Due to its asynchronous, event-driven nature, JavaScript, similar to concurrent programs, suffers from the problem of data races. Past research provides methods for automatically exploring a web application to generate a trace and uses an offline dynamic race detector to find data races in the application. However, the existing random exploration techniques fail to identify races that require complex interactions with the application. While more sophisticated approaches to explore websites exist, these are not targeted towards finding data races. We conduct a study of data race bugs in open source software that shows most data race bugs are related to AJAX requests. Motivated by these findings, we present an approach for UI-level test generation which explores a website with the goal to find additional data races.