Due to its relatively low cost and flexibility, the internet has become an increasingly attractive platform for survey practitioners. Moreover, the increasing ubiquity of high-speed internet has quelled past concerns about undercoverage of certain populations. Internet access for lower SES populations is known to be disproportionately mobile and thus dependent on smartphone or tablet-friendly instruments for success. As the dependence on mobile devices for internet access correlated with SES status, we would expect a high degree of clustering in web responses by mobile as opposed to traditional desktop computing. At question is how geographic clustering or autocorrelation in web platform and access, specifically traditional computer plus broadband vs. smartphone or tablet with mobile connections, can influence response rate, and substantive data in a web study. If certain parts of the United States tended to respond using given platform correlated with different response rates, degrees of breakoff, or substantive data, there exists a risk of spatial bias. We analyze paradata and substantive responses from a web survey of health in Illinois to understand the geographic nature of the interaction between internet platform, operating system, screen-size, and response. Specifically, use geographic information systems (GIS), local indicators of spatial autocorrelation (LISA), and spatial-lag modeling to visualize and quantify any clustering and its potential impact on survey data and response. While previous research has shown no correlation between broad band access and response rates (Fiorio et al. 2013), the current poster provides an extension by examining the impact of device ownership, mobile internet access, and internet use frequency. Our research is valuable to designers of web surveys who depend on a combination of mobile and traditional computer users.