The purpose of this study was to review the diagnostic accuracy of MRI in detecting perineural spreading (PNS) of head and neck tumors using histopathological or surgical evidence from the afflicted nerve as the reference standard. Previous studies in the English language published in the last 30 years were searched from PubMed and Embase databases. We included studies that used magnetic resonance imaging (MRI) (with and without contrast enhancement) to detect PNS, as well as the histological or surgical confirmation of PNS, and that reported the exact numbers of patients required for assessing diagnostic accuracy. The outcome measures were sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Heterogeneity was assessed with the Higgins inconsistency test (I2). P-values smaller than 0.05 were considered statistically significant. A total of 11 retrospective studies were found, reporting 319 nerve samples from 245 patients. Meta-analytic estimates and their 95% confidence intervals were as follows: sensitivity 0.85 (0.70–0.95), specificity 0.85 (0.80–0.89), PPV 0.86 (0.70–0.94), and NPV 0.85 (0.71–0.93). We found statistically significant heterogeneity for sensitivity (I2 = 72%, p = 0.003) and PPV (I2 = 70%, p = 0.038), but not for NPV (I2 = 65%, p = 0.119) or specificity (I2 = 12%, p = 0.842). The most frequent MRI features of PNS were nerve enlargement and enhancement. Squamous cell carcinoma and adenoid cystic carcinoma were the most common tumor types, and the facial and trigeminal nerves were the most commonly affected nerves in PNS. Only a few studies provided examples of false MRI diagnoses. MRI demonstrated high diagnostic accuracy in depicting PNS of cranial nerves, yet this statement was based on scarce and heterogeneous evidence.