Including genotyped females in the reference population (RP) is an obvious way to increase RP but caution is needed because of potential preferential treatment of the genotyped cows and lower reliabilities of phenotypes compared with proven bulls. Denmark, Finland and Sweden have implemented a female genotyping project with voluntary genotyping of entire herds using low-density chip (LD project). The objective of the present study was to examine the effect of adding different sources of genotyped females to RP for Nordic Jersey. Five scenarios for building RP were considered: (1) bulls only; (2) bulls with females from LD project; (3) bulls with females from LD project plus non-LD project females genotyped before their first calving date; (4) bulls with females from LD project plus non-LD project females genotyped after their first calving date; (5) bulls with all genotyped females included. Genomically enhanced breeding values (GEBV) were predicted for 8 the traits in the Nordic Total Merit (NTM) index through a genomic BLUP (GBLUP) model using deregressed proofs (DRP). The validation population (VP) was formed by a cut-off using birth year of 2005 based on the genotyped bulls with DRPs. Average gain in reliability over the 8 traits ranged from 1.8% to 4.5% points compared with the scenario with only bulls in RP (scenario 1). Adding all the genotyped females in the RP achieved highest gain in reliability (scenario 5), followed by scenario 3, scenario 2 and scenario 4. The mean reliability of scenario 3 was 0.5% points higher than scenario 2 due to a slightly larger size of RP, and a decrease of 1.1% points in mean reliability were observed when including the extra 143 genotypes cows in scenario 4 compared with scenario 2. The mean reliabilities of scenario 2 and 3 were 1.6 and 1.1% points lower than of scenario 5. All scenarios led to inflated GEBVs since the regression coefficients are below 1. However, scenario 2 and scenario 3 led to less bias of genomic predictions than scenario 5 with the mean regression coefficients closer to 1. The results suggest adding unselected females in the RP significantly improve the reliabilities and tend to reduce the prediction bias compared with adding selectively genotyped females.