To more accurately determine the seroprevalence of hepatitis G virus (HGV) infection, we surveyed antibody to HGV (anti-E2) by enzyme-linked immunosorbent assay (ELISA) and HGV RNA by nested polymerase chain reaction (PCR) in 298 residents of a hepatitis C virus (HCV)-endemic area of Japan and in 225 hemodialysis patients. We then compared these findings with known HCV and hepatitis B virus (HBV) infection prevalences. Anti-E2 and HGV RNA prevalences were 32 (10.7%) and 5 (1.7%) in the residents and 24 (10.7%) and 10 (4.4%) in the hemodialysis patients, respectively. Anti-E2 and HGV RNA concurrence was found in two of the hemodialysis patients. Total HGV marker (anti-E2 and/or HGV RNA) prevalences [37 (12.4%) in residents and 32 (14.2%) in hemodialysis patients], were significantly lower than the prevalences of antibody to HCV (anti-HCV) by ELISA [59 (19.8%) and 96 (42.7%)], and antibody to hepatitis B core antigen (anti-HBc) by radioimmunoassay (RIA) [87 (29.2%) and 101 (44.9%)] (P<0.05). The anti-HCV prevalence in subjects with total HGV marker was significantly higher than in those without total HGV marker. There was no significant difference in anti-HBc prevalence between those with and without total HGV marker. The viremic rate was highest in HCV infection (HCV RNA by PCR/anti-HCV) (83.2%), with HGV infection (HGV RNA/total HGV marker) (21.7%) intermediate, and HBV infection (hepatitis B surface antigen by RIA/anti-HBc) (5.3%) lowest (P<0.05). These findings indicate that HGV infection was less endemic than HCV and HBV. HGV was eliminated naturally more frequently than HCV infection and less frequently than HBV infection.