Background. The increase in antibiotic resistance in human-impacted environments, such as coastal waters with aquaculture activity, is related to the widespread use of antibiotics, even at sub-lethal concentrations. In Chile, the world's second largest producer of salmon, aquaculture is considered the main source of antibiotics in coastal waters. In this work, we aimed to characterize the genetic and phenotypic profiles of antibiotic resistance in bacterial communities from salmon farms. Methods. Bacterial metagenomes from an intensive aquaculture zone in southern Chile were sequenced, and the composition, abundance and sequence of antibiotic resistance genes (ARGs) were analyzed using assembled and raw read data. Total DNA from bacterial communities was used as a template to recover floR gene variants, which were tested by heterologous expression and functional characterization of phenicol resistance. Results. Prediction of ARGs in salmon farm metagenomes using more permissive parameters yielded significantly more results than the default Resistance Gene Identifier (RGI) software. ARGs grouped into drug classes showed similar abundance profiles to global ocean bacteria. The floR gene was the most abundant phenicol-resistance gene with the lowest gene counts, showing a conserved sequence although with variations from the reference floR. These differences were recovered by RGI prediction and, in greater depth, by mapping reads to the floR sequence using SNP base-calling. These variants were analyzed by heterologous expression, revealing the co-existence of high- and low-resistance sequences in the environmental bacteria. Conclusions. This study highlights the importance of combining metagenomic and phenotypic approaches to study the genetic variability in and evolution of antibiotic-resistant bacteria associated with salmon farms.