WEB surveys through online panels have become an important data collection mode in survey research. Although Internet is rapidly penetrating more and more households, online panels face a delicate problem, of panel attrition as well as turning respondents into professionals. Consequently, online panel providers strive to reduce as much as possible these phenomena using different approaches: implementing fraud detection algorithms (straight lining, digital fingerprinting), data quality modules, web layout enrichment (web design enhancements though using various HTML and web2.0 elements: background pictures, fonts & colors, dynamic web pages, etc). On one hand, all these together aim to detect bad respondents in order to be excluded from further invites and on the other hand they strive to increase good respondent’s loyalty as making web questionnaires more attractive to them. However, during this chase for respondent’s loyalty, online panel providers tend to ignore that using various visual “enhancements” might affect respondent’s perception of the questions asked on the web questionnaires which directly affects the answers they give. In this regard, I have designed an experiment to investigate the influence of three types of visual stimuli (plain text, background images and answer-lists with pictures) on the responses to a set of 18 lifestyle questions. With other words, I am trying to determine whether the online respondents tend to give different answers when facing more complex web designs (using pictures as background and/or answer options) than when responding to simple text questions. Furthermore, using this experiment, I will try to demonstrate empirically that the response process (Tourangeau, Rips, Rasinski, 2000; Cannell, Miller and Oksemberg,1981) is influenced (biased) by the suitability of pictures with question/answer texts, as items associated with less suitable images tend to be chosen less than items associated with more suitable images. Finally, I will discuss about the opportunity of using pictures (as background or associated with answer options) inside online questionnaires and its implication for online data collection.