A molecular method for the detection of Paenibacillus azotofixans in soil and the wheat rhizosphere was developed. The system consisted of polymerase chain reaction (PCR) amplification of part of the variable V1 to V4 regions of the 16S ribosomal RNA gene, followed by hybridization with a specific oligonucleotide probe homologous to part of the intervening region. In vitro specificity tests showed that the detection system worked specifically for P. azotofixans strains, and did not detect other Paenibacillus species or species of other bacterial genera. Vegetative cells of a rifampicin resistant P. azotofixans derivative were trackable in Flevo silt loam (FSL) soil in 24 h experiments using both selective plating and most probable number (MPN)-PCR combined with probing, and plate counts parallelled MPN-PCR estimations of numbers of specific targets. MPN-PCR allowed for the detection of down to 102 introduced cells per g of dry soil. Introduced P. azotofixans spores did not form colonies on selective plates, but were detectable via PCR. The P. azotofixans populations introduced into the silt loam soil suffered a slow decline of the detectable plate count over a period of 14 days. MPN-PCR revealed a similar decline of the number of specific DNA targets. Greater numbers of targets were found in wheat rhizosphere from Flevo silt loam soil, and these numbers persisted throughout the experiment. Soil drying resulted in enhanced persistence of the target sequences, whereas in a constantly moist soil the numbers of target sequences declined. Rewetting of dried soil resulted in declining target sequence numbers. The MPN-PCR detection method is adequate to assess the impact of stress conditions affecting P. azotofixans in FSL and probably other soils, since it abolishes the need for culturing or specific markers and is direct and unambiguous due to its high specificity.