The significance of androgen receptor (AR) in breast cancer (BC) management is not fully defined, and it is still ambiguous how the level of AR expression influences oestrogen receptor-positive (ER+) tumours. The aim of the present study was to analyse the prognostic impact of AR/ER ratio, evaluated by immunohistochemistry (IHC), correlating this value with clinical, pathological and molecular characteristics. We retrospectively selected a cohort of 402 ER+BC patients. On each tumour, IHC analyses for AR, ER, PgR, HER2 and Ki67 were performed and AR+ cases were used to calculate the AR/ER value. A cut-off of ≥2 was selected using receiver-operating characteristic (ROC) curve analyses. RNA from 19 cases with AR/ER≥2 was extracted and used for Prosigna-PAM50 assays. Tumours with AR/ER≥2 (6%) showed more frequent metastatic lymph nodes, larger size, higher histological grade and lower PgR levels than cases with AR/ER<2. Multivariate analysis confirmed that patients with AR/ER≥2 had worse disease-free interval (DFI) and disease-specific survival (DSS) (hazard ratios (HR) = 4.96 for DFI and HR = 8.69 for DSS, both ≤ 0.004). According to the Prosigna-PAM50 assay, 63% (12/19) of these cases resulted in intermediate or high risk of recurrence categories. Additionally, although all samples were positive for ER assessed by IHC, the molecular test assigned 47.4% (9/19) of BCs to intrinsic non-luminal subtypes. In conclusion, the AR/ER ratio ≥2 identifies a subgroup of patients with aggressive biological features and may represent an additional independent marker of worse BC prognosis. Moreover, the Prosigna-PAM50 results indicate that a significant number of cases with AR/ER≥2 could be non-luminal tumours.