The objective of our study was to evaluate the accuracy of frozen section (FS) in borderline ovarian tumours (BOT) and to define the factors associated with misdiagnosis during FS evaluation. We performed a retrospective review of patients who underwent exploratory laparotomy for an adnexal mass, from January 2007 to July 2012, at a tertiary oncology centre in Turkey. Patients with a diagnosis of BOT either in FS or in permanent pathology were identified. Agreement between FS diagnosis and permanent histology was observed in 37/59 patients (62.7%), which gave a sensitivity and a positive predictive value of 71.2% and 84.1%, respectively. In patients with a diagnosis of BOT by frozen section only (n = 44), the diagnosis was consistent with permanent histopathology in 37/44 patients (84.1%). Frozen section interpreted a malignant tumour as BOT (under-diagnosis) in 6/44 (13.6%) of cases and interpreted a benign lesion as BOT (over-diagnosis) in 1/44 (2.3%) of cases. Slide review of discrepant cases revealed that major pathological causes of under-diagnosis were misinterpretation and sampling errors. Univariate analysis showed that presence of bilateral tumour and positive peritoneal cytology were associated with under-diagnosis. We concluded that, despite significant risk of under-diagnosis, FS analysis is an accurate method for intraoperative diagnosis of BOTs.