Objective: To evaluate the diagnostic accuracy of bronchoscopy in patients with clinical or radiological suspicion of tuberculosis who were unable to produce sputum or with negative sputum smear microscopy results. Methods: A prospective cross-sectional study involving 286 patients under clinical or radiological suspicion of having pulmonary tuberculosis and submitted to bronchoscopy-BAL and transbronchial biopsy (TBB). The BAL specimens were submitted to direct testing and culture for AFB and fungi, whereas the TBB specimens were submitted to histopathological examination. Results: Of the 286 patients studied, 225 (79%) were diagnosed on the basis of bronchoscopic findings, as follows: pulmonary tuberculosis, in 127 (44%); nonspecific chronic inflammation, in 51 (18%); pneumocystis, fungal infections, or nocardiosis, in 20 (7%); bronchiolitis obliterans organizing pneumonia, alveolites, or pneumoconiosis, in 14 (5%); lung or metastatic neoplasms, in 7 (2%); and nontuberculous mycobacterium infections, in 6 (2%). For the diagnosis of tuberculosis, BAL showed a sensitivity and a specificity of 60% and 100%, respectively. Adding the TBB findings significantly increased this sensitivity (to 84%), as did adding the post-bronchoscopy sputum smear microscopy results (total sensitivity, 94%). Minor post-procedure complications occurred in 5.6% of the cases. Conclusions: Bronchoscopy is a reliable method for the diagnosis of pulmonary tuberculosis, with low complication rates. The combination of TBB and BAL increases the sensitivity of the method and facilitates the differential diagnosis with other diseases.