Population declines of once-abundant species have often preceded understanding of their roles within ecosystems. Consequently, important drivers of environmental change may remain undiagnosed because we simply do not know how species that are now rare or extinct shaped ecosystems in the past. Australia's desert rodents are thought to have little numerical impact on seed fate and vegetation recruitment when compared with ants or with desert rodents on other continents. However most research on granivory by Australian desert rodents has occurred in areas where rodents were rare or functionally extinct. Here we ask if the paradigm that rodents are relatively un-important granivores in Australian deserts is an artefact of their historical decline. In the Strzelecki Desert, the endangered rodent, Notomys fuscus is rare where introduced mesopredators are abundant but common where dingoes (an apex predator) suppress mesopredator populations. We used foraging trays to compare rates of seed removal for a common shrub (Dodonaea viscosa angustissima hopbush) between areas where N. fuscus, hopbush shrubs and their seedlings were rare and common and found that seed removal was consistently higher where rodents were common and hopbush rare. By excluding ants and rodents from foraging trays we show that ants removed more seeds than rodents where rodents were rare but rodents removed far more seeds than ants where rodents were common. By manipulating rodents’ access to the soil seed-bank we show that hopbush seeds persisted in greater numbers where rodents were excluded than where they had access. Our results support the hypothesis that granivory by rodents may once have been a far more important process influencing the fate of seeds and shaping plant communities in arid Australia and suggest that dingo extirpation has cascading effects on shrub seeds. Our study highlights that functional extinction of rodents may be an under-appreciated driver of vegetation change.