Blooms of toxic microalgae and the accumulation of their toxic compounds in bivalve molluscs are a regular worldwide problem to both consumers and producers. To select populations that accumulate a less amount of toxins is very important to understand the metabolism of biotoxins. The study of gene expression patterns by real-time quantitative polymerase chain reaction (RT-qPCR) may provide insights into the genes involved in detoxification metabolic pathways. One of the critical steps using RT-qPCR is that the expression results have to be normalized using internal reference genes. A systematic evaluation of reference genes in mussels has not been done yet and genes are frequently used as reference genes without validation. In this study, eight commonly used candidate reference genes have been tested as suitable in the mussel Mytilus galloprovincialis and their expression in mussels exposed to a toxic tide was studied by RT-qPCR. Their expression stabilities were evaluated using three different Excel applets (geNorm, NormFinder, and BestKeeper) which produced highly comparable results. The most suitable combination of reference genes for the normalization was glyceraldehyde-3-phosphate dehydrogenase (gapdh), 40S ribosomal protein S4 (rps4), and cytochrome c oxidase subunit 1 in digestive gland and gill and gapdh, rps4, and 40S ribosomal protein S27 in mantle. The M. galloprovincialis mrp1 and mrp2 genes were used to assess the quality of these reference genes. Our results show that some of the most widely used reference genes are not always suitable. This work underlies the importance of the validation of reference genes for each experimental situation and will be useful for the identification of genes involved in detoxification pathways in M. galloprovincialis.