Numerous studies have enriched our knowledge of the microbial community composition and metabolic versatility of contaminated soil. However, there remains a substantial gap regarding the bioassembly patterns of the indigenous microbial community distribution in contaminated deep soils. Herein, the indigenous microbial community structure diversity, function, and co-occurrence relationships in aged PAH-contaminated deep soil collected from an abandoned chemical facility were investigated using high-throughput sequencing. The results showed that the dominant phyla in all samples were responsible for PAH degradation and included Proteobacteria (20.86%–81.37%), Chloroflexi (2.03%–28.44%), Firmicutes (3.06%–31.16%), Actinobacteria (2.92%–11.91%), Acidobacteria (0.41%–12.68%), and Nitrospirae (0.81%–9.21%). Eighty biomarkers were obtained by linear discriminant analysis of effect size (LEfSe), and most of these biomarkers were PAH degraders. Functional predictions using Tax4Fun indicated that the aged contaminated soil has the potential for PAH degradation. Statistical analysis showed that in contrast with the PAH concentration, edaphic properties (nutrients and pH) were significantly correlated (r > 0.25, P < 0.01) with the bacterial community and functional composition. Co-occurrence network analysis (modularity index of 0.781) revealed non-random assembly patterns of the bacterial communities in the PAH-contaminated soils. The modules in the network were mainly involved in carbon and nitrogen cycles, organic substance degradation, and biological electron transfer processes. Microbes from the same module had strong ecological linkages. Additionally, SAR202 clade, Thermoanaerobaculum, Nitrospira, and Xanthomonadales, which were identified as keystone species, played an irreplaceable role in the network. Overall, our results suggested that environmental factors such as nutrients and pH, together with ecological function, are the main factors driving the assembly of microbial communities in aged PAH-contaminated deep soils.