A primary goal of metabolomics is the characterization of a potentially very large number of metabolites that are part of complex mixtures. Application to biofluids and tissue samples offers insights into biochemical metabolic pathways and their role in health and disease. 1D 1H and 2D 13C‐1H HSQC NMR spectra are most commonly used for this purpose. They yield quantitative information about each proton of the mixture, but not which protons belong to the same molecule. Interpretation requires the use of NMR spectral databases naturally limited to known metabolites. Here, we present a new method that uses complementary ion exchange resin beads to differentially attenuate 2D NMR cross‐peaks that belong to different metabolites. Based on their characteristic attenuation patterns, cross‐peaks could be clustered and assigned to individual molecules, including unknown metabolites with multiple spin systems, as demonstrated for a metabolite model mixture and E. coli cell lysate.