Adeno-associated viral (AAV) vectors are currently being tested in multiple clinical trials for liver-directed gene transfer to treat the bleeding disorders hemophilia A and B and metabolic disorders. The optimal viral capsid for transduction of human hepatocytes has been under active investigation, but results across various models are inconsistent. We tested in vivo transduction in "humanized" mice. Methods to quantitate percent AAV transduced human and murine hepatocytes in chimeric livers were optimized using flow cytometry and confocal microscopy with image analysis. Distinct transduction efficiencies were noted following peripheral vein administration of a self-complementary vector expressing a gfp reporter gene. An engineered AAV3 capsid with two amino acid changes, S663V+T492V (AAV3-ST), showed best efficiency for human hepatocytes (~3-times, ~8-times, and ~80-times higher than for AAV9, AAV8, and AAV5, respectively). AAV5, 8, and 9 were more efficient in transducing murine than human hepatocytes. AAV8 yielded the highest transduction rate of murine hepatocytes, which was 19-times higher than that for human hepatocytes. In summary, our data show substantial differences among AAV serotypes in transduction of human and mouse hepatocytes, are the first to report on AAV5 in humanized mice, and support the use of AAV3-based vectors for human liver gene transfer.