Fraudulent behavior is prevalent in many markets, particularly, when behavior at the individual level is not observable. Results from aggregate analyses of experimental reporting tasks suggest that preferences for truth-telling are hetereogeneous and determined by the combination of an intrinsic motivation to be honest and the desire to be seen as honest. However, such aggregate data do not allow us to understand which truth-telling types (co-)exist and how prevalent each type is. We propose a novel reporting paradigm to measure preferences for truth-telling that keeps reports unobservable but still identifies both motives independently at the individual level. The key idea of the measure is to rely on decision makers’ intention to i) be honest and ii) be seen as honest, instead of relying on their actual reports. Results from a student and a general population sample reveal systematic heterogeneity in preference types. Using different validation procedures we find that intention-based measures may allow for the identification of underlying preferences without monitoring reports directly.