Genomic uracil resulting from spontaneously deaminated cytosine generates mutagenic U:G mismatches that are usually corrected by error-free base excision repair (BER). However, in B-cells, activation-induced cytosine deaminase (AID) generates U:G mismatches in hot-spot sequences at Ig loci. These are subject to mutagenic processing during somatic hypermutation (SHM) and class switch recombination (CSR). Uracil N-glycosylases UNG2 and SMUG1 (single strand-selective monofunctional uracil-DNA glycosylase 1) initiate error-free BER in most DNA contexts, but UNG2 is also involved in mutagenic processing of AID-induced uracil during the antibody diversification process, the regulation of which is not understood. AID is strictly single strand-specific. Here we show that in the presence of Mg2+ and monovalent salts, human and mouse SMUG1 are essentially double strand-specific, whereas UNG2 efficiently removes uracil from both single and double stranded DNA under all tested conditions. Furthermore, SMUG1 and UNG2 display widely different sequence preferences. Interestingly, uracil in a hot-spot sequence for AID is 200-fold more efficiently removed from single stranded DNA by UNG2 than by SMUG1. This may explain why SMUG1, which is not excluded from Ig loci, is unable to replace UNG2 in antibody diversification. We suggest a model for mutagenic processing in which replication protein A (RPA) recruits UNG2 to sites of deamination and keeps DNA in a single stranded conformation, thus avoiding error-free BER of the deaminated cytosine.