This article describes the development of a safer sex algorithm, which considers the characteristics of a woman's relationship and the HIV risk profile of herself and her primary partner. A sample of 306 low-income, predominantly African American women was recruited to participate in a study of the effectiveness of a sexual health seminar. These women were interviewed three times, at one month prior to seminar administration, three months after the seminar and again nine months after the seminar. Data from these women indicate that using an algorithm that considers the probability of HIV transmission between partners decreases the measured prevalence of unsafe sex in this population by about 17% and lowers the estimate of the average number of unsafe incidents by about four incidents in three months. The algorithm results in measures with adequate levels of temporal stability, which are similar to the more commonly used measure, vaginal or anal intercourse without a condom.