AI Preparedness Guidelines for Archivists

Practical Insights from the ARA’s New Guidelines

AI is no doubt a very hot topic right now, and here at PastView it's something that we are increasingly integrating into our software platform.

PastView’s AI and machine learning tools help users explore, understand and unlock value from their digital collections through features such as Optical Character Recognition (OCR), Handwritten Text Recognition (OCR), Audiovisual Transcription (AVT), and Facial Recognition & Automated Image Tagging. These tools can standardize and enrich metadata, make audio and video content searchable and accessible, and uncover connections across visual collections, while giving users full control over when and how AI is applied.

An essential foundation for all of this work is ensuring that content is properly migrated, ingested and structured before AI is introduced.

That's why PastView’s Data Migration and Ingest Service plays a vital role in preparing collections for advanced processing. By helping organizations move content from legacy systems into well organized and documented collections, this service supports clean data, consistent metadata, and reliable structure. These are the exact conditions the archival community needs before AI can be used effectively.

As AI becomes more embedded in archival workflows across the Galleries, Libraries, Archives and Museums (GLAM) sector, an important question follows; How can archives ensure that AI is used responsibly, transparently, and in alignment with professional standards?

The Archives and Records Association (UK and Ireland) addressed this question in its AI Preparedness Guidelines for Archivists, published in February 2026. The guidelines were authored by Professor Giovanni Colavizza and Professor Lise Jaillant as part of the FLAME project, which explores the responsible use of AI across libraries, archives and museums.

Our article below draws on those guidelines and reflects on what AI readiness means in practice for archival institutions.

AI Readiness Is About Preparation, Not Hype

One of the most important messages in the ARA guidelines is that AI readiness is not about adopting tools quickly. It is about preparing collections, documentation, governance and workflows so that AI can be used safely and effectively.

The document reframes the conversation. Rather than asking “What can AI do for us?”, archivists are encouraged to ask:

• Do we understand our collections well enough?
• Is our metadata sufficiently structured and consistent?
• Have we documented exclusions, gaps and biases?
• Do we have clear oversight and evaluation processes?

Without these foundations, AI systems risk amplifying existing weaknesses in description, context or access controls.

1. Completeness and Transparency

Archives are rarely complete. Collections may contain gaps due to loss, selection policies, historical bias or digitization priorities.

The guidelines emphasize that AI systems must not be treated as neutral interpreters of archival material. If a collection is partial or unrepresentative, an AI system trained or applied to it will reflect those limitations.

AI readiness therefore includes:

• Documenting what is included and what is excluded
• Recording known gaps or sampling decisions
• Acknowledging historical or structural bias
• Ensuring contextual information accompanies digital materials

Transparency is not optional. It is fundamental to responsible automation.

2. Metadata Quality and Context

AI tools depend on metadata. Poorly structured or inconsistent metadata reduces the reliability of automated outputs.

The guidelines highlight the importance of:

• Item level description where feasible
• Clear provenance information
• Structured relationships between records
• Consistent naming conventions
• Explicit access and restriction metadata

Narrative description also matters. Rich contextual metadata allows AI systems to generate more meaningful suggestions, whether drafting descriptions, enabling semantic search, or supporting discovery interfaces.

For organizations developing AI enabled systems, this reinforces a simple truth. AI works best when archival practice is already strong.

3. Data Formats, Structure and Derivatives

AI systems require machine readable and predictable data structures. However, this must not come at the expense of archival integrity.

The ARA guidelines recommend:

• Preserving original files in their authentic form
• Creating structured derivative copies for AI processing
• Maintaining clear links between originals and derivatives
• Using consistent and well documented file formats

This approach ensures that AI processing enhances access without altering or compromising the evidential value of archival records.

4. Human Oversight and Evaluation

AI should support professional expertise, not replace it.

The guidelines encourage defining clear use cases before deployment. For example:

• Drafting metadata to accelerate cataloging workflows
• Flagging potentially sensitive personal data for review
• Enhancing search through retrieval based systems
• Assisting with transcription or content recognition

Crucially, all outputs should be reviewed, validated and measured. Institutions should define success criteria, such as time saved, accuracy levels, or improvements in user experience.

Human oversight remains central at every stage.

Responsible AI in Practice

For technology providers and archival institutions alike, the ARA’s guidance reinforces a shared responsibility. AI tools can unlock extraordinary efficiencies and new forms of access. They can make previously hidden collections searchable, improve accessibility for users, and support sensitivity review processes.

However, these benefits depend on preparation, governance and professional judgement.

At PastView, this aligns closely with our approach. AI is designed to augment archival workflows, not override them. Users retain control over when AI is applied, how outputs are reviewed, and how metadata is managed. Responsible implementation means embedding AI within existing professional standards rather than treating it as a shortcut.

Conclusion

The AI Preparedness Guidelines for Archivists provide timely and practical direction for a sector navigating rapid technological change. They remind us that AI readiness is not a technical milestone but a professional one.

Strong documentation, structured metadata, preserved provenance, and human oversight are not optional extras. They are the conditions that make meaningful, ethical and sustainable AI integration possible.

As AI continues to evolve, archives that invest in these foundations will be best placed to harness its benefits while maintaining the trust, integrity and contextual depth that define the profession.

Interested in exploring how AI can support your archival collections responsibly?
Get in touch with the PastView team to discuss how structured ingest, strong metadata foundations and carefully applied AI tools can help prepare your archive for the future.

Credit:

This article draws on the AI Preparedness Guidelines for Archivists published by the Archives and Records Association (UK and Ireland) in February 2026, authored by Professor Giovanni Colavizza and Professor Lise Jaillant as part of the FLAME project. Readers are encouraged to consult the full guidelines for comprehensive detail and recommendations.