Some oncolytic viruses, such as myxoma virus (MYXV), can selectively target malignant hematopoietic cells, while sparing normal hematopoietic cells. This capacity for discrimination creates an opportunity to use oncolytic viruses as ex vivo purging agents of autologous hematopoietic cell grafts in patients with hematologic malignancies. However, the mechanisms by which oncolytic viruses select malignant hematopoietic cells are poorly understood. In this study, we investigated how MYXV specifically targets human AML cells. MYXV prevented chloroma formation and bone marrow engraftment of two human AML cell lines, KG-1 and THP-1. The reduction in human leukemia engraftment after ex vivo MYXV treatment was dose-dependent and required a minimum MOI of 3. Both AML cell lines demonstrated MYXV binding to leukemia cell membranes following co-incubation: however, evidence of productive MYXV infection was observed only in THP-1 cells. This observation, that KG-1 can be targeted in vivo even in the absence of in vitro permissive viral infection, contrasts with the current understanding of oncolytic virotherapy, which assumes that virus infection and productive replication is a requirement. Preventing MYXV binding to AML cells with heparin abrogated the purging capacity of MYXV, indicating that binding of infectious virus particles is a necessary step for effective viral oncolysis. Our results challenge the current dogma of oncolytic virotherapy and show that in vitro permissiveness to an oncolytic virus is not necessarily an accurate predictor of oncolytic potency in vivo.