The quantitative real-time PCR (qRT-PCR) is an efficient and sensitive method for determining gene expression levels, but the accuracy of the results substantially depends on the stability of the reference gene (RG). Therefore, choosing an appropriate reference gene is a critical step in normalizing qRT-PCR data. Prunella vulgaris L. is a traditional Chinese medicine herb widely used in China. Its main medicinal part is the fruiting spike which is termed Spica Prunellae. However, thus far, few studies have been conducted on the mechanism of Spica Prunellae development. Meanwhile, no reliable RGs have been reported in P. vulgaris. The expression levels of 14 candidate RGs were analyzed in this study in various organs and at different stages of Spica Prunellae development. Four statistical algorithms (Delta Ct, BestKeeper, NormFinder, and geNorm) were utilized to identify the RGs’ stability, and an integrated stability rating was generated via the RefFinder website online. The final ranking results revealed that eIF-2 was the most stable RG, whereas VAB2 was the least suitable as an RG. Furthermore, eIF-2 + Histon3.3 was identified as the best RG combination in different periods and the total samples. Finally, the expressions of the PvTAT and Pv4CL2 genes related to the regulation of rosmarinic acid synthesis in different organs were used to verify the stable and unstable RGs. The stable RGs in P. vulgaris were originally identified and verified in this work. This achievement provides strong support for obtaining a reliable qPCR analysis and lays the foundation for in-depth research on the developmental mechanism of Spica Prunellae.