Arid and semi-arid grassland ecosystems cover 41% of the global land surface but are currently being degraded by increases in aridity and changes in land use. As a consequence, ecosystem services, including carbon (C) and nitrogen (N) sequestration, are being reduced. In this study, we determined whether stable isotopic C (δ13C) and N (δ15N) values can be used to explain patterns of multiple ecosystem functions related to C and N cycling in arid and semi-arid grasslands at the regional scale. Ecosystem (plant and soil) δ13C and δ15N values and ecosystem functions related to C and N cycling were investigated along an aridity gradient on the Mongolian Plateau. The gradient covered four vegetation grassland types with distinct characteristics of aridity and soil. We found that ecosystem δ13C and δ15N values increased while ecosystem functions decreased as aridity increased. Structural equation modelling revealed that aridity directly and indirectly affected plant δ13C values by affecting plant C/N and indirectly affected soil δ13C values by affecting plant and soil C/N and plant δ13C. Aridity directly and indirectly affected soil δ15N values by affecting soil C/N and soil pH, and indirectly affected plant δ15N values by affecting plant and soil C/N and soil δ15N. As indicated by ecosystem δ13C and δ15N values, the negative effects of aridity on ecosystem functions were mainly derived from the direct and indirect effects of aridity on C cycling and N cycling, respectively. Our findings suggest that ecosystem δ13C and δ15N values can be used to explain patterns of multiple ecosystem functions related to C and N cycling in arid and semi-arid grasslands at the regional scale. Consideration of ecosystem δ13C and δ15N values in predictive models will greatly increase the ability to accurately simulate how ecosystem functions change in response to the rapid degradation of semi-arid grasslands.