The effectiveness of HIV antibody counseling and testing as a prevention intervention is limited: persons testing seronegative do not usually change their risk behaviors, some actually increase their risk behaviors, and decreases in risk behaviors are usually short-lived. Referrals to additional prevention and other needed services are therefore recommended, although the extent and determinants of referral provision for persons testing seronegative are unknown. We assessed the prevalence of referrals and the association between risk behaviors and prevention referrals among seronegatives. We reviewed HIV testing and referral data on all persons receiving confidential seronegative test results in San Francisco (SF) in the first 10 months of 1995 (n = 5,595), and gathered more detailed referral information at the municipal STD clinic from November 1995 through May 1996 (n = 747). The overall prevalence of referrals was low: a referral was given to 19.1% of the SF sample and 10.6% of the STD clinic sample; 15.4% of the SF sample and 5.9% of the STD clinic sample received a prevention referral. Injection drug users (IDUs) were the most likely to receive a prevention referral (48.5% of SF IDUs, 36.4% of STD clinic IDUs); men having sex with men and women with high-risk partners were also more likely to get a prevention referral than others. For SF IDUs, unsafe sex and needle sharing were not associated with an increased likelihood of receiving a prevention referral. Opportunities to link high-risk clients from counseling and testing to HIV prevention services are being missed. The referral component of HIV counseling and testing should be improved.