Campylobacter jejuni infections are a leading cause of bacterial food-borne diarrhoeal illness worldwide, and Campylobacter infections in children are associated with stunted growth and therefore long-term deficits into adulthood. Despite this global impact on health and human capital, how zoonotic C. jejuni responds to the human host remains unclear. Unlike other intestinal pathogens, C. jejuni does not harbour pathogen-defining toxins that explicitly contribute to disease in humans. This makes understanding Campylobacter pathogenesis challenging and supports a broad examination of bacterial factors that contribute to C. jejuni infection. Here, we use a controlled human infection model to characterize C. jejuni transcriptional and genetic adaptations in vivo, along with a non-human primate infection model to validate our approach. We found that variation in 11 genes is associated with either acute or persistent human infections and includes products involved in host cell invasion, bile sensing and flagella modification, plus additional potential therapeutic targets. In particular, a functional version of the cell invasion protein A (cipA) gene product is strongly associated with persistently infecting bacteria and we identified its biochemical role in flagella modification. These data characterize the adaptive C. jejuni response to primate infections and suggest therapy design should consider the intrinsic differences between acute and persistently infecting bacteria. In addition, RNA sequencing revealed conserved responses during natural host commensalism and human infections. Thirty-nine genes were differentially regulated in vivo across hosts, lifestyles and C. jejuni strains. This conserved in vivo response highlights important C. jejuni survival mechanisms such as iron acquisition and evasion of the host mucosal immune response. These advances highlight pathogen adaptability across host species and demonstrate the utility of multidisciplinary collaborations in future clinical trials to study pathogens in vivo.