This chapter addresses the phenomenon of speeding in web surveys as an indicator of a lack of respondent attention. Building on the response process model, we suggest that speeding implies skipping necessary steps in the process of providing valid answers to survey questions. Following this line of reasoning, we study the prevalence and impact of speeding on substantive results in a probability-based recruited online survey and a nonprobability online panel to explore whether the nature of the sample makes a difference. These samples may differ in several aspects regarding the recruitment process and the intention to participate, and thus, may vary also in the proportion of speeders. In particular, compared to freshly probability-based recruited respondents, a larger proportion of nonprobability online panels are recruited via a highly self-selective process and might thus differ in attention when completing in online surveys. To explore this question, we rely on data from online surveys in two countries: The German Longitudinal Election Study (nonprobability online panel) and the American National Election Study (freshly RDD recruitment respondents especially for the national election). Utilizing a page-specific procedure to identify speeding, we demonstrate that there are a noticeable number of respondents who “speed” through the pages of the web surveys. Moreover, speeding is somewhat more frequent in the nonprobability sample than in the freshly recruited probability-based sample. What is more, irrespective of whether utilizing data from probability-based or nonprobability online surveys, we find that excluding too fast responses does not make a considerable difference in substantive results in terms of marginal distributions and explanatory models.