Clinical HIV risk assessments have not typically integrated questions about sex partners' HIV status with questions about condom use and type of sex. Since 2001, we have asked all men who have sex with men (MSM) evaluated in an urban sexually transmitted disease (STD) clinic how often in the preceding 12 months they used condoms for anal sex with partners who were HIV-positive, HIV-negative, and of unknown HIV status. Overall, MSM displayed a pattern of assortative mixing by HIV status, particularly for unprotected anal intercourse (UAI). Nevertheless, 433 (27%) of 1580 MSM who denied knowing they were HIV-positive and 93 (43%) of 217 HIV-positive MSM reported having UAI with a partner of opposite or unknown HIV status. Among men who denied previously knowing they were HIV-positive, 24 (9.6%) of 251 MSM who reported having UAI with an HIV-positive partner or partner of unknown HIV status compared with 11 (1.7%) of 620 MSM who denied such exposure tested HIV-positive (odds ratio=5.8, 95% confidence interval: 2.8-12.1). UAI with an HIV-positive partner or partner with unknown HIV status was 69% sensitive and 73% specific in identifying men with previously undiagnosed HIV infection; UAI regardless of partner HIV status was 80% sensitive but only 45% specific. The positive predictive value was highest for risk assessments that included partner HIV status. Integrating questions about anal sex partner HIV status and condom use identifies MSM at greatest risk for HIV acquisition and transmission. These risk criteria might be effectively used to triage MSM into more intensive prevention interventions.