As local and global disturbances reshape African savannas, an understanding of how animal communities recover and respond to landscape features can inform conservation and restoration. Here, we explored the spatial ecology of a wildlife community in Gorongosa National Park, Mozambique, where conservation efforts have fostered the recovery of large mammal populations after their near‐extirpation during Mozambique’s civil war. We deployed a grid of 60 camera traps and used a hierarchical, multi‐species occupancy modeling approach to examine patterns of occupancy and its environmental and anthropogenic correlates for different functional groups and species. Our survey provides strong evidence that wildlife in Gorongosa is recovering. Throughout the study area, modeled species richness was comparable to richness in less‐disturbed savanna systems in Tanzania and Botswana, and exceeded estimates of richness from a mixed‐use landscape outside the park and from postwar (1997–2002) aerial surveys. However, the mammal community in Gorongosa differs from prewar conditions and from those of more intact systems, with few large carnivores, low occupancy probabilities for large ungulate species that were dominant prior to the war, and high occupancy for other ungulates that are now ubiquitous. Associations with tree cover varied among species and guilds. Contrary to our expectation, there was no effect of lake proximity on occupancy, given that previously dominant floodplain ungulate species now occupy more wooded areas. Mammals were more likely to occupy areas that burned frequently, as post‐fire vegetation regrowth provides high‐quality forage, highlighting the importance of Gorongosa’s fire regime. Occupancy was lower in areas with more illegal hunting, and higher closer to roads, potentially because roads were established in areas of high wildlife density and facilitate animal movement. Continued multi‐species monitoring in Gorongosa can shed light on the different recovery trajectories of ungulate species and the consequences of ongoing large carnivore restoration, guiding conservation interventions.

R script, JAGS model, and data used to estimate occupancy of mammals using camera trap records from Gorongosa National Park, Mozambique