Reference genes are often used to normalize expression of data from real-time quantitative reverse transcription-polymerase chain reaction (RT-qPCR), and only a validation of their stability during a given experimental paradigm leads to reliable interpretations. The present study was thus designed to validate potential reference genes in a mouse model of mesiotemporal lobe epilepsy (MTLE) with focal seizures after unilateral intrahippocampal injection of kainate (KA). Ipsilateral and contralateral hippocampi were removed during nonconvulsive status epilepticus (5 hr), epileptogenesis (7 days), and the chronic period of recurrent focal seizures (21 days). Naive animals were equally studied. The stability of eight potential reference genes (hypoxanthine phosphoribosyltransferase, Hprt1; peptidylprolyl isomerase A, Ppia; TATA box binding protein, Tbp; beta-actin, Actb; acidic ribosomal phosphoprotein P0, Arbp; glyceraldehyde-3-phosphate dehydrogenase, Gapdh; ribosomal RNA 18S, 18S rRNA; and glucuronidase beta, Gusb) were determined using geNorm and NormFinder software. The first five (Hprt1, Ppia, Tbp, Actb, and Arbp) were found to be stable across the different phases of the disease and appeared adequate for normalizing RT-qPCR data in this model. This was in contrast to the other three (18S rRNA, Gapdh, and Gusb), which showed unstable expressions and should be avoided. The analysis of KA-induced changes in the expression of glial fibrillary acidic protein (Gfap) gene resulted in various relative expressions or even a completely different pattern when unstable reference genes were used. These results highlight the absolute need to validate the reference genes for a correct interpretation of mRNA quantification.