The distribution of the bilby Macrotis lagotis was assessed in the Tanami Desert using stratified random plots, repetitively sampled transects, aerial survey transects, and ground truth plots. Compared to a previous assessment of distribution, the extent of occurrence has changed little in the last 20 yr. However, the area of occupancy is small relative to the extent of occurrence and <25% of the current geographic range has bilby sign <20 km apart. Generalised linear modelling was used to determine the strength of association between bilby occurrence and habitat variables and identify refugia characteristics. Four competing candidate models were examined to determine whether bilby occurrence associated significantly with productive substrates and introduced herbivores, the distribution of key predator species, the pattern of fire, and climatic gradients including rainfall and temperature. For the entire study area, bilby presence associated most strongly with variables of mean annual rainfall, substrate type and the probability of dingo occurrence. Proximity to recently burnt habitat formed a significant predictor of bilby occurrence in a model derived for a reduced part of the study area where most sign was found. The work suggested that the current frameworks underpinning understanding of biotic distributions in arid Australia are deficient, and that climatic gradients, lateritic and rocky systems, and predators need to be incorporated into our thinking in the future. The extent of occurrence based on outlier records from opportunistic reports provided a misleading indication of the true status of the bilby.