Livestock effects on plant communities through overgrazing (desertification) should affect the structure and functioning of semarid rangeland communities. We measured plant, granivorous ant and rodent communities and rates of seed removal by rodents and ants in grazed (by livestock) and ungrazed desert grasslands as well as mesquite and creosotebush shrublands to test hypotheses on the effects of grazing and desertification on ecosystem structure and functioning. In desert grasslands grazing reduced the cover of perennial grasses, particularly the dominant Bouteloua eriopoda, but the cover of forbs and shrubs did not differ between treatments. One species of perennial grass, Dasyochloa pulchellum, increased in grazed grasslands compared with grassland exclosures. Detrended correspondence analysis showed that grazing caused desert grasslands to shift in community structure towards the shrublands. There were more seed harvesting ant and rodent species in the creosotebush shrublands than in the grasslands and mesquite shrublands. Grazing had no effect on the diversity of ants or rodents within grasslands, and detrended correspondence analysis revealed no clear trends in granivorous ant community structure in the grazed and ungrazed grasslands or the mesquite and creosotebush shrublands. Ants removed more seeds than did rodents in the grassland sites but rodents removed more seeds than did ants in the creosotebush sites and seed removal rates by rodents and ants were the same in the mesquite sites. Our data support the hypothesis that livestock grazing leads to a shift from grassland to shrubland in the Chihuahuan Desert, with associated changes in the structure and functioning of faunal communities. Because grasslands support few species and low densities of rodents, seed harvesting ants are the most important granivores in these desert grasslands. On a larger scale, we therefore hypothesize that the observed dominance of rodents as seed harvesters in the Chihuahuan desert is a function of the desertification of desert grasslands to shrublands by livestock, and that associated feedback effects may complicate the regeneration of degraded communities.