One of the most common DNA lesions arising in cells is an apurinic/apyrimidinic (AP) site resulting from base loss. Although a template strand AP site impedes DNA synthesis, translesion synthesis (TLS) DNA polymerases can bypass an AP site. Because this bypass is expected to be highly mutagenic because of loss of base coding potential, here we quantify the efficiency and the specificity of AP site bypass by two Y family TLS enzymes, Sulfolobus solfataricus DNA polymerase 4 (Dpo4) and human DNA polymerase eta (Pol eta). During a single cycle of processive DNA synthesis, Dpo4 and Pol eta bypass synthetic AP sites with 13-30 and 10-13%, respectively, of the bypass efficiency for undamaged bases in the same sequence contexts. These efficiencies are higher than for the A family, exonuclease-deficient Klenow fragment of Escherichia coli DNA polymerase I. We then determined AP site bypass specificity for complete bypass, requiring insertion or misalignment at the AP site followed by multiple incorporations using the aberrant primer templates. Although Dpo4, Pol eta, and Klenow polymerase have different fidelity when copying undamaged DNA, bypass of AP sites lacking A or G by all three polymerases is nearly 100% mutagenic. The majority (70-80%) of bypass events made by all three polymerases are insertion of dAMP opposite the AP site. Single base deletion errors comprise 10-25% of bypass events, with other base insertions observed at lower rates. Given that mammalian cells contain five polymerases implicated in TLS, and given that a large number of AP sites are generated per mammalian cell per day, even moderately efficient AP site bypass could be a source of substitution and frameshift mutagenesis in vivo.