but citizens often experience segregation in their daily urban movements. Using GPS location data, we identify patterns of this experienced segregation across US cities, differentiating between neighborhoods that are sources and sinks—exporters and importers—of diversity. By clustering areas with similar mobility signatures, capturing both the diversity of visitors and the exposure of neighbourhoods to diversity, we uncover a generic mesoscopic structure: rings of isolation around cities and internal pockets of segregation. Using a decision tree, we identify the key predictors of isolation and segregation: race, wealth and geographic centrality. We show that these patterns are persistent across time and prevalent across all US cities, with a trend toward larger rings and stronger pockets after the pandemic. These findings offer insights into the dynamics that contribute to inequality between neighbourhoods, so that targeted interventions promoting economic opportunity can be developed.