Environmental surveillance of antibiotic resistance genes (ARGs) is critical for understanding and mitigating the spread of antimicrobial resistance. Current short-read-based ARG profiling methods are limited in their ability to provide detailed host information, which is indispensable for tracking the transmission and assessing the risk of ARGs. Here, we present Argo, a novel approach that leverages long-read overlapping to rapidly identify and quantify ARGs in complex environmental metagenomes at the species level. Argo significantly enhances the resolution of ARG detection by assigning taxonomic labels collectively to clusters of reads, rather than to individual reads. By benchmarking the performance in host identification using simulation, we confirm the advantage of long-read overlapping over existing metagenomic profiling strategies in terms of accuracy. Using sequenced mock communities with varying quality scores and read lengths, along with a global fecal dataset comprising 329 human and non-human primate samples, we demonstrate Argo's capability to deliver comprehensive and species-resolved ARG profiles in real settings.