We proposed to integrate cognitive and social factors in the study of unprotected commercial sex. Data from 159 female entertainment workers from 15 establishments in Shanghai who reported commercial sex in the month prior to interview were used to test the approach. Two-sample t tests and multivariate logistic regression were conducted to examine if and how individual cognitive and social influence factors affect the odds of consistent condom use. About 21% of participants reported consistent condom use. Both individual cognitive and social influence factors were important correlates of condom use; models containing either alone suffered significantly in explanatory power. Perceived easiness in condom use was the most proximate determinant of condom use. It helped to translate HIV knowledge and prevention motivation into behavior and to mediate the impact of sexual relationship power. Prevention intervention is urgently needed and should focus on self-efficacy training and promoting supportive social and working environments.