The identification of significant habitats for highly mobile marine vertebrates is essential for their conservation. Evidence is often difficult to obtain for deep‐diving species such as sperm whales (Physeter macrocephalus), where standard visual survey methods are not sufficient to detect the species. Sperm whales rely on sound for most of their activities, so acoustics is a crucial tool to locate them in the environment and collect information about their daily life. We used a maximum entropy (MaxEnt) modeling approach to predict potential habitats for sperm whales during 2007–2015 in an area of the Mediterranean Sea (characterized by submarine canyon systems) where sperm whale singletons, social units of females and calves, and clusters with immature males, were regularly encountered in sympatry. Models to test species’ distribution and the potential differences between groups of varying composition and life stages were based on 3 independent variables (depth, slope, and Euclidean distance from the nearest coast) and a combination of presence‐only visual and acoustic data from boat‐based surveys. One variable (depth) was the strongest predictor in all encounters (pooled data) and clusters, whereas distance from coast and slope best predicted encounters with singletons and social units, respectively. The model predicted suitable locations in areas that were well‐known sperm whale habitat and in new regions of previously overlooked habitat, which possibly represent key areas for this endangered species in the Mediterranean. This study highlights that consideration should be taken regarding type of social aggregation when using modeling techniques for generating suitable habitat maps for conservation purposes. © 2018 The Wildlife Society.