AbstractAimAgriculture is a key threat to biodiversity; however, its relationship with biodiversity patterns is understudied. Here, we evaluate how the extent, intensity and history of croplands relate to the global distribution of threatened mammals. We propose two hypotheses to explain these relationships: shelter, which predicts that threatened species concentrate in areas with low human land use; and threat, according to which threatened species should concentrate in areas of high human land use.LocationGlobal.Time periodc. 6000 BC – AD 2014.Major taxa studiedTerrestrial mammals.MethodsWe used boosted regression trees (BRT) that include spatial autocorrelation to investigate the relationship between the proportion of threatened terrestrial mammals [as defined by the International Union for Conservation of Nature (IUCN) Red List] and multiple metrics describing agricultural extent, intensity and history derived from remote sensing data and statistical projections. Data were analysed with a grain size of c. 110 km × 110 km at both global and biogeographical‐realm scales.ResultsAgricultural extent and intensity were the most relevant indicator types, with specific metrics important for each realm. Forest cover (extent) was identified as important in several regions. Tropical regions in early agricultural transition stages (e.g. frontier landscapes) were consistent with the shelter hypothesis, whereas patterns found for regions in later stages (e.g. intensified agricultural landscapes) were mostly found in temperate regions and agreed with the threat hypothesis.Main conclusionsThese results highlight the need to consider multiple land‐use indicators when addressing threats to biodiversity and to separately assess areas with divergent human and ecological histories in global‐scale studies. Different relationships associated with different agricultural transition stages suggest that high concentrations of threatened species may have contrasting meanings in different regions worldwide. We propose a new unifying hypothesis following a cyclic relationship along agricultural transition stages resulting in alternating negative and positive relationships between agriculture and threatened species richness.