Loss of native herbivores and introduction of livestock in many arid and semi-arid ecosystems around the world has shifted the competitive balance from herbaceous to woody plants, leading to biodiversity loss, reduced plant productivity, and soil erosion. To restore functions of these ecosystems, ecological replacements have been proposed as substitutes for extinct native herbivores. Here we predict how an ecological replacement giant tortoise population (Chelonoidis spp.) would interact with woody plants on Pinta Island in the Galápagos Archipelago, where a small group of replacement tortoises was introduced in 2010 to initiate restoration of the island's plant community. We developed an individual-based, spatially explicit simulation model that incorporated field-derived tortoise behavior and tortoise–plant interaction data to test whether tortoise introductions could lead to broad-scale changes in the plant community and, if so, at what tortoise densities. Tortoises reduced vegetation density in most (81%) 50-year-long simulations if the tortoise density was at least 0.7 per hectare, a value well below typical densities. In a smaller proportion of simulations (30%), tortoises increased local vegetation patchiness. Our results suggest that even moderate-density tortoise populations can reverse woody plant encroachment. Deployment of ecological replacement giant tortoises may therefore be a viable approach for restoring other arid and semi-arid ecosystems where a native herbivore that previously had strong interactions with the plant community has gone extinct.