Effective risk assessment and control of environmental antibiotic resistance depend on comprehensive information about antibiotic resistance genes (ARGs) and their microbial hosts. Advances in sequencing technologies and bioinformatics have enabled the identification of ARG hosts using metagenome-assembled contigs and genomes. However, these approaches often suffer from information loss and require extensive computational resources. Here we introduce a bioinformatic strategy that identifies ARG hosts by prescreening ARG-like reads (ALRs) directly from total metagenomic datasets. This ALR-based method offers several advantages: (1) it enables the detection of low-abundance ARG hosts with higher accuracy in complex environments; (2) it establishes a direct relationship between the abundance of ARGs and their hosts; and (3) it reduces computation time by approximately 44-96% compared to strategies relying on assembled contigs and genomes. We applied our ALR-based strategy alongside two traditional methods to investigate a typical human-impacted environment. The results were consistent across all methods, revealing that ARGs are predominantly carried by Gammaproteobacteria and Bacilli, and their distribution patterns may indicate the impact of wastewater discharge on coastal resistome. Our strategy provides rapid and accurate identification of antibiotic-resistant bacteria, offering valuable insights for the high-throughput surveillance of environmental antibiotic resistance. This study further expands our knowledge of ARG-related risk management in future.