Comparing bacterial 16S rDNA sequences to GenBank and other large public databases via BLAST often provides results of little use for identification and taxonomic assignment of the organisms of interest. The human microbiome, and in particular the oral microbiome, includes many taxa, and accurate identification of sequence data is essential for studies of these communities. For this purpose, a phylogenetically curated 16S rDNA database of the core oral microbiome, CORE, was developed. The goal was to include a comprehensive and minimally redundant representation of the bacteria that regularly reside in the human oral cavity with computationally robust classification at the level of species and genus. Clades of cultivated and uncultivated taxa were formed based on sequence analyses using multiple criteria, including maximum-likelihood-based topology and bootstrap support, genetic distance, and previous naming. A number of classification inconsistencies for previously named species, especially at the level of genus, were resolved. The performance of the CORE database for identifying clinical sequences was compared to that of three publicly available databases, GenBank nr/nt, RDP and HOMD, using a set of sequencing reads that had not been used in creation of the database. CORE offered improved performance compared to other public databases for identification of human oral bacterial 16S sequences by a number of criteria. In addition, the CORE database and phylogenetic tree provide a framework for measures of community divergence, and the focused size of the database offers advantages of efficiency for BLAST searching of large datasets. The CORE database is available as a searchable interface and for download at http://microbiome.osu.edu.