Syphilis diagnosis is based on clinical observation, serological analysis, and dark-field microscopy (DFM) detection of Treponema pallidum subsp. pallidum, the etiological agent of syphilis, in skin ulcers. We performed a nested PCR (nPCR) assay specifically amplifying the tpp47 gene of T. pallidum from swab and blood specimens. We studied a cohort of 294 patients with suspected syphilis and 35 healthy volunteers. Eighty-seven of the 294 patients had primary syphilis, 103 had secondary syphilis, 40 had latent syphilis, and 64 were found not to have syphilis. The T. pallidum nPCR results for swab specimens were highly concordant with syphilis diagnosis, with a sensitivity of 82% and a specificity of 95%. Reasonable agreement was observed between the results obtained with the nPCR and DFM methods (kappa = 0.53). No agreement was found between the nPCR detection of T. pallidum in blood and the diagnosis of syphilis, with sensitivities of 29, 18, 14.7, and 24% and specificities of 96, 92, 93, and 97% for peripheral blood mononuclear cell (PBMC), plasma, serum, and whole-blood fractions, respectively. HIV status did not affect the frequency of T. pallidum detection in any of the specimens tested. Swab specimens from mucosal or skin lesions seemed to be more useful than blood for the efficient detection of the T. pallidum genome and, thus, for the diagnosis of syphilis.