OBJECTIVE: To explore behaviour change, baseline risk behaviour, perception of risk, HIV disclosure and life events in health centre-based voluntary counselling and testing (VCT) clients. DESIGN AND SETTING: Single-arm prospective cohort with before-after design at three (one urban and two rural) government health centres in Kenya; study duration 2 years, 1999-2001. SUBJECTS: Consecutive eligible adult clients. MAIN OUTCOME MEASURES: Numbers of sexual partners, partner type, condom use, reported symptoms of sexually transmitted infection, HIV disclosure and life events. RESULTS: High rates of enrollment and follow-up provided a demographically representative sample of 401 clients with mean time to follow-up of 7.5 months. Baseline indicators showed that clients were at higher risk than the general population, but reported a poor perception of risk. Clients with multiple partners showed a significant reduction of sexual partners at follow-up (16% to 6%; p<0.001), and numbers reporting symptoms of sexually transmitted infection decreased significantly also (from 40% to 15%; p<0.001). Condom use improved from a low baseline. Low rates of disclosure (55%) were reported by HIV-positive clients. Overall, no changes in rates of life events were seen. CONCLUSION: This study suggests that significant prevention gains can be recorded in clients receiving health centre-based VCT services in Africa. Prevention issues should be considered when refining counselling and testing policies for expanding treatment programmes.