The El Niño-Southern Oscillation (ENSO) is the dominant mode of interannual variability in the climate system, altering the global distribution of temperature and precipitation. These altered climate patterns impact remote regions, a relationship known as teleconnections. Due to the sensitivity of photosynthesis to changes in temperature, precipitation and radiation, ENSO teleconnections have large implications for the carbon cycle and the magnitude of carbon uptake by terrestrial vegetation at interannual timescales. We quantify the ENSO impact on gross primary productivity (GPP) and identify the physical mechanisms driving the terrestrial ENSO response for different regions around the world. Of the ten CMIP6 Earth System Models we analyzed, most models exhibit consistent global, spatial patterns of ENSO teleconnections with GPP, with a notable exception in Brazil, characterized by a reduced uptake of carbon by terrestrial vegetation during El Niño years. This reduced carbon uptake is primarily driven by a suppression of GPP in the Amazon Basin region due to reduced precipitation, increased surface temperatures, and increased surface downwelling shortwave radiation. In response to ENSO, the physical climate variations among models are more similar than the GPP response, suggesting an important role for the sensitivities of land processes to changes in physical climate within the ensemble. Three models, IPSL-CM6A-LR, MIROC-ES2L, and MPI-ESM1-2-LR, exhibit a weaker reduction, or an enhancement, in GPP in the Amazon Basin region during El Niño years compared to other models due to GPP exhibiting a positive dependence on radiation in these models. These results suggest general agreement among models on the ENSO impact on GPP in key regions, with differences being attributed to the land model response to ENSO-driven climate. Particularly, our results suggest that differing sensitivities of GPP to surface downwelling shortwave radiation in the tropics alter the magnitude of carbon uptake and thus have implications for the atmospheric CO₂ growth rates. Our results suggest that improved constraints on the sensitivity of tropical forests to variations in radiation may enable more robust model predictions of the impact of ENSO on carbon fluxes.