AI ready: how to prepare archives for Artificial Intelligence, improve discoverability, and enhance access

Artificial Intelligence (AI) and machine learning offer significant potential for archives: they can help identify sensitive content at scale, support the creation and enhancement of descriptive metadata, and enable new forms of access and interpretation. Yet AI also risks compounding long-standing challenges around archival description, bias, and power, especially where collections are fragmentary, under-described, or documented using outdated or harmful terminology.This article asks: how can archives prepare their collections for AI in ways that enhance discoverability and access, while remaining grounded in archival principles and ethical commitments? Co-authored by a computer scientist and a digital humanist with experience in cultural heritage, the article offers a cross-disciplinary literature review and practice-oriented guidelines for “AI data preparedness” in archives. Drawing on work in data documentation, data readiness, collections-as-data, and archival theory, we argue that concepts such as provenance, appraisal, authenticity, reliability, and trustworthiness should shape AI applications in records and archives.We distinguish between task-specific AI systems and generative AI, and propose workflows in which AI assists rather than replaces archivists, including description, terminology mapping, and sensitivity review. AI readiness, we argue, is an ethical and interventionist approach to data curation.