Studies of HIV molecular evolution and pathogenesis have relied on the polymerase chain reaction (PCR) to provide sequence information from infected tissues. Until recently, studies have been constrained by the limited length of fragments that can be reliably amplified. The addition of a thermostable 3'-exonuclease activity and altered cycling profiles has increased the length of target sequences that can be amplified by more than 10-fold. We have evaluated the fidelity of long PCR (LPCR). We determined that LPCR amplification maintains the distribution of sequences found in a heterogeneous sample and introduces nucleotide misincorporations at a rate comparable to that found with routine PCR. However, a significant proportion of the LPCR-amplified DNA fragments resulted from recombination events. This result suggests that LPCR amplification may have limited utility in the production and analysis of full-length HIV clones.