Nitrogenase-dependent H₂ production by photosynthetic bacteria, such as Rhodobacter capsulatus, has been extensively investigated. An important limitation to increase H₂ production using genetic manipulation is the scarcity of high-throughput screening methods to detect possible overproducing mutants. Previously, we engineered R. capsulatus strains that emitted fluorescence in response to H₂ and used them to identify mutations in the nitrogenase Fe protein leading to H₂ overproduction. Here, we used ultraviolet light to induce random mutations in the genome of the engineered H₂-sensing strain, and fluorescent-activated cell sorting to detect and isolate the H₂-overproducing cells from libraries containing 5 × 10⁵ mutants. Three rounds of mutagenesis and strain selection gradually increased H₂ production up to 3-fold. The whole genomes of five H₂ overproducing strains were sequenced and compared to that of the parental sensor strain to determine the basis for H₂ overproduction. No mutations were present in well-characterized functions related to nitrogen fixation, except for the transcriptional activator nifA2. However, several mutations mapped to energy-generating systems and to carbon metabolism-related functions, which could feed reducing power or ATP to nitrogenase. Time-course experiments of nitrogenase depression in batch cultures exposed mismatches between nitrogenase protein levels and their H₂ and ethylene production activities that suggested energy limitation. Consistently, cultivating in a chemostat produced up to 19-fold more H₂ than the corresponding batch cultures, revealing the potential of selected H₂ overproducing strains.