The spatial arrangement of seagrass beds varies from scales of centimeters to meters (rhizomes, shoot groups), meters to tens of meters (patches), to tens of meters to kilometers (seagrass landscapes). In this study we examine the role of patch scale (patch size, seagrass % cover, seagrass biomass), landscape scale (fractal geometry, patch isolation) and wave exposure (mean wind velocity and exceedance) variables in influencing benthic community composition in seagrass beds at three intertidal sites in northern New Zealand (two sites in Manukau Harbour and one site in Whangapoua Harbour). Analysis of univariate community measures (numbers of individuals and species, species richness, diversity and evenness) and multivariate analyses indicated that there were significant differences in community composition inside and outside of seagrass patches at each of the three sites. Partialling out the spatial and temporal components of the ecological variation indicated that seagrass patch variables explained only 3–4% of the patch scale variation in benthic community composition at each of the sites. The temporal component was more important, explaining 12–14% of the variation. The unexplained variation was high (about 75%) at all three sites, indicating that other factors were influencing variation in community composition at the scale of the patches, or that there was a large amount of stochastic variation. Landscape and wave exposure variables explained 62.5% of the variation in the species abundance data, and the unexplained variation at the landscape level was correspondingly low (12%). Canonical correspondence analysis produced an ordination that suggests that, while mean wind velocity and exceedance were important in explaining the differences between the communities in the two harbours, spatial patterning of the habitat, primarily fractal dimension, and secondarily patch isolation (or some factors that were similarly correlated), were important in contributing to variability in community composition at the two sites in Manukau Harbour. This study suggests that spatial patterning of seagrass habitat at landscape scales, independent of the patch scale characteristics of the seagrass beds, can affect benthic community composition. Community composition inside and outside seagrass habitats involves responses to seagrass bed structure at a series of hierarchical levels, and we need to consider more than one spatial scale if we are to understand community dynamics in seagrass habitats.