Global dimming, a decadal decrease in incident global radiation, is often accompanied with an increase in the diffuse-radiation fraction, and therefore the impact of global dimming on crop production is hard to predict. A popular approach to quantify this impact is statistical analysis of historical climate and crop data, or use of dynamic crop simulation modelling approach. Here we show that statistical analysis of historical data did not provide plausible values for the effect of diffuse radiation versus direct radiation on rice or wheat yield. In contrast, our field experimental study of three years demonstrated a fertilization effect of increased diffuse-radiation fraction, which partly offset yield losses caused by decreased global radiation, in both crops. The fertilization effect was not attributed to any improved canopy light interception but mainly to the increased radiation use efficiency (RUE). The increased RUE was explained not only by the saturating shape of photosynthetic light-response curves but also by plant acclimation to dimming that gradually increased leaf nitrogen concentration. Crop harvest index slightly decreased under dimming, thereby discounting the fertilization effect on crop yields. These results challenge existing modelling paradigms, which assume that the fertilization effect on crop yields is mainly attributed to an improved light interception. Further studies on the physiological mechanism of plant acclimation are required to better quantify the global-dimming impact on agroecosystem productivity under future climate change. This article is protected by copyright. All rights reserved.