Although urban areas are often considered to have uniformly negative effects on biodiversity, cities are most accurately characterized as heterogeneous mosaics of buildings, streets, parks, and gardens that include both ‘good’ and ‘bad’ areas for wildlife. However, to date, few studies have evaluated how human impacts vary in direction and magnitude across a heterogeneous urban landscape. In this study, we assessed the distribution of floral resources and flower-visiting insects across a variety of land uses in New York City. We visited both green spaces (e.g. parks, cemeteries) and heavily developed neighborhood blocks (e.g. with high or low density residential zoning) and used structural equation modeling (SEM) to evaluate the direct and indirect effects of median income, vegetation, and development intensity on floral resources and insects in both settings. Abundance and taxonomic richness of flower-visiting insects was significantly greater in green spaces than neighborhood blocks. The SEM results indicated that heavily-developed neighborhoods generally had fewer flower-visiting insects consistent with reductions in floral resources. However, some low-density residential neighborhoods maintained high levels of floral resources and flower-visiting insects. We found that the effects of surrounding vegetation on floral resources, and thus indirect effects on insects, varied considerably between green spaces and neighborhood blocks. Along neighborhood blocks, vegetation consisted of a mosaic of open gardens and sparsely distributed trees and had a positive indirect effect on flower-visiting insects. In contrast, vegetation in urban green spaces was associated with increased canopy cover and thus had a negative indirect effect on flower-visiting insects through reductions in floral resources. In both neighborhood blocks and green spaces, vegetation had a positive direct effect on flower-visiting insects independent of the influence of vegetation on floral resources. Our results demonstrate how inter-related components of an urban ecosystem can vary with respect to one another across a heterogeneous urban landscape, suggesting that it is inappropriate to generalize about urban systems as a whole without first addressing differences among component land use types.