Preclinical studies suggest that some effects of conventional chemotherapy, and in particular, gemcitabine, are mediated through enhanced antitumor immune responses. The objective of this study was to use material from a randomized clinical trial to evaluate whether patients with preexisting immune infiltrates responded better to treatment with gemcitabine + docetaxel (GD) compared to docetaxel alone. Formalin fixed, paraffin-embedded breast cancer tissues from SBG0102 phase 3 trial patients randomly assigned to treatment with GD or docetaxel were used. Immunohistochemical staining for CD8, FOXP3, LAG3, PD-1, PD-L1 and CD163 was performed. Tumor infiltrating lymphocytes (TILs) and tumor associated macrophages were evaluated. Prespecified statistical analyses were performed in a formal prospective-retrospective design. Time to progression was primary endpoint and overall survival secondary endpoint. Correlations between biomarker status and endpoints were evaluated using the Kaplan–Meier method and Cox proportional hazards models. Biomarker data was obtained for 237 patients. There was no difference in treatment effect according to biomarker status for the whole cohort. In planned subgroup analysis by PAM50 subtype, in non-luminal (basal-like and HER2E) breast cancers FOXP3 was a significant predictor of treatment effect with GD compared to docetaxel, with a HR of 0.22 (0.09–0.52) for tumors with low FOXP3 compared to HR 0.92 (0.47–1.80) for high FOXP3 TILs (P_interaction = 0.01). Immune biomarkers were not predictive of added benefit of gemcitabine in a cohort of mixed breast cancer subtypes. However, in non-luminal breast cancers, patients with low FOXP3+ TILs may have significant benefit from added gemcitabine.