Background: Empirically assessing the impact of preselection on subsequent genetic evaluations of preselected animals requires comparison of scenarios with and without preselection. However, preselection almost always takes place in animal breeding programs, so it is difficult, if not impossible, to have a dataset without preselection. Hence most studies on preselection used simulated datasets, concluding that subsequent genomic estimated breeding values (GEBV) from single-step genomic best linear unbiased prediction (ssGBLUP) are unbiased. The aim of this study was to investigate the impact of genomic preselection, using real data, on accuracy and bias of GEBV of validation animals. Methods: We used data on four pig production traits from one sire-line and one dam-line, with more intense original preselection in the dam-line than in the sire-line. The traits are average daily gain during performance testing, average daily gain throughout life, backfat, and loin depth. Per line, we ran ssGBLUP with the entire data until validation generation and considered this scenario as the reference scenario. We then implemented two scenarios with additional layers of genomic preselection by removing all animals without progeny either i) only in the validation generation, or ii) in all generations. In computing accuracy and bias, we compared GEBV against progeny yield deviation of validation animals. Results: Results showed only a limited loss in accuracy due to the additional layers of genomic preselection. This is true in both lines, for all traits, and regardless of whether validation animals had records or not. Bias too was largely absent, and did not differ greatly among corresponding scenarios with or without additional layers of genomic preselection. Conclusion: We concluded that impact of recent and/or historical genomic preselection is minimal on subsequent genetic evaluations of selection candidates, if these subsequent genetic evaluations are done using ssGBLUP.