We wished to identify opportunities for improving the yield of positive HIV test results from federally funded HIV counseling and testing programs. We reviewed client records from 1992 and 1993 for targeting opportunities at the site level based on site type (i.e., family planning clinic) and the seropositivity in the past and at the client level based on the client's history of a past negative test, demographics, and risk history. We studied 1,281,606 records from 1992. The number of tests and opportunities for site-level targeting varied by project area. Seropositivity varied by site type, but the best predictor of seropositivity was seropositivity at that site in the past. Of 1,102 sites with <1% of tests positive in 1992, only five had >3% positive in 1993. Sites with no positive tests in 1992 performed 99,468 tests in 1993, and only 292 (0.3%) were positive. Clients with a past negative test had a slightly lower seropositivity (1.5%) than clients with no previous test (2.0%). In sites with a low (0.1-2.0%) seropositivity, clients with no transmission risk by history were unlikely to be infected (0.8% for black men). However, in sites with high (> or =5%) seropositivity, clients without risk were often infected (5.7% for black men). Opportunities for targeting were identified. They vary considerably by project area and testing site. These opportunities for targeting should be considered by sites as AIDS prevention strategies evolve.