Background: rapid and accurate diagnosis of extrapulmonary tuberculosis (EPTB) continues to be a challenge. Although culture remains the most sensitive method for confirmation of TB, the prolonged time required for culture, limit its contribution to clinical decision making. Our study aims to assess the performance of Xpert MTB/RIF test against smear and culture-confirmed cases of extrapulmonary TB. Methods: A total, 272 non-respiratory specimens (tissues, 88 (32.4%), pleural fluids, 60 (22.1%), CSF, 48 (17.6), aspirate 36 (13.2), ascetic fluid, 20 (7.4), urine, 8 (2.9), body fluid and pus 5 (1.8) each, and blood 2 (0.73) submitted to the laboratory for Mycobacteria over two-year period were comparatively investigated with the molecular-based Gene Xpert MTB/RIF assay system and conventional smear and solid culture methods. Result: The reliability indices of the Gene Xpert MTB/RIF assay are higher compared to smear. The overall sensitivity (82.1%) of the Gene Xpert MTB/RIF assay is significantly higher than smear (46.9 %). The highest Mtb positivity agreement between the Xpert MTB/RIF assay and Mtb culture was found in pus and CSF specimens (100 % [95% CI, 91.1 % to 100 %]) each, while the lowest Mtb positivity agreement was found in the ascetic fluids specimens (50% [95% CI, 44% to 58%]). The Area under the curve (AUC) of the Receiver Operator Characteristic (ROC) curve to assess the accuracy of smear was 0.737 (with 95% CI of 0.600-0.874), whereas the AUC of the ROC curve for Xpert MTB/RIF was 0.946 (with 95% CI of 0.000-1.00). Conclusion: The Xpert assay showed superior performance over the conventional smear for the rapid detection of M. tuberculosis, in (EPTB).