The use of dispersants in marine environments is a common practice worldwide for oil spill remediation. While the effects of chemical dispersants have been extensively studied, those of biosurfactants, mainly surfactin that is considered one of the most effective surfactants produced by bacteria, have been less considered. We constructed microcosms containing marine water collected from Grumari beach (W_GB, Brazil) and from Schiermonnikoog beach (W_SI, The Netherlands) with the addition of oil (WO), Ultrasperse II plus oil (WOS), surfactin plus oil (WOB), and both dispersants (WS or WB) individually. In these treatments, the composition of bacterial communities and their predictive biodegradation potential were determined over time. High-throughput sequencing of the rrs gene encoding bacterial 16S rRNA revealed that Bacteroidetes (Flavobacteria class) and Proteobacteria (mainly Gammaproteobacteria and Alphaproteobacteria classes) were the most abundant phyla found among the W_GB and W_SI microbiomes, and the relative abundance of the bacterial types in the different microcosms varied based on the treatment applied. Non-metrical multidimensional scaling (NMDS) revealed a clear clustering based on the addition of oil and on the dispersant type added to the GB or SI microcosms, i.e., WB and WOB were separated from WS and WOS in both marine ecosystems studied. The potential presence of diverse enzymes involved in oil degradation was indicated by predictive bacterial metagenome reconstruction. The abundance of predicted genes for degradation of petroleum hydrocarbons increased more in surfactin-treated microcosms than those treated with Ultrasperse II, mainly in the marine water samples from Grumari beach.