Aim To determine the diagnostic accuracy of ultrasound to detect partial and complete thickness rotator cuff tears based on all available clinical trials. Materials and methods An electronic search of databases registering published and unpublished literature was conducted. All diagnostic accuracy studies that directly compared the accuracy of ultrasound (the index test) to either arthroscopic or open surgical findings (the reference test) for rotator cuff tear were included. The methodological quality of each included study was assessed using the QUADAS form. When appropriate, pooled sensitivity and specificity analysis was conducted, with an assessment of the summary receiver operating characteristic (ROC) curve for each analysis. Results Sixty-two studies assessing 6007 patients and 6066 shoulders were included. Ultrasonography had good sensitivity and specificity for the assessment of partial thickness (sensitivity 0.84; specificity 0.89), and full-thickness rotator cuff tears (sensitivity 0.96; specificity 0.93). However, the literature poorly described population characteristics, assessor blinding, and was based on limited sample sizes. The literature assessing transducer frequency was particularly small in size. Conclusion Ultrasonography is an appropriate radiological technique for the assessment of rotator cuff tears with an acceptable sensitivity and specificity. The diagnostic test accuracy of ultrasound is superior for the detection of full-thickness compared to partial-thickness cuff tears. Further study assessing the effect of transducer frequency is warranted.