Recombination plays a pivotal role in generating within-host diversity and enabling HIV's evolutionary success, particularly in evading the host immune response. Despite this, the variability in recombination rates across different settings and the underlying factors that drive these differences remain poorly understood. In this study, we analysed a large dataset encompassing hundreds of untreated, longitudinally sampled infections using both whole-genome long-read and short-read sequencing datasets. By quantifying recombination rates, we uncover substantial variation across subtypes, viral loads, and stages of infection. We also map recombination hot and cold spots across the genome using a sliding window approach, finding that previously reported inter-subtype regions of high or low recombination are replicated at the within-host level. Importantly, our findings reveal the significant influence of selection on recombination, showing that the presence and success of recombinant genomes is strongly interconnected with the fitness landscape. These results offer valuable insights into the contribution of recombination to evolutionary dynamics and demonstrate the enhanced resolution that long-read sequencing offers for studying viral evolution.