To help identify an epidemiological link between the consumption of buffalo mozzarella prepared with raw milk (not heat-treated) and cases of human brucellosis, 80 samples of raw buffalo milk and 315 samples of mozzarella were tested for the presence of Brucella spp. Samples originated from Caserta, the province with the highest number of Brucella-positive buffalo herds in Campania, the region in which 96.02% of all cases of human brucellosis reported in 2000-2005 were recorded. To take into account possible seasonal variations, between February 2006 and March 2007, samples were purchased directly from 72 dairy outlets that were representative of the province. Samples were tested for Brucella spp. using the polymerase chain reaction (PCR) and bacterial isolation. Samples tested negative for Brucella using both methods. Spiked samples were then prepared and tested by PCR and bacterial isolation and the sensitivity, specificity, repeatability, reproducibility and limit of detection were determined. The PCR limit of detection was below 1 colony-forming unit (cfu)/g. The repeatability and reproducibility of the method were 100% (p = 0.95), the sensitivity was 96.7% (p = 0.95) and the specificity was 100% (p = 0.95).