While central to the VAA, the questionnaire with which respondents provide their positions on political statements has only been the subject of limited investigation. Moreover, these investigations either lack an experimental set-up or are restricted to a specific section of the questionnaire. This is worrying, as the design of the questionnaire may not only affect the user's response, but their eventual recommendations as well. To assess to which degree this is problematic in VAAs, this study makes use of the data-set generated by Stem-Consult, a VAA launched before the elections for the House of Representatives in the Netherlands in March 2017. Stem-Consult contained an experimental set-up in which users were randomly assigned to one of two possible versions, which were distinguished by differences in the positive or negative wording of certain statements. The effects of these changes are discussed both in terms of their influence on user responses as well as their effect on the party recommendations presented to the user. These findings can help designers of VAA to gain a deeper understanding of the way their questionnaires function and hence aid them to improve their design.