Epstein-Barr virus (EBV) is a ubiquitous pathogen of humans that can cause several types of lymphoma and carcinoma. Like other herpesviruses, EBV has diversified through both coevolution with its host and genetic exchange between virus strains. Sequence analysis of the EBV genome is unusually challenging because of the large number and lengths of repeat regions within the virus. Here we describe the sequence assembly and analysis of the large internal repeat 1 of EBV (IR1; also known as the BamW repeats) for more than 70 strains. The diversity of the latency protein EBV nuclear antigen leader protein (EBNA-LP) resides predominantly within the exons downstream of IR1. The integrity of the putative BWRF1 open reading frame (ORF) is retained in over 80% of strains, and deletions truncating IR1 always spare BWRF1. Conserved regions include the IR1 latency promoter (Wp) and one zone upstream of and two within BWRF1. IR1 is heterogeneous in 70% of strains, and this heterogeneity arises from sequence exchange between strains as well as from spontaneous mutation, with interstrain recombination being more common in tumor-derived viruses. This genetic exchange often incorporates regions of <1 kb, and allelic gene conversion changes the frequency of small regions within the repeat but not close to the flanks. These observations suggest that IR1—and, by extension, EBV—diversifies through both recombination and breakpoint repair, while concerted evolution of IR1 is driven by gene conversion of small regions. Finally, the prototype EBV strain B95-8 contains four nonconsensus variants within a single IR1 repeat unit, including a stop codon in the EBNA-LP gene. Repairing IR1 improves EBNA-LP levels and the quality of transformation by the B95-8 bacterial artificial chromosome (BAC).IMPORTANCE Epstein-Barr virus (EBV) infects the majority of the world population but causes illness in only a small minority of people. Nevertheless, over 1% of cancers worldwide are attributable to EBV. Recent sequencing projects investigating virus diversity to see if different strains have different disease impacts have excluded regions of repeating sequence, as they are more technically challenging. Here we analyze the sequence of the largest repeat in EBV (IR1). We first characterized the variations in protein sequences encoded across IR1. In studying variations within the repeat of each strain, we identified a mutation in the main laboratory strain of EBV that impairs virus function, and we suggest that tumor-associated viruses may be more likely to contain DNA mixed from two strains. The patterns of this mixing suggest that sequences can spread between strains (and also within the repeat) by copying sequence from another strain (or repeat unit) to repair DNA damage.