Rickettsia typhi and R. felis are flea-transmitted human pathogenic rickettsial species. To investigate the distributional dynamics of these rickettsiae we designed a micro-immunofluorescence assay (MIF) using species-specific monoclonal antibodies (MAbs) applied to flea cryosections. Our assay was performed in less than 3 h and its applicability was demonstrated by the detection of R. typhi in 50 artificially infected human body lice but in none of 50 uninfected lice. With MIF, we identified 31 positive among 32 fleas proven with PCR to be naturally infected with R. felis; and 7 positive among 32 fleas proven with PCR to be naturally infected with R. typhi. No cross-detection was observed with both MAbs. Fresh R. felis-infected fleas were significantly more MIF-positive than long conserved R. typhi-infected fleas (31/32 vs. 7/32, P < 0.01). This discrepancy may be linked to degradation of antigens by long-term freezing. For R. typhi-infected fleas, our assay was significantly more efficient when applied to fleas in early stages of infection (less than 15 days) by comparison with fleas frozen more than 20 days after infection (7/15 vs. 0/17, P = 0.01). This difference may be related to an antigenic modification caused by selection pressure in the vector and host process. The sensitivity of the described method did not exceed 47% (7/15) for R. typhi but, in contrast, was 97% for R. felis. Thus, our method appears to be useful for surveillance in R. felis infections, but requires further studies for the detection of R. typhi.