Grasslands are an imperiled ecosystem, and grassland bird abundance is declining across North America. One of the strongest drivers for these declines is woody plant encroachment of grasslands. In the Great Plains and Sagebrush biomes of North America, spatial covariance—a remote-sensing metric for tracking boundaries between vegetation types—is emerging as a new method to identify and strategize conservation of grassland cores in the face of woody plant encroachment. However, the relationship between spatial covariance and grassland bird community occupancy is unknown. Here, we used Bayesian multispecies occupancy models to understand how occupancy probability of six declining grassland species responded to spatial covariance at three scales (0.81, 7.29, and 65.61 ha) and tree cover in fragmented grasslands of Arkansas, USA. Model selection revealed that the smallest spatial scale (0.81 ha) best explained grassland bird occupancy. Tree cover alone was a poor predictor of grassland bird occupancy compared to models that included spatial covariance at the 0.81- and 7.29-ha scales. Grassland bird occupancy declined at tree-grass boundaries (negative spatial covariance at the 0.81-ha scale) and increased in grassland cores (near-zero or slightly positive spatial covariance at the 0.81-ha scale). At low tree cover, Dickcissel (Spiza americana), Eastern Kingbird (Tyrannus tyrannus), Loggerhead Shrike (Lanius ludovicianus), Northern Bobwhite (Colinus virginianus), and Scissor-tailed Flycatcher (Tyrannus forficatus) occupancy probability more than doubled in grassland cores (where spatial covariance approached zero). Eastern Meadowlark (Sturnella magna) had the weakest relationship with spatial covariance. Our results suggest that spatial covariance can identify grassland cores and serve as a powerful predictor of grassland bird community occupancy, even in highly fragmented grasslands. Identifying grassland cores empowers defending core grasslands from woody plant encroachment and then growing cores via active restoration.