Carbon isotopes are commonly used in trophic ecology to estimate consumer diet composition. This estimation is complicated by the fact that lipids exhibit a more depleted carbon signature (δ13C) than other macromolecules, and are often found at different concentrations among individual organisms. Some researchers argue that lipids bias diet reconstructions using stable isotopes and should be accounted for prior to analysis in food web mixing models, whereas others contend that removing lipids may result in erroneous interpretations of the trophic interactions under study. To highlight this disagreement on best practices for applying δ13C in food web studies, we sampled the recent literature to determine the frequency and method of lipid-correction. We then quantified the potential magnitude and source of bias in mixing model results from a theoretical example and case study of diet reconstruction. The literature was split nearly evenly as to whether lipid-correction was applied to δ13C data in mixing model estimates of diet composition. Comparative mixing model scenarios demonstrated that lipid-correction can substantially alter the estimated diet composition and interpretation of consumer foraging habits. Given the lack of consensus on whether or not to lipid-correct prey and/or consumers, and the associated variation in mixing model results, we call for the establishment of a unified framework that will guide diet reconstruction in stable isotope ecology. Uncertainty in the prevalence of direct routing versus de novo synthesis of lipids across ecosystems, taxa, and trophic levels must be resolved to better guide treatment of lipids in isotope studies using carbon.