The Library of Congress (LOC) is an inherently political institution with immense reach. With 151.6 million visits and 520.3 million page views in 2022, its digital collections put the LOC’s repository of materials in the hands of users around the world, informing the kinds of narratives we tell about our past for purposes of the present. While more accessible, these collections are not always appropriately or transparently contextualized, creating significant barriers to access and often perpetuating biased or offensive language and attitudes. This matter stems from principles of provenance and metadata schemas, standards that govern how context is preserved and made available. As scholars working with digital information and literacy argue, the ubiquity of attributing authority to web-based information makes nuanced, accurate, and accessible context for digital collections increasingly necessary. Shortcomings in contemporary provenance and metadata practice are even sharper in the case of image and graphic narrative collections since prevailing descriptive standards were not designed with visual content in mind. These intersecting and at times contradictory concerns demonstrate both the complicated tension between provenance’s failures and its apparent necessity, and the ways it continues to affect applications of metadata. Exemplifying these complexities, we discuss two LOC case studies: the Webcomics Web Archive and Liljenquist Family Collection of Civil War Photographs. Illustrating the constraints of provenance and its circulation in metadata, these collections highlight the accessibility and equity issues that particularly impact visual materials.