AIDS continues to have a devastating effect on developing countries, particularly in Sub-Saharan Africa. The lack of a proven effective vaccine to stop HIV transmission has lead to much of public policy putting an emphasis on information campaigns in order to reduce HIV-prevalence. In this paper we examine the impact of HIV/AIDSknowledge from two sides. First, we examine to what extent the campaigns have been successful at inducing the expected behavioural change with regards to HIV-related attitudes. Second, we examine the impact of HIV/AIDS knowledge on HIV status. The basic policy issue can be expressed as follows: even if individuals have acquired sufficient and necessary information on the basic facts about AIDS, factors such as innate risk attitudes or cultural background could undermine the effects of the campaigns. Using the Kenya Demographic and Health Survey (2003) we elicit empirical evidence on the relation between declared HIV/AIDS-knowledge, behavioural attitudes related to HIV/AIDS situations and the relation between knowledge and observed HIV-status. Overall, our empirical findings suggest that information campaigns have been effective at equipping the adult population in Kenya with the required knowledge to avoid becoming HIV-positive. However, when HIV-status is measured objectively we find that the relation between correctly declared attitudes and actual behaviour is only statistically significant for females that have arrived into sexuality late enough to benefit from such campaigns: it is for these females that the impact of the information campaigns has been to statistically reduce the probability of becoming HIV-positive, as intended. In the case of males we find that there is no statistical relation between either knowledge or timing of the information campaigns and a positive HIV-status. Nevertheless, another important finding refers to the selection bias induced by males who are sampled randomly but decline to take the HIV-test. The consequences of this bias are twofold; first, the estimated policy parameters for males should be interpreted with caution, but more importantly, estimating the population level HIV-prevalence for Kenyan males based on the DHS implies underestimating the true and unknown prevalence rate. Our analysis controls for individual characteristics, selection bias and endogeneity effects thus allowing us to make inferences for the full population and with regards to policy implementation.