OBJECTIVE To compare the diagnostic accuracy of dual-energy computed tomography arthrography (DE-CTA) and magnetic resonance arthrography (MRA) of the shoulder in depicting glenoid labral tears. METHODS This prospective institutional review board-approved study included 47 consecutive patients (28 males, 19 females; mean age of 34.2 years) studied between January 2017 and October 2018. All patients underwent DE-CTA and MRA the same day. Two radiologists (25 and 11 years of experience, respectively), blinded to clinical data, evaluated the presence labral tears on virtual-blended 120 kV standard CTA and on DE-CTA images. A third radiologist (18 years of experience) evaluated the MRA images. Diagnostic accuracy values were calculated by using surgery as standard of reference. Inter-observer and intra-observer agreements were calculated with k statistics. A value of p < 0.05 was considered statistically significant. RESULTS Surgery revealed the presence of labral tears in 38/47 patients (80.9%). Sensitivity and specificity values in diagnosing labral tears were 84.2% and 77.8% for MRA (Reader 3), 84.2% and 77.8% for CTA (Reader 1), 84.2% and 88.9% for CTA (Reader 2), 89.5% and 88.9% for DE-CTA (Reader 1), and 92.1% and 88.9% for DE-CTA (Reader 2). A nonsignificant increase in AUC values with respect to MRA was obtained by reading the CTA (p = 0.470) and DE-CTA dataset (p = 0.217), respectively. Inter-observer agreements were near perfect for CTA (k = 0.84) and substantial for DE-CTA reading (k = 0.76). Intra-observer agreements were near perfect both for CTA (k = 0.88) and for DE-CTA reading (k = 0.82). CONCLUSIONS DE-CTA and MRA were not different in terms of diagnostic performance.