Summary Introduction Polymerase chain reaction (PCR)-based techniques to detect Mycobacterium tuberculosis DNA in respiratory specimens have been increasingly used to diagnose pulmonary tuberculosis. Their use in non-respiratory specimens to diagnose extrapulmonary tuberculosis is, however, controversial. In this study, we estimated the accuracy of three in-country commercialized PCR-based diagnostic techniques in pleural fluid samples for the diagnosis of pleural tuberculosis. Methods Patients underwent thoracenthesis for diagnosis purposes; pleural fluid aliquots were frozen and subsequently submitted to two real time PCR tests (COBAS®TAQMAN®MTB and Xpert®MTB/Rif) and one conventional PCR test (Detect-TB®). Two different reference standards were considered: probable tuberculosis (based on clinical grounds) and confirmed tuberculosis (bacteriologically or histologically). Results Ninety-three patients were included, of whom 65 with pleural tuberculosis, 35 of them confirmed. Sensitivities were 29% for COBAS®TAQMAN®MTB, 3% for Xpert®MTB/Rif and 3% for Detect-TB®; specificities were 86%, 100% and 97% respectively, considering confirmed tuberculosis. Considering all cases, sensitivities were 16%, 3% and 2%, and specificities, 86%, 100%, and 97%. Discussion Compared to the 95% sensitivity of adenosine deaminase, the most sensitive test for pleural tuberculosis, the sensitivities of the three PCR-based tests were very low. We conclude that at present, there is no major place for such tests in routine clinical use.