Purpose <p>To evaluate the diagnostic accuracy of preoperative imaging in patients with suspected primary facial nerve tumours and to assess interobserver agreement in radiological interpretation.</p> Methods <p>A retrospective study included patients who underwent surgical treatment for suspected primary facial nerve tumours between 1990 and 2024. Concordance between the primary radiological diagnosis and histopathology was analysed. Diagnostic performance metrics were calculated using histopathology as the reference standard. A secondary blinded imaging review was performed by three radiologists. Interobserver agreement was assessed using Cohen’s and Fleiss’ kappa coefficients.</p> Results <p>Twenty-eight patients were included in the primary analysis. Schwannoma was the most frequent radiological diagnosis (53.6%) and histopathological subtype (50%). Overall concordance between the primary radiological diagnosis and histopathology was 57.1%. In the blinded radiological review (<i>n</i> = 23), diagnostic accuracy ranged from 47.8% to 60.9%. Overall interobserver agreement for the final diagnosis was fair (Fleiss’ κ = 0.30). Accuracy was higher for common tumours (schwannomas and haemangiomas) than for rare tumours (68.9% vs. 25%, <i>p</i> = 0.001). Haemangioma showed high diagnostic performance (sensitivity and specificity) in the primary analysis but lower reproducibility in the blinded review.</p> Conclusion <p>Preoperative imaging shows moderate diagnostic accuracy and limited interobserver agreement. Diagnostic performance appears to be higher for common tumours such as schwannomas, whereas haemangiomas and other lesions may show more variable and less reproducible imaging features. Despite these limitations, imaging remains essential for tumour localisation and clinical decision-making.</p>