Dryland biodiversity is increasingly threatened by habitat loss, land use change, and climate extremes, particularly drought. The endangered eastern subspecies of the endemic Pink Cockatoo (Cacatua leadbeateri leadbeateri) is a rarely studied, iconic parrot inhabiting Australia's drylands. This study used passive acoustic monitoring and artificial intelligence to investigate the spatial and temporal habitat use of Pink Cockatoos across three dryland ecosystems in eastern Australia. A total of 91,646 h of recordings, collected from an array of 26 sensors for the spatial analysis and eight sensors for the temporal analysis, were processed using a convolutional neural network recogniser, with detections related to environmental variables including rainfall, temperature, vegetation greenness, and distance to water. Detections were highest in riparian habitats, with proximity to permanent water and River Red Gum +/- Coolabah vegetation emerging as key predictors. Tree diameter at breast height was also significant, suggesting a reliance on mature hollows for nesting. Detection rates increased during droughts and declined after rainfall, highlighting reliance on climate refugia and potential dispersal or population decline after drought. These findings demonstrate the importance of riparian zones and large trees for dryland specialists and show the value of ecoacoustic monitoring for conservation in remote regions. The study provides critical ecological insight into a declining species and informs future habitat protection and climate-adaptive conservation planning in arid Australia.