We evaluated the upgraded Amplified Mycobacterium Tuberculosis Direct Test kit (AMTD) (Gen-Probe Inc.) for the direct detection of Mycobacterium tuberculosis in respiratory and non-respiratory specimens, and compared the results between the traditional 30,000 RLUs cutoff criteria (C) and three equivocal ranges (30,000-100,000, R1; 30,000-500,000, R2; and 30,000-1,000,000, R3). We tested 663 respiratory and 238 non-respiratory samples from 464 patients. The gold standard was considered to be the combination of culture and clinical data. One hundred and nineteen samples were from 56 patients with pulmonary tuberculosis, and 36 samples were from 19 patients with extrapulmonary tuberculosis. When C criteria was applied, the sensitivity and specificity values were 90.8 and 93.0% for respiratory specimens, while they were 88.9 and 92.1% for non-respiratory specimens (p = NS). The sensitivity was significantly higher in smear-positive specimens (96.7%) than in smear-negative ones (81.0%) (p < 0.05). When compared with C criteria, the overall sensitivity was maintained at 90.3% for R1 criteria, and slightly decreased to 89.7% for R2 and R3 criteria (p = NS). Overall, specificity increased significantly from 92.9% (C) to 97.5% (R1), 99.1% (R2), and 99.2% (R3). Application of R2 or R3 criteria improved significantly the specificity of the test with little decrease in sensitivity.