Studies evaluating public knowledge of the warning signs and risk factors associated with cancer have varied in the question format used. Those using a prompted (recognition) format have tended to find higher levels of knowledge than those using an unprompted, recall format. The aim of this study was to quantify the effect of prompting on knowledge of the seven warning signs of cancer, and risk factors for breast and bowel cancer, using data from large representative samples of the UK population. We also tested for demographic differences in the effect of prompting, hypothesizing that prompting would have the greatest impact on groups with least knowledge, specifically men, older and younger people, and those with least education. Analysis of data from four ONS surveys (total n = 5,863) demonstrated significantly higher knowledge of all signs and risk factors in the prompted compared with the unprompted condition. Contrary to our hypothesis, the pattern of interaction of prompting with gender and level of education was inconsistent, and the effect of prompting decreased with increasing age. Implications for future research on cancer knowledge and the most appropriate question format are discussed.