To assess the representativeness of drug use treatment samples for measuring HIV seroprevalence among injecting drug users (IDU) in community settings. Seroprevalence was determined in two cross-sectional, convenience samples including an unlinked survey of IDU entrants to all publicly-funded drug-treatment programs and a survey of community-recruited IDU. Unconditional logistic regression [odds ratio (OR)] was used to calculate unadjusted and adjusted OR to measure the association between HIV seropositivity and site of recruitment. Between 1988 and 1989, 25% of 870 community-recruited IDU were seropositive, compared with 13% of 671 entrants to drug-treatment programs. This twofold risk of HIV seropositivity among community-recruited IDU remained after adjustment for sample differences in gender, race-ethnicity, and age group (adjusted OR, 2.09; 95% confidence interval, 1.58-2.78). These results suggest the importance of extending HIV surveillance outside of drug-treatment facilities. Active serologic surveillance may be feasible by coupling recent saliva and fingerstick sampling techniques with existing community outreach education efforts.