β-Lactoglobulin (β-LG) genotypes are associated with differences in bovine milk protein composition. Therefore, β-LG genotypes are of direct relevance for the dairy industry. In this study, we predicted β-lactoglobulin genotypes based on routinely recorded milk Fourier transform infrared spectra using 500 calibration samples. The results show that 76% of the cows carrying the β-LG AA genotype, 80% of the cows carrying the β-LG AB genotype, and 66% of the cows carrying the β-LG BB genotype were predicted correctly. Furthermore, the prediction of β-LG genotypes based on Fourier transform infrared spectra showed a repeatability of 0.85. We discuss how the combined use of predicted β-LG genotypes, pedigree information, and β-LG genotypes derived using other methods could lead to further improvement in the percentage of correctly predicted β-LG genotypes. The presented methodology is easy and inexpensive and could ultimately provide β-LG genotypes at the individual cow level.