Tuberculosis (TB) remains a global public health threat. Misdiagnosis and delayed therapy of sputum smear-negative TB can affect the treatment outcomes and promote pathogen transmission. The application of Xpert MTB/RIF assay in bronchoalveolar lavage fluid (BALF) has been recommended but needs clinical evidence. We carried out a prospective study in the Nanjing Public Health Medical Center from September 2018 to August 2019. Pulmonary tuberculosis (PTB) patients were enrolled in the study if they had negative results of sputum smear. We compared the performance of Xpert MTB/RIF assay in sputum and BALF using sputum culture as the reference. In addition to this, we applied parallel tests using sputum culture, sputum-based Xpert MTB/RIF assay and BALF-based Xpert MTB/RIF assay to jointly detect smear-negative PTB using clinical diagnosis as the reference. With mycobacterial culture as the reference standard, Xpert MTB/RIF of BALF showed a higher sensitivity (14/16, 87.5%), but a relatively lower specificity (57/92, 62.0%). Xpert MTB/RIF of sputum showed relatively lower sensitivity (6/10, 60.0%) and higher specificity (63/88, 71.6%). Compared with sputum culture, Xpert MTB /RIF assay reduced the median detection time of MTB from 30 to 0 days, which significantly shortened the diagnosis time of the smear-negative TB patients. Among the combined detections, the positive detection proportion was improved with significant differences comparing with sputum culture only, from 11.1% (10/90) to 46.7% (42/90) (P < 0.05). Our study showed Xpert MTB/RIF in BALF had a better performance in detecting MTB of smear-negative patients.