G protein-coupled estrogen receptor (GPER), or GPR30, is a membrane receptor reported to mediate non-genomic estrogen responses. Tamoxifen is a partial agonist at GPER in vitro. Here, we investigated if GPER expression is prognostic in primary breast cancer, if the receptor is treatment-predictive for adjuvant tamoxifen, and if receptor subcellular localization has any impact on the prognostic value. Total and plasma membrane (PM) GPER expression was analyzed by immunohistochemistry in breast tumors from 742 postmenopausal lymph node-negative patients subsequently randomized for tamoxifen treatment for 2-5 years versus no systemic treatment, regardless of estrogen receptor (ER) status, and with a median follow-up of 17 years for patients free of event. PM GPER expression was a strong independent prognostic factor for poor prognosis in breast cancer without treatment-predictive information for tamoxifen. In the tamoxifen-treated ER-positive and progesterone receptor (PgR)-positive patient subgroup, the absence of PM GPER (53 % of all ER-positive tumors) predicted 91 % 20-year distant disease-free survival, compared to 73 % in the presence of GPER (p = 0.001). Total GPER expression showed positive correlations with ER and PgR and negative correlation with histological grade, but the correlations were biphasic. On the other hand, PM GPER expression showed strong negative correlations with ER and PgR, and strong positive correlation with HER2 overexpression and high histological grade. GPER overexpression and PM localization are critical events in breast cancer progression, and lack of GPER in the PM is associated with excellent long-term prognosis in ER-positive and PgR-positive tamoxifen-treated primary breast cancer.