Adenosquamous carcinoma (ASC) is a relatively uncommon histological subtype of cervical cancer (CC). A point of controversy is the relative prognosis of ASC compared to squamous cell carcinoma (SCC). We hypothesized that ASC could be classified into two intrinsic molecular subtypes with different outcomes. We examined 143 biopsy samples of CC patients to identify a molecule for classification using microarray expression analysis and immunohistochemical analysis (IHA). We found specific expression profiles of candidate genes that distinguished two clusters. All adenocarcinoma (AC) patients were classified into one cluster, and most SCC patients fell into the other cluster. ASC patients were classified across the two clusters, which showed significantly different prognoses. The SCC-like expression signature comprised ANXA8, CK5, IFI16, and nectin-1; and the AC-like signature comprised EpCAM, and TMEM98. These signature-specific genes hypothetically indicated specific pathways by ontological analysis. The AC-like signature suggested an epithelial-mesenchymal transition and activated β-catenin pathway, while the SCC-like signature suggested keratinocyte differentiation, HPV infection, and p53-mediated apoptosis. IHA revealed that positive expression of the most promising protein, EpCAM, was significantly associated with poor prognosis. In addition, the inhibition of EpCAM expression using siRNA significantly increased radiation-induced cell death in the cervical cell line, ME-180. In conclusion, we identified two possible ASC subtypes associated with different expression profiles and different prognoses. This work provides a novel set of genes that could be used as independent prognostic markers and therapy targets.