Abstract: This paper introduces a simple mechanism to incentivize truth-telling even when the underlying truth is unverifiable. Most similar mechanisms in the literature rely on the existence of a common prior about the distribution of possible answers and on Bayesian arguments. In our mechanism, respondents to a question bet on the answers of others, relative to the answers given to other questions. For instance, people are asked to bet whether they think a given movie will get higher ratings than another, random movie. The bet reveals whether people themselves liked the movie. We call this method “top-flop betting” and show that it provides incentives to truthfully reveal private information, even in the presence of biases in the answers the bets are based on. Unlike existing methods, our method (i) relaxes assumptions on common prior; (ii) is robust to risk aversion and certainty effects, basically requiring first-order stochastic dominance only; (iii) leads to truth-telling as a dominant strategy (not needing a Nash equilibrium).