Reverse transcription real-time quantitative PCR (RT-qPCR) is the predominant method of choice for the quantification of mRNA transcripts of a selected gene of interest. Here reference genes are commonly used to normalize non-biological variation in mRNA levels and their appropriate selection is therefore essential for the accurate interpretation the collected data. In recent years the use of multiple validated references genes has been shown to substantially increase the robustness of the normalization. It is therefore considered good practice to experimentally validate putative reference genes under specific experimental conditions, determine the optimal number of reference genes to be employed, and report the method or methods used. Under this premise, we assessed the current state of reference gene base normalization in RT-qPCR bivalve ecotoxicology studies (post 2011), employing a systematic quantitative literature review. A total of 52 papers published met our criteria and were analysed for the gene or genes used, whether they employed multiple reference genes, as well as the validation method employed. In addition we performed a case study using primary hemocytes from the marine bivalve Ruditapes philippinarum after in vitro copper exposure. Herein we further critically discuss methods for reference gene validation, including the established algorithms geNorm, NormFinder and BestKeeper, as well as the popular online tool RefFinder. We identified that RT-qPCR normalization in bivalve ecotoxicology studies is largely performed using single reference genes, while less than 40% of the studies attempted to experimentally validate the expression stability of the reference genes used. 18s rRNA and β-Actin were the most popular genes, yet their un-validated use did introduce artefactual variance that altered the interpretation of the resulting data, while the use of appropriately validated reference genes did substantially improve normalization. Our findings further suggest that combining the results from multiple individual algorithms and calculating the overall best-ranked gene, as e.g. computed by the RefFinder tool, does not by default lead to the identification of the most suitable reference gene or combination of reference genes.