A new type of political attitude survey that has gained popularity in Europe and in the United States is the voting advice application (VAA). VAAs provide users with a voting advice based on their answers to a set of attitude questions. In the calculation of this advice, no-opinion answers are excluded. We tested the hypothesis that negative VAA questions lead to more no-opinion answers than their positive equivalents. In a field experiment, visitors (N=41,505) of a VAA developed for the municipality of Utrecht in the Netherlands, were randomly guided to one of the versions of the tool in which the polarity of 16 questions was manipulated. Results do not show an overall effect of question polarity. This overall null finding appears to be caused by contrasting effects for two subtypes of negative questions: Explicit negatives (e.g. not allow) yield more no-opinion answers than their positive counterparts (e.g. allow) do, while the reverse holds for implicit negatives (e.g. forbid).