The heterogeneous expression of EGFRvIII [variant III mutant of epidermal growth factor receptor (EGFR)] in glioblastoma has significant impact on the clinical response to the treatment of EGFRvIII-specific chimeric antigen receptor-engineered T (CAR T) cells. We hypothesized that CAR T cells that could target both EGFRvIII and the form of EGFR expressed on tumor cells, but not EGFR on normal cells, would greatly improve efficacy without inducing on-target, off-tumor toxicity. Therefore, we developed a humanized single-chain antibody, M27, with a single specificity that binds to an epitope found both on wild-type EGFR- and EGFRvIII-overexpressing tumor cells, but not EGFR-expressing normal cells, including primary keratinocytes, on which wild-type EGFR is highly expressed. M27-derived CAR T cells effectively lysed EGFRvIII- or EGFR-overexpressing tumor cells, but showed no observable toxicity on normal cells. Inclusion of the CD137 (4-1BB) costimulatory intracellular domain in the M27-28BBZ CAR provided CAR T cells with higher tumor lysis activity than when not included (as in the M27-28Z CAR). The growth of established EGFR- or EGFRvIII-overexpressing glioma xenografts was suppressed by M27-28BBZ CAR T cells as well. The growth of heterogeneic xenograft tumors, created by mixing EGFR- and EGFR-overexpressing glioblastoma cells, was also effectively inhibited by M27-28BBZ CAR T cells. The survival of mice in the orthotopic models was significantly prolonged after M27-28BBZ CAR T-cell infusion. These results suggested that tumor-selective, bitargeted anti-EGFR/EGFRvIII CAR T cells may be a promising modality for the treatment of patients with EGFR/EGFRvIII-overexpressing glioblastoma. .