Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) is a rapid, reliable and widely used method of studying gene expression profiles that requires appropriate normalization for accurate and reliable results. Reference genes are usually used to normalize mRNA levels; however, the expression levels of these reference genes may vary between cell types, developmental stages, species and experimental conditions. Therefore, a normalization strategy is an important precondition for reliable conclusions, with endogenous controls requiring determination for every experimental system. In the present study, 18 reference genes used in various prior studies were analyzed to determine their applicability in bladder cancer. A total of 35 matched malignant and non-malignant bladder cancer (specifically transitional cell carcinoma) tissue specimens were examined. RNA and cDNA quality was stringently controlled. Candidate reference genes were assessed using SYBR-Green RT-qPCR. mRNA abundance was compared and reference genes with distinct ranges of expression to possible target genes were excluded. Genes that were differentially expressed in matched non-cancerous and cancerous samples were also excluded, using quantification cycle analysis. Subsequently, the stability of the selected reference genes was analyzed using three different methods: geNorm, NormFinder and BestKeeper. The rarely used ribosomal protein S23 (RPS23) was the most stable single reference gene, with RPS23, tumor protein, translationally controlled 1 and RPS13 comprising the optimal reference gene set for all the bladder samples. These stable reference genes should be employed in normalization and quantification of transcript levels in future expression studies of bladder cancer-associated genes.