Discussing the current AHRC/LABEX-funded EyCon (Early Conflict Photography 1890–1918 and Visual AI) project, this article considers potentially problematic metadata and how it affects the accessibility of digital visual archives. The authors deliberate how metadata creation and enrichment could be improved through Artificial Intelligence (AI) tools and explore the practical applications of AI-reliant tools to analyze a large corpus of photographs and create or enrich metadata. The amount of visual data created by digitization efforts is not always followed by the creation of contextual metadata, which is a major problem for archival institutions and their users, as metadata directly affects the accessibility of digitized records. Moreover, the scale of digitization efforts means it is often beyond the scope of archivists and other record managers to individually assess problematic or sensitive images and their metadata. Additionally, existing metadata for photographic and visual records are presenting issues in terms of outdated descriptions or inconsistent contextual information. As more attention is given to the creation of accessible digital content within archival institutions, we argue that too little is being given to the enrichment of record data. In this article, the authors ask how new tools can address incomplete or inaccurate metadata and improve the transparency and accessibility of digital visual records.