The use of predatory bacteria as live antibiotics has been proposed for managing bacterial infections, especially for those caused by antibiotic multiresistant isolates for which there are few therapeutic options. However, the current knowledge in this field is scarce, with most of the available data based on environmental isolates, with a significant lack of human clinical samples. Here, we evaluated the predatory spectrum of the reference strain Bdellovibrio bacteriovorus 109J on 13 Serratia marcescens (including carbapenemase producers) and 78 Pseudomonas aeruginosa clinical isolates from respiratory (colonizing lungs of cystic fibrosis patients) or bacteremic infections, differentiated by phenotype (mucoid or not), antibiotic resistance phenotype (including multidrug-resistant isolates), and genetic lineage (frequent and rare sequence types). The source of the isolates was significantly associated with predation efficiency (100% for S. marcescens, 67% for P. aeruginosa from cystic fibrosis, and 25% for P. aeruginosa from bacteremia). In contrast, no correlation with colonial morphotype, genetic background, or antibiotic susceptibility was found. To evaluate the influence of the predator on the predation, we employed a more aggressive B. bacteriovorus mutant 109J preying upon the same 48 bacteremic P. aeruginosa isolates. The mutant's predation efficiency was higher than that of their wild-type counterpart (43% vs. 25%), pointing out that predation is specific to each prey-predator pair. Our results provide the most extensive study of clinical prey susceptibility published to date and show that the prey-predator interaction is influenced by the origin of the isolates rather than by their genetic background or their antibiotic susceptibility phenotype.