In the genomic era, new techniques and criteria are proposed to improve the traditionally phenotypic and biochemical test–based approaches for prokaryotic species definition. Among them, average nucleotide identity (ANI) mirrors DNA-DNA hybridization and is widely used by the microbial research community. However, our test shows that ANI possibly defines distinct taxa as the same species when they shared highly homologous sequences in a very short genomic region. In this study, we propose an improved algorithm named total nucleotide identity (TNI) for use in bacterial taxonomy; this algorithm provided higher accuracy for species classification than ANI. Furthermore, we developed a species identification system for prokaryotes (SISP) based on pairwise TNI of 3,073 genomes acquired from GenBank. For a submitted query genome, SISP can quickly find its most closely related genome from the established database based on the TNI calculation and infer the possible species of the query genome. Given a criterion of TNI > 70%, SISP has an accuracy that was above 90% for 3,596 prokaryotic genomes. SISP is open source and is available at https://github.com/chjp/SISProkaryotes.