News sites on the World Wide Web pose challenges for information retrieval due to their dynamic content and a tendency to produce multiple versions of the same story. In this study, we advance a method we refer to as micro-longitudinal sampling that automates the capture and reduction of news site content at one-minute intervals to a manageable size for qualitative analysis. This method was used to mine text and images from the front pages of three major news sites over a three-week period, and the changes identified were manually analyzed using content analysis and critical linguistics methods. The results reveal multiple motivations for changes, only some of which, it is argued, need to be attended to in searching and archiving online news.