Leptospirosis is still one of the most important health problems in developing countries located in humid tropical and subtropical regions. Human infections are generally caused by exposure to water, soil or food contaminated with the urine of infected wild and domestic animals such as rodents and dogs. The clinical course of leptospirosis is variable and may be difficult to distinguish from many other infectious diseases. The dark-field microscopy (DFM), serology and nucleic acid amplification techniques are used to diagnose leptospirosis, however, a distinctive standard reference method is still lacking. Therefore, in this study, we aimed to determine the presence of Leptospira spp., to differentiate the pathogenic L. interrogans and the non-pathogenic L. biflexa, and also to determine the sensitivity and specificity values of molecular methods as an alternative to conventional ones. A total of 133 serum samples, from 47 humans and 86 cattle were evaluated by two conventional tests: the Microagglutination Test (MAT) and the DFM, as well as three molecular methods, the 16S rRNA-PCR followed by Restriction Fragment Lenght Polymorphism (RFLP) of the amplification products 16S rRNA-PCR-RFLP, LipL32-PCR and OmpL1-PCR. In this study, for L. interrogans, the specificity and sensitivity rates of the 16S rRNA-PCR and the LipL32-PCR were considered similar (100% versus 98.25% and 100% versus 98.68%, respectively). The OmpL1-PCR was able to classify L. interrogans into two intergroups, but this PCR was less sensitive (87.01%) than the other two PCR methods. The 16S rRNA-PCR-RFLP could detect L. biflexa DNA, but LipL32-PCR and OmpL1-PCR could not. The 16S rRNAPCR- RFLP provided an early and accurate diagnosis and was able to distinguish pathogenic and non-pathogenic Leptospira species, hence it may be used as an alternative method to the conventional gold standard techniques for the rapid disgnosis of leptospirosis.