Directed evolution through genetic engineering of viral capsids followed by selection has emerged as a powerful means to create novel recombinant adeno-associated virus (rAAV) vectors with desired tropism and enhanced properties. One of the most effective approaches uses rAAV-based random peptide display libraries. Here we report a novel system based on an infection-defective rAAV1.9-3 as a platform for random peptide display, and show that biopanning of the libraries in vitro effectively identifies the peptides that restore and enhance rAAV transduction. rAAV1.9-3 has a genetically engineered AAV1 capsid with amino acids 445-568 being replaced with those of AAV9, and has been identified as a variant exhibiting significantly impaired infectivity and delayed blood clearance when infused into mice. In this study, we generated rAAV1.9-3 variant libraries in which 7- or 12-mer random peptides were expressed at the capsid amino acid position 590. Three rounds of positive selection for primary human dermal fibroblasts successfully identified new rAAV-peptide variants that transduce them more efficiently than the prototype rAAV2. Thus our study demonstrates that an infection-defective rAAV variant serves as a novel detargeted platform for random peptide display libraries. We also describe a brief review of recent progress in rAAV-based random peptide display library approaches.