Bacteriophages, viruses that infect bacteria, shape microbial communities, including the rumen microbiome, yet assigning them to hosts in complex environments remains challenging. We applied proximity-ligation (Hi-C) metagenomics to rumen contents from twelve dairy cows across four pasture-based diets to recover genome-resolved viral sequences alongside microbial metagenome-assembled genomes (MAGs) and plasmids. Because Hi-C captures physical interactions at the time of sampling, we contrasted these contact-based host links with in silico prediction (iPHoP), which leverages historical signals (e.g., CRISPR matches). We annotated genomes to catalogue auxiliary viral genes (AVGs) and to identify antimicrobial resistance (AMR) genes in their genomic context. We resolved a large rumen phage catalogue and established thousands of virus–host links, including numerous prophages. Hi-C and iPHoP showed broad concordance on shared assignments, with Hi-C adding time-specific interaction evidence. Ecologically, viral productivity decreased as host abundance increased (consistent with kill-the-winner dynamic among rarer taxa), whereas dominant hosts exhibited higher lysogeny (aligning with piggyback-the-winner dynamics). AMR genes predominantly localised to bacterial chromosomes or plasmids, while carriage on viral genomes was rare, suggesting phage impacts on AMR act mainly via predation on ARG-carrying hosts rather than viral mobilisation of ARGs. With Hi-C capturing phage–host interaction dynamics, its host assignments can inform putative hosts for closely related rumen phages in external datasets, including uncultivated viruses represented only in reference databases.  A Hi-C, host-resolved framework offers a practical route to characterise rumen phage–host networks and responsibly extend host annotations to related and uncultivated phages, supporting future ecological and applied studies.