The roles of Generative AI in enabling or disabling inclusive research and accessible cultural heritage

Generative AI tools like ChatGPT are producing distorted accounts of 19th-century Black British history, and this project will test how to fix that. The problem is twofold. First, commercial AI models trained on mainstream datasets often omit or misrepresent marginalised histories—particularly Black British sailors in the Admiralty records held at The National Archives. Second, researchers and archivists lack clear guidance on how to use AI ethically when cataloguing and presenting such diverse histories. This PhD project combines expertise from The National Archives and the University of Roehampton to address both gaps. If successful, the research will produce practical advice for archivists, historians, and cultural institutions on how to prompt AI tools critically and how to represent underrepresented histories accurately. This could change how public-facing archives handle discovery and cataloguing—systems most people never see, but which quietly shape whose stories get told and how. The project does not aim to build new AI tools; it focuses on interrogating existing commercial ones and developing ethical guidelines for their use in cultural heritage.