Abstract: The composition and distribution of bacterial communities in petroleum contaminated sites vary greatly and are affected by many factors. The effect of petroleum contamination on the change of bacterial communities in the contaminated site is still unclear. The community assembly process in petroleum-contaminated sites remains to be studied. In a petrochemical site in Gansu Province, four media with different contamination characteristics were selected to investigate the petroleum concentration, bacterial community structure diversity and community assembly process in sludge, petroleum contaminated soil, uncontaminated soil, and contaminated aquifer sediments. The results showed that genotypes with the ability to degrade petroleum (alkB and nah genes) could be selected in an aerobic environment with high petroleum concentration, and the degradation genes in the sludge environment could reach 10⁷ copies/g. The copies of genes that degrade petroleum anaerobically was low in an anaerobic environment such as sediment and the gene of depredating petroleum was only 10⁶ copies/g. Anaerobic degradation genes such as ass and bamA were not highly expressed. The α-diversity of the bacterial community in the site was reduced after petroleum contaminated soil and was significantly negatively correlated with petroleum concentration (P < 0.05). Compared to uncontaminated soils, sludge and contaminated sediments reduced α-diversity by about 50%, whereas petroleum contaminated soils reduced α-diversity by about 33%. In aerobic environments such as sludge, petroleum-contaminated soil and uncontaminated soil, community assembly was dominated by stochastic processes such as diversification and drift, regardless of the degree of contamination. In anaerobic environments, such as contaminated aquifer sediments, microbial community assembly was dominated by deterministic processes such as environmental selection and interspecies competition. This study shows that the diversity and assembly processes of bacterial communities in oil-contaminated sites are mainly affected by oil concentration, but the feedback to the environment, such as selection of degradation genes and assembly processes, requires higher oil concentration to be shown.