Despite longstanding promise and many known examples, antimicrobial peptides (AMPs) have failed, thus far, to impact human medicine. On the basis of the physical chemistry and mechanism of action of AMPs, we hypothesized that host cell interactions could contribute to a loss of activity in vivo where host cells are highly concentrated. To test this idea, we characterized AMP activity in the presence of human red blood cells (RBC). Indeed, we show that most of a representative set of natural and synthetic AMPs tested are significantly inhibited by preincubation with host cells and would be effectively inactive at physiological cell density. We studied an example broad-spectrum AMP, ARVA (RRGWALRLVLAY), in a direct, label-free binding assay. We show that weak binding to host cells, coupled with their high concentration, is sufficient to account for a loss of useful activity, for at least some AMPs, because >1 × 108 peptides must be bound to each bacterial cell to achieve sterilization. The effect of host cell preincubation on AMP activity is comparable to that of serum protein binding. Feasible changes in host cell binding could lead to AMPs that do not lose activity through interaction with host cells. We suggest that the intentional identification of AMPs that are active in the presence of concentrated host cells can be achieved with a paradigm shift in the way AMPs are discovered.