Reverse transcription–quantitative polymerase chain reaction (RT‐qPCR) has emerged as an important technique for gene‐expression analysis. However, for accurate and reliable results, the data normalization using appropriated reference genes is critical, and a systematic validation of reference genes in cowpea (Vigna unguiculata L. Walp), a high stress‐tolerant leguminous, has not been performed. To provide suitable reference genes in this strategic leguminous species under drought stress, we evaluated the expression stability of eight candidate genes using geNorm and NormFinder algorithms. The analyses were performed in nodules and leaves separately (organ‐specific analysis) or grouping the two organs (global analysis). Our results showed VuPp2A and VuUbq28 as the best reference genes—by both algorithms—for global analysis. For organ‐specific analyses, geNorm identified VuPp2A/VuYls8 as the most stable genes for nodules and NormFinder identified VuPolyP/VuPp2A as the most stable ones for leaves. Moreover, VuUbq28 was always identified among the top three genes. The proposed reference genes were validated by two stress‐responsive genes (VusHsp17.7 and VuNced1). In conclusion, VuPp2A was the best reference gene in all analyses, followed by VuUbq28, VuYls8, and VuPolyP as the most stable genes for RT‐qPCR data normalization in cowpea under drought stress. These genes could be a good starting point for reference‐gene validation in other leguminous species under stress and BNF conditions.