Abstract We use an experimental approach to evaluate the effectiveness of the most commonly employed bias-mitigation tool in nonmarket valuation surveys: the cheap talk script. Our experimental design allows us to estimate treatment effects on two margins of choice separately: the decision to enter the market at all (the extensive margin) and the choices among alternatives offered (the intensive margin). The key result of this study is to show that a cheap talk script appears to affect both margins in ways distinctly different than when choices involve actual payments. Specifically, participants in hypothetical choice experiments including cheap talk are more inclined to enter the market but are also more price-sensitive as compared to when payments are real. Interestingly, the average influence of cheap talk on market participation and price-sensitiveness could result in total willingness to pay (WTP) estimates that are similar to real payment treatments since the two effects identified act in opposite directions when computing WTP. However, they may do so by inducing behavior that is distinctly different than those of consumers facing real choices. Our results highlight that future reliance on cheap talk as a bias mitigation tool requires extensive testing for empirical regularities to gain any confidence that the tool can be effective, and under what circumstances.