Conservation biologists, evolutionary ecologists and agricultural biologists require an improved understanding of how pollinators utilize space and share resources. Using microsatellite markers, we conducted a genetic analysis of space use and resource sharing at several spatial scales among workers of two ecologically dissimilar bumble bee species (Bombus terrestris and B. pascuorum) foraging in an urban landscape (London, UK). At fine scales, the relatedness of workers visiting small patches of flowers did not differ significantly from zero. Therefore, colonies shared flower patches randomly with other colonies, suggesting that worker scent-marks deterring visits to unrewarding flowers have not evolved as signals benefiting nestmates. To investigate space use at intermediate scales, we developed a program based on Thomas & Hill's maximum likelihood sibship reconstruction method to estimate the number of colonies utilizing single sites. The average number of colonies (95% confidence limits) sending workers to forage at sites of ≈ 1 ha in area was 96 colonies (84–118) in B. terrestris and 66 colonies (61–76) in B. pascuorum. These values are surprisingly high and suggested that workers travelled far from their colonies to visit the sites. At the landscape scale, there was little or no genetic differentiation between sites. We conclude that urban habitats support large bumble bee populations and are potentially valuable in terms of bumble bee conservation. In addition, bumble bee-mediated gene flow in plants is likely to occur over large distances and plant–bumble bee conservation requires landscape-scale action.