Accurate quantification of quantitative PCR (qPCR) data requires a set of stable reference genes (RGs) for normalisation. Despite its importance to mechanistic studies, no evaluation of RG stability has been conducted for pregnant human myometrium. A systematic search of the literature was performed to identify the most used RGs in human myometrial gene expression studies. The stability of these genes, and others, was then evaluated using geNorm and NormFinder algorithms, in samples of myometrium from singleton or twin pregnancies (n = 7 per group) delivering at term or preterm. The most frequently cited RGs were GAPDH, ACTB, B2M and 18s. There was strong agreement between algorithms on the most and least stable genes: Both indicated CYC1, YWHAZ and ATP5B were the most stably expressed. Despite being some of the most used RGs, B2M, 18s and ACTB expression was least stable and was too variable for use as accurate normalisation factors. Pairwise variation analysis determined that the optimal number of RGs for accurate normalisation is two. Validation of the choice of RGs by comparing relative expression of oxytocin receptors (OXTR) using the least stable 18s and B2M, with the most stable, CYC1 and YWHAZ, erroneously demonstrated significantly increased OXTR expression in myometrium in singleton pregnancies compared to twins. This study demonstrates the importance of appropriate RG selection for accurate quantification of relative expression in pregnant human myometrium qPCR studies. For normalisation, the geometric mean of CYC1 and YWHAZ or ATP5B is suggested. The use of ACTB, 18s and B2M, is not recommended.