Several low-grade persistent viral infections induce and sustain very large numbers of virus-specific effector T cells. This was first described as a response to cytomegalovirus (CMV), a herpesvirus that establishes a life-long persistent/latent infection, and sustains the largest known effector T cell populations in healthy people. These T cells remain functional and traffic systemically, which has led to the recent exploration of CMV as a persistent vaccine vector. However, the maintenance of this remarkable response is not understood. Current models propose that reservoirs of viral antigen and/or latently infected cells in lymph nodes stimulate T cell proliferation and effector differentiation, followed by migration of progeny to non-lymphoid tissues where they control CMV reactivation. We tested this model using murine CMV (MCMV), a natural mouse pathogen and homologue of human CMV (HCMV). While T cells within draining lymph nodes divided at a higher rate than cells elsewhere, antigen-dependent proliferation of MCMV-specific effector T cells was observed systemically. Strikingly, inhibition of T cell egress from lymph nodes failed to eliminate systemic T cell division, and did not prevent the maintenance of the inflationary populations. In fact, we found that the vast majority of inflationary cells, including most cells undergoing antigen-driven division, had not migrated into the parenchyma of non-lymphoid tissues but were instead exposed to the blood supply. Indeed, the immunodominance and effector phenotype of inflationary cells, both of which are primary hallmarks of memory inflation, were largely confined to blood-localized T cells. Together these results support a new model of MCMV-driven memory inflation in which most immune surveillance occurs in circulation, and in which most inflationary effector T cells are produced in response to viral antigen presented by cells that are accessible to the blood supply.