The satellite‐derived Normalized Difference Vegetation Index (NDVI) is commonly used by researchers and managers to represent ungulate forage conditions in landscapes across the globe, despite limited information about how it compares to empirical measurements of forage quality and quantity. The application of NDVI as a forage metric is particularly appealing for studying migratory caribou (Rangifer tarandus) in remote Arctic ecosystems, where field assessments are logistically and financially prohibitive, and climate‐mediated changes in vegetation have been hypothesized to influence population declines. To determine the utility of NDVI for adequately representing caribou forage conditions, we compared NDVI derived from Moderate Resolution Imaging Spectroradiometer (MODIS) satellite imagery to empirical measures of caribou forage biomass, nitrogen, digestible nitrogen, and digestible energy within the summer range of the Central Arctic Caribou Herd on the North Slope of Alaska. Specifically, we determined the strength of forage–NDVI relationships at the start of the growing season and across the summer, assessed the efficacy of NDVI variables for modeling spatiotemporal variation in field measurements of different forage components, and used long‐term MODIS data to estimate temporal changes in forage between 2000 and 2016. We found that NDVI values were weakly correlated with caribou forage quality at the start of the growing season and throughout the summer. Although linear models of forage–NDVI relationships performed poorly, NDVI variables (NDVI and the number of days from when NDVI reached its maximum value) were useful for modeling spatiotemporal variation in empirical measurements of forage components across the growing season, but only when we incorporated nonlinear forage–NDVI relationships and other habitat covariates. Phenological advances in the date of peak NDVI were associated with significant changes in forage conditions, particularly nitrogen, which exhibited earlier seasonal declines. Using long‐term MODIS data, predicted values of forage nitrogen declined between 2000 and 2016, driven by exceedingly low values in 2014 and 2015. Given our results, we caution the application of NDVI as a general (linear) proxy of caribou forage conditions across the growing season, and encourage practitioners to use NDVI variables to model spatiotemporal variation in specific forage conditions from empirical field data, accounting for nonlinear forage–NDVI relationships.