•A model of local and landscape disturbance effects on seed dispersal is proposed.•The model is tested regionally with a keystone semiarid scrub dispersed by mammals.•Disturbance at each scale impacts on seed dispersal via distinct pathways.•Fruit intake is affected by remnant while disperser abundance by landscape features.•Attention to the scale of disturbance is needed to recover seed dispersal functions. The scale of disturbance is of paramount importance to understand the disruptive effect of anthropogenic perturbation on animal seed dispersal and its consequences for plant species conservation at regional level. However, the intricate ways by which landscape and local-scale disturbances affect seed dispersal remain unclear. We propose a conceptual scheme of the direct and indirect effects that large-scale (landscape) and local-scale (within-habitat remnant) disturbances may have on seed dispersal. We evaluate this scheme regionally with Ziziphus lotus, a mammalian-dispersed keystone scrub of threatened European semiarid habitats dispersed by mammals. Using Structural Equation Modeling we disentangled whether disturbance effects on seed dispersal function happen via landscape/remnant structural changes or by shortage in food provision (fruit and carnivores’ prey abundance) and if such changes cascade on seed dispersal by modifying primarily the fruit consumption rates or the animal-disperser assemblages. Disturbances at both spatial scales impacted on seed dispersal, which collapsed in some localities, especially those without juveniles. Rates of fruit consumption by seed dispersers increased with fruit availability that, in turn, depended on the level of alteration at the local scale of each remnant. However, the abundance and diversity of seed dispersers responded to structural changes in the landscape, like natural habitat cover and complexity. Attention to both scales of disturbance is thus needed to recover the seed dispersal function. Management within remnants to improve their state and enhance food provision will favor fruit consumption rates, while increasing natural habitat cover and landscape complexity will favor the disperser-assemblage abundance and diversity.