Abstract Conjugative plasmids play a key role in the dissemination of antimicrobial resistance (AMR) genes across bacterial pathogens. AMR plasmids are widespread in clinical settings, but their distribution is not random, and certain associations between plasmids and bacterial clones are particularly successful. For example, the globally spread carbapenem resistance plasmid pOXA-48 can use a wide range of enterobacterial species as hosts, but it is usually associated with a small number of specific Klebsiella pneumoniae clones. These successful associations represent an important threat for hospitalized patients. However, knowledge remains limited about the factors determining AMR plasmid distribution in clinically relevant bacteria. Here, we combined in vitro and in vivo experimental approaches to analyze pOXA-48-associated AMR levels and conjugation dynamics in a collection of wild type enterobacterial strains isolated from hospitalized patients. Our results reveal significant variability in these traits across different bacterial hosts, with Klebsiella spp. strains showing higher pOXA-48-mediated AMR and conjugation frequencies than Escherichia coli strains. Using experimentally determined parameters, we developed a simple mathematical model to interrogate the contribution of AMR levels and conjugation permissiveness to plasmid distribution in bacterial communities. The simulations revealed that a small subset of clones, combining high AMR levels and conjugation permissiveness, play a critical role in stabilizing the plasmid in different polyclonal microbial communities. These results help to explain the preferential association of plasmid pOXA-48 with K. pneumoniae clones in clinical settings. More generally, our study reveals that species- and strain-specific variability in plasmid-associated phenotypes shape AMR evolution in clinically relevant bacterial communities. Significance statement Conjugative plasmids disseminate AMR genes across bacterial pathogens. Understanding the rules governing plasmid dynamics in bacterial communities is therefore crucial to controlling the global AMR crisis. In this study, we analyzed the dynamics of an AMR plasmid of great clinical relevance, pOXA-48, in a collection of wild type bacteria recovered from hospitalized patients. We reported a high degree of variability in two key plasmid-associated phenotypes, AMR level and conjugation ability, across the collection of clinical bacteria. Using simulations based on the experimental results, we studied how successful associations between AMR plasmids and clinical strains can arise in bacterial communities. Our results revealed that accounting for variability in plasmid-associated phenotypes help to understand the evolution of AMR in clinical settings.