Cancer-associated fibroblasts (CAFs) were originally presumed to represent a homogeneous population uniformly driving tumorigenesis, united by their morphology and peritumoural location. Our understanding of CAFs has since been shaped by sophisticated in vitro and in vivo experiments, pathological association and, more recently, ablation, and it is now widely appreciated that CAFs form a group of highly heterogeneous cells with no single overarching marker. Studies have demonstrated that the CAF population contains different subtypes based on the expression of marker proteins with the capacity to promote or inhibit cancer, with their biological role as accomplices or adversaries dependent on many factors, including the cancer stage. So, while CAFs have been endlessly shown to promote the growth, survival and spread of tumours via improvements in functionality and an altered secretome, they are also capable of retarding tumorigenesis via largely unknown mechanisms. It is important to reconcile these disparate results so that the functions of, or factors produced by, tumour-promoting subtypes can be specifically targeted to improve cancer patient outcomes. This review will dissect out CAF complexity and CAF-directed cancer treatment strategies in order to provide a case for future, rational therapies.