Background: The relationship between the conceptus and the maternal uterine environment is crucial for the successful establishment and maintenance of pregnancy in cattle. Gene expression analysis of the conceptus and maternal reproductive tissues is a favorable method to assess the embryonic maternal interaction. The reliability of the commonly used method reverse transcription-quantitative polymerase chain reaction (RT-qPCR) depends on proper normalization to stable reference genes (RGs). The objective of this study was to determine the expression stability of ten potential RGs (SUZ12, CNOT11, ACTB, RPL19, RPS9, GAPDH, TBP, HPRT1, SDHA and PPIA) in maternal reproductive tissues and fetal tissues, and to analyze the effect of RG selection on the calculation of the relative expression of target genes. Results: The expression stability of ten potential RGs was analyzed in eight different tissues (caruncular endometrium, intercaruncular endometrium, corpus luteum, ovary, oviduct, mammary gland, embryonic disc and trophoblast) from three pregnant dairy cows. Three programs—GeNorm, NormFinder and Bestkeeper—were used to identify the best RGs. According to all three programs, the most stable RG was CNOT11, whereas the least stable RGs were GAPDH and HPRT1. GeNorm analysis showed that a combination of five RGs (SDHA, PPIA, CNOT11, RPS9 and RPL19) was necessary for appropriate data normalization. However, NormFinder analysis indicated that the combination of CNOT11 and PPIA was the most suitable. When target genes were normalized to these RGs, the relative expression of the Radical S-adenosyl methionine domain containing 2 gene was not affected by the choice of RGs, whereas a large difference was observed in the expression profile of the Nuclear erythroid2-related factor 2 gene between the most stable RGs and least stable RGs. Conclusions: The results indicate that careful selection of RGs is crucial under different conditions, especially for target genes with relatively small fold changes. Furthermore, the results provide useful information for the selection of RGs for evaluating genes affecting bovine reproduction.