Abstract Water diversion and pollution are two pervasive stressors for river ecosystems that often co‐occur. The individual effects of both stressors on river communities and energy transfer across the food webs are well described; however, how they interact remains unknown. We hypothesised that low‐to‐moderate nutrient pollution gradient would cause a mild increase in invertebrate driven herbivory and water diversion a strong reduction in detritivory, whereas their joint effect would reduce invertebrate abundance and diversity, as well as total energy fluxes (from basal resources to invertebrates and fish). We also expected a shift in body size spectra slope with increased energy transfer between trophic levels with moderate pollution, but not with water diversion. To test these hypotheses, we selected four rivers across a range of nutrient pollution levels (a proxy of water quality) subject to similar water diversion schemes and compared food webs upstream and downstream of their diversion weirs. Both stressors changed the availability of basal food resources. Nutrient pollution induced changes in the green food web (i.e., biofilm‐based) by enhancing biofilm stocks, whereas water diversion affected the brown food web (i.e., detritus‐based) by decreasing stocks of detritus. The propagation of the effects to higher trophic levels differed with each stressor: pollution increased the homogeneity of community within the reach, whereas water diversion made communities more heterogeneous. Moreover, pollution induced changes within omnivores, increasing herbivory and carnivory, whereas diversion reduced the total energy fluxes through a decrease in detritivory, especially with pollution. Although most of the variables studied seemed to be more sensitive to water diversion, pollution often accentuated the response, being the interaction between both stressors more explanatory than any of the two stressors on its own. The effects of water diversion on diversity and energy flow through food webs are more detrimental to moderately polluted rivers than to systems with high quality water. We show that the use of tools merging knowledge on trophic relationships among species and their metabolic requirements enables disturbances to be detected that would otherwise go unnoticed.