Chimeric antigen receptor (CAR) T cells have been approved for use in patients with B cell malignancies or relapsed and/or refractory multiple myeloma, yet efficacy against most solid tumours remains elusive. The limited imaging and biopsy data from clinical trials in this setting continues to hinder understanding, necessitating a reliance on imperfect preclinical models. In this Perspective, I re-evaluate current data and suggest potential pathways towards greater success, drawing lessons from the few successful trials testing CAR T cells in patients with solid tumours and the clinical experience with tumour-infiltrating lymphocytes. The most promising approaches include the use of pluripotent stem cells, co-targeting multiple mechanisms of immune evasion, employing multiple co-stimulatory domains, and CAR ligand-targeting vaccines. An alternative strategy focused on administering multiple doses of short-lived CAR T cells in an attempt to pre-empt exhaustion and maintain a functional effector pool should also be considered.