To examine the potential role of seroprevalence surveys in HIV epidemic analysis and prediction, seroprevalence was modeled in terms of age- and time-specific incidence, while taking account of the differential inclusion of infected and uninfected individuals in serosurveys. Differential inclusion had two components, one reflecting disease progression and the other background demographic factors that could change over time. Seroprevalence and progression marker data generated by simulated epidemics showed that age dependence in HIV incidence is a key factor in data interpretation. Time trends in seroprevalence are difficult to interpret, and incidence cannot be estimated from these data unless the effects of disease progression on inclusion are taken into account. Furthermore, changes in incidence could be hard to distinguish from changes in background differential inclusion, casting doubt on the value of smaller sentinel surveys. Uncertainty about differential inclusion severely limits the value of seroprevalence data in improving the precision of HIV/AIDS prediction. Progression marker data are even harder to interpret, since changes in incidence will have effects on the marker value distribution, the size, direction, and timing of which are highly age-dependent. Independent data on differential inclusion would enhance the value of data from seroprevalence surveys.