Effective planning for species conservation often requires an understanding of habitat use. The resources an animal selects within the landscape relate to its behavioral state and, therefore, incorporating behavior into habitat selection analyses can help inform management of threatened species. Here we present an approach for developing behavior-specific spatial habitat-use models using large quantities of GPS telemetry data. Using hidden Markov models, we first characterize 231,478 GPS fixes from 22 recently fledged endangered Tasmanian wedge-tailed eagles (Aquila audax fleayi) as reflective of either perching, short-distance flight, or long-distance flight. We then use a multivariate habitat selection ratio to develop spatial models predicting where these behavioral states occur. Recently fledged Tasmanian wedge-tailed eagles selected for areas close to forest edges during perching and short distance flights, whereas they selected more strongly for areas with steep topography (slopes > 15°) and further from forest edges for longer flights. Models using distance to forest edge and topographic slope effectively predicted where eagles engaged in long flights (R² > 0.91, r_s > 0.90) in each of six regions, whereas the performance varied by region for models describing perching (R² = 0.43–0.97, r_s = 0.80–0.97) and short flights (R² = 0.34–0.93, r_s = 0.63–1.00). Our study provides a detailed understanding of habitat use by young Tasmanian wedge-tailed eagles, which has multiple applications in the ongoing conservation of the population. Our method illustrates a framework for spatially explicit and behavior-specific habitat selection analyses that can be applied to other species of conservation concern.