AbstractFood–web complexity often hinders disentangling functionally relevant aspects of food–web structure and its relationships to biodiversity. Here, we present a theoretical framework to evaluate food–web complexity in terms of biodiversity. Food network unfolding is a theoretical method to transform a complex food web into a linear food chain based on ecosystem processes. Based on this method, we can define three biodiversity indices, horizontal diversity (DH), vertical diversity (DV) and range diversity (DR), which are associated with the species diversity within each trophic level, diversity of trophic levels, and diversity in resource use, respectively. These indices are related to Shannon's diversity index (H′), whereH′ = DH + DV − DR. Application of the framework to three riverine macroinvertebrate communities revealed thatDindices, calculated from biomass and stable isotope features, captured well the anthropogenic, seasonal, or other within‐site changes in food–web structures that could not be captured withH′alone.