The golden-cheeked warbler (Dendroica chrysoparia) is a federally listed endangered species with an extremely limited breeding range (Edwards Plateau, Tex., USA) and specific habitat requirements. Although localized habitat composition and structure correlate well with golden-cheeked warbler occurrence and nest location, the effects of landscape composition and structure are largely untested. To investigate landscape-scale influences on habitat occupancy, we surveyed 202 point locations for golden-cheeked warbler presence. We used logistic regression to compare the relative influence of 4 landscape-scale variables (largest patch index, percent of the landscape in woodland, mean-nearest neighbor, and edge density) measured at 4 spatial scales (3 ha, 12 ha, 50 ha, and 200 ha) in predicting occurrence. The amount of mature juniper–oak (Juniperus ashei and Quercus spp.) woodland in the landscape was the single most important variable across spatial scales and average values of woodland cover remained stable across scales. Golden-cheeked warblers occurred in a local habitat only when landscape composition exceeded 40% woodland and the likelihood of occurrence was greater than 0.5 only when landscape composition exceeded 80% woodland. By contrast, common measures of habitat fragmentation, including edge density and mean-nearest neighbor, were poor predictors of species occurrence across all spatial scales. Our results suggest that golden-cheeked warblers may be less sensitive to fragmentation then they are to overall habitat loss, which may be due to natural habitat heterogeneity related to the variable terrain. Our results do not preclude a relationship between habitat fragmentation and overall habitat quality as measured by nesting success, survival, and recruitment. While our results hold promise for increasing the reliability of predicting species occurrence, future studies are needed to determine the influence of landscape composition and structure on the ultimate measures of habitat quality.