Summary Food web structure is a fundamental feature of ecosystems. Stable isotopes (δ15N and δ13C) are used to estimate the relative contribution of food sources to consumer's diets (e.g. mixing models). In complex food webs, the use of δ15N and δ13C measurements cannot always solve trophic interactions and distinguish among aquatic organisms with different feeding habits if little intra‐ and interspecific isotopic differentiation occurs. We have developed a method to characterize trophic relationships using Bayesian stable isotope mixing models in combination with trace metal data as prior information. Trace metal information is useful because of the high correspondence between trace metal profiles in consumers and their food sources, as we show here in an example of concentrations of fish and their expected dietary items. Trace metal concentration allows a more accurate estimation of relative contributions of food sources to consumer species compared to estimates based only on stable isotope values. We show the improvement of the procedure using four freshwater fish species with well‐known feeding habits. The method provides a better estimation of the inter‐ and intraspecific dietary variability and correspondence with the feeding habits of these species. The approach described shows a considerable potential as a tool to assess trophic links in situations in which stable isotope methods are not conclusive. The method can be applied using other compounds that bioaccumulate in consumers (e.g. persistent organic pollutants).