Previous results from our laboratory have shown that optical sorter–based indirect selection reduced deoxynivalenol (DON) accumulation in soft red winter wheat (SRWW). In this paper we evaluate the efficacy of optical sorter–augmented genomic selection (OSA‐GS) for lowering DON accumulation at three selection intensities across 2 yr. In total, 758 SRWW breeding lines were genotyped and then phenotyped in an inoculated Fusarium head blight (FHB) nursery. Accumulation of DON was measured on all breeding lines. The proportion of Fusarium‐damaged kernels estimated using an optical sorter (FDKos) was measured on 120 lines; these data were used to train a genomic prediction model. Genomic estimated breeding values (GEBVs) for FDKos were computed for all lines without actual FDKos data. The top 20, 30, and 40% of lines without actual FDKos data were selected based on FDKos GEBVs. The same was done using actual measured DON values. Both strategies lowered DON, but traditional direct phenotypic selection based on actual DON values outperformed OSA‐GS. In other words, phenotypic selection was necessary to achieve the greatest reductions in DON. However, using previously published cost estimates for the price of an optical sorter, DON analysis, and genotyping, we determined that OSA‐GS required less financial investment than phenotypic selection based on measured DON. Taken together, our findings indicate that OSA‐GS is a cost‐effective method for lowering DON accumulation and support the usefulness of an optical sorter as a tool for FHB resistance breeding in SRWW.