In the “information age,” readers encounter information about socio-scientific issues repeatedly from multiple documents and sources. In turn, this information found across multiple sources may reactivate and strengthen inaccurate prior knowledge or misconceptions. Much existing theoretical and empirical work has examined how readers represent and process multiple documents and sources. Likewise, existing work has also provided an understanding of the conditions that promote the revision of preexisting misconceptions (i.e., knowledge revision) during reading. However, currently lacking is an understanding of how knowledge revision unfolds when readers engage with multiple documents from different sources. To address this gap, I present a new theoretical account that integrates key representational and processing aspects from existing accounts of multiple-document comprehension and sourcing to expand our current understanding of knowledge revision to account for multiple documents and sources (i.e., the Knowledge Revision Components Framework – Multiple Documents; KReC-MD). In a set of two experiments, I tested core hypotheses derived from KReC-MD regarding the influence of text structure, source credibility, and intertextual integration on knowledge revision. In Experiment 1, readers engaged with a set of three documents addressing misconceptions related to genetically modified organisms (GMOs) that varied in text structure (refutation vs. non-refutation) and source credibility (high vs. low). Readers demonstrated superior intertextual integration after engaging with refutation texts, as well as high-credibility sources. In turn, readers who engaged in more intertextual integration demonstrated superior knowledge revision outcomes. In addition to examining knowledge revision outcomes, it was also critical to examine the processes that readers engage in moment-by-moment during reading. Thus, in Experiment 2, I used a typed think-aloud methodology to examine the integration, sourcing, and knowledge revision processes readers engaged in during reading of refutation texts from either high- or low-credibility sources. Readers engaged in more source evaluations in the low-credibility condition, yet they demonstrated superior intertextual integration and knowledge revision in the high-credibility condition. Thus, readers may have engaged in more evaluation of low-credibility sources as a means of subsequently rejecting information from those sources. These results are discussed in light of existing research regarding multiple-document comprehension and sourcing, and critically, are used to refine the initial proposal of KReC-MD and inform future work.