The application of complex network theory to community ecology has enabled quantification of interactions among large suites of species and clarified patterns of community structure across systems. Past analyses, however, have assumed that ecological networks are temporally static and persistent and spatially homogeneous, which could confound inference if species interactions vary over time and space. To evaluate temporal and spatial variation in mutualistic networks, therefore, we evaluated the consistency of a nectarivory/pollination network across years, by season, and over space. We tracked nectaring interactions among 37 butterfly and 58 flowering plant taxa during an 11‐yr period (2007–2017), across each summer and over a grassland landscape in Pennsylvania, USA. The composition of butterflies, plants, and their interactions varied markedly across years, months, and sites. Despite this compositional variation, one metric of network structure, nestedness, was invariant, with interactions much more nested than random across all years, months, and sites. Together with previous studies, this result suggests ecological interaction networks are generally more nested than expected by chance. Other measures of network structure were more variable, especially over time. Numbers of plants and interactions varied by year, month, and site. Connectance and numbers of butterflies varied annually and seasonally. Temporal variation in specialization was also evident for some species at an annual level and for the community across the season. We further found highly stable species were almost always generalists, while highly specialized species were almost always temporally and spatially variable, with few exceptions. Together, these results suggest communities are comprised of a reliable core of generalist species, accompanied by a changing suite of specialist species that, when participating in the network, primarily interact with the reliable core species. Our finding of nested mutualistic network centered on stable‐generalist species accompanied by a changing suite of sporadic specialists indicates dynamic changes in ecological communities vary with topological position. This finding further suggests that, even as rare species are highly threatened by species invasion, climate change, and other anthropogenic perturbations, network structure may be robust to species loss and compositional change from these perturbations.