Spinocerebellar ataxia 3 (SCA3) is a genetic disorder resulting from the expansion of the CAG repeats in the ATXN3 gene. The pathogenesis of SCA3 is based on the toxic function of the mutant ataxin-3 protein, but the exact mechanism of the disease remains elusive. Various types of transgenic mouse models explore different aspects of SCA3 pathogenesis, but a knock-in humanized mouse has not yet been created. The initial aim of this study was to generate an ataxin-3 humanized mouse model using a knock-in strategy. The human cDNA for ataxin-3 containing 69 CAG repeats was cloned from SCA3 patient and introduced into the mouse ataxin-3 locus at exon 2, deleting it along with exon 3 and intron 2. Although the human transgene was inserted correctly, the resulting mice acquired the knock-out properties and did not express ataxin-3 protein in any analyzed tissues, as confirmed by western blot and immunohistochemistry. Analyses of RNA expression revealed that the entire locus consisting of human and mouse exons was expressed and alternatively spliced. We detected mRNA isoforms composed of exon 1 spliced with mouse exon 4 or with human exon 7. After applying 37 PCR cycles, we also detected a very low level of the correct exon 1/exon 2 isoform. Additionally, we confirmed by bioinformatic analysis that the structure and power of the splicing site between mouse intron 1 and human exon 2 (the targeted locus) was not changed compared with the native mouse locus. We hypothesized that these splicing aberrations result from the deletion of further splicing sites and the presence of a strong splicing site in exon 4, which was confirmed by bioinformatic analysis. In summary, we created a functional ataxin-3 knock-out mouse model that is viable and fertile and does not present a reduced life span. Our work provides new insights into the splicing characteristics of the Atxn3 gene and provides useful information for future attempts to create knock-in SCA3 models.