The duplex structure of DNA, with its internal base pairing, protects the nucleobases from chemical damage, but it also poses a barrier to DNA-modifying enzymes, including the enzymes that recognize and repair DNA damage. It is known that unpaired (or bulged) nucleotides are significantly more accessible, but it is not known whether they might be recognized by nucleotide-flipping enzymes. We have investigated this question with human alkyladenine DNA glycosylase (AAG). AAG recognizes a wide variety of structurally disparate lesions, including deoxyinosine (I), which results from the spontaneous oxidative deamination of adenosine, and catalyzes the hydrolysis of the N-glycosidic bond to release the lesion base and initiate the base excision repair pathway. We used single-turnover kinetics to characterize the reactions of AAG with synthetic 25-mer oligonucleotides containing a single I lesion in single-stranded, mismatched, or single-nucleotide bulge contexts. We found that AAG has the highest catalytic efficiency toward a lesion that is presented in a single-nucleotide bulge. In contrast, AAG has more than 2000-fold reduced catalytic efficiency toward a single-stranded I-containing oligonucleotide relative to the duplexes. We have observed 20-fold differences in catalytic efficiency for the excision of the presumed biological target (paired with T) relative to alternative pairings such as C that might be formed by the replication of an unrepaired I. Furthermore, a linear free-energy relationship shows a strong inverse correlation between duplex stability and catalytic efficiency (slope = -0.6 to -1.0), indicating that gaining access to the base lesion provides a substantial barrier to AAG-catalyzed initiation of DNA repair. The observation that AAG recognizes a single-nucleotide bulge as efficiently as a mismatch implies that the recognition of DNA damage is remarkably plastic.