Author: Lustre Network

GLOW-Ready (Getting Government Archives GenAI-Ready for Users) was a research project led by Professor Lise Jaillant. Running from April to July 2026, the project built on the earlier LUSTRE and LUSTRE/GLOW projects and continued their focus on improving access to born-digital government records through collaboration between researchers, archivists, government professionals and users. It explored how generative artificial intelligence could support more responsible, transparent and user-centred access to digital government archives.

Government archives increasingly contain large volumes of emails, digital messages and other born-digital records. These collections can be difficult to search, navigate and interpret using conventional archival systems. At the same time, generative AI is changing how users expect to find and engage with information.

GLOW-Ready examined how these technologies could be used in archival discovery while addressing concerns around reliability, provenance, transparency, trust and user control. The project brought together archive and knowledge and information management professionals, government representatives, researchers, technology specialists and archive users.

Workshops

The project organised two online workshops on 18 and 25 June 2026.

The workshops brought together 155 speakers and participants from government, academia, archives, libraries, museums and other professional organisations in the UK and internationally.

Presentations and discussions explored:

• how generative AI is changing archive users’ expectations;
• the opportunities and limitations of AI-assisted discovery;
• reliability, provenance and transparency in AI-generated answers;
• user-centred approaches to archival search;
• practical examples of AI tools for digital records;
• the organisational skills and resources needed to adopt AI responsibly.

The workshops combined academic research, professional experience and practical demonstrations. They also created opportunities for participants from different sectors to discuss shared challenges and identify priorities for future work.

Workshop presentations and reports are available on the LUSTRE website.

Watch Workshop 1 presentations: GLOW-Ready Workshop 1

Watch Workshop 2 presentations: GLOW-Ready Workshop 2

Read the workshop report: GLOW-Ready Workshops Report

Research with archive users and professionals

GLOW-Ready gathered evidence through surveys, workshop discussions and follow-up interviews.

The research explored how people are currently using generative AI, what they expect from AI-supported archive services and what would make these services trustworthy and useful.

Key questions included:

• How should AI-supported archival search explain where its answers come from?
• How should supporting records and citations be presented?
• How much control should users have over generated answers and summaries?
• How can archives reduce the risk of inaccurate or unsupported responses?
• What training and organisational support do archive professionals need?

The findings informed an article on generative AI and user expectations of archival discovery, submitted to Digital Scholarship in the Humanities.

Review of AI research and practice in archives

The project also completed a systematic review of 179 publications on AI and archives published between 2020 and 2025.

The review examines how AI is currently being used in archival work and identifies gaps in research, professional skills and organisational readiness. It considers applications including description, classification, search, access and the management of large digital collections.

An article based on the review has been submitted to Archival Science.

RAG demonstrators

GLOW-Ready developed retrieval-augmented generation, or RAG, demonstrators to explore how generative AI could support access to large collections of email records.

Using the Enron email dataset, the main demonstrator allows users to ask questions in natural language and receive concise answers or summaries based on relevant retrieved emails.

The system:

• cleans and divides email records into smaller sections;
• identifies the main issue in a user’s question;
• retrieves relevant messages;
• checks the available evidence;
• generates an answer or summary;
• provides references to the supporting emails.

The demonstrator was designed to reduce unsupported responses and make AI-generated answers easier to verify.

A related model was also developed to identify spam and summarise email records. Together, the demonstrators provide practical examples of how archives could test AI-supported access within controlled environments.

Project outputs

The project produced:

• two international online workshops;
• surveys and follow-up interviews with participants;
• a study of generative AI and archival discovery;
• a systematic review of AI research and practice in archives;
• RAG demonstrators for searching and summarising email collections;
• practical findings to support future training and development in the archival sector.

The presentation slides and recordings from this workshop can be accessed here:

GLOW-Ready Workshop 7

GLOW-Ready Workshop 8

Speaker: Prof Leif Isaksen, University of Exeter 

Abstract: This presentation will make the case that very recent developments in AI ‘agency’ – the ability to use digital tools, and iterate and reason over abstract and concrete tasks – present a radically new environment for the delivery of public information resources. The cost of interface development has historically been so high that in practice data and presentational format are often tightly bound. It is rapidly becoming possible to produce bespoke, even ‘disposable’, interfaces that tailor individual or combined datasets to the specific needs of an individual or stakeholder group. This allows both for much greater agency by those users, as well as potential risks and opportunities to the content providers as material is presented in ways that may radically depart from their original intent. Here I hope to open up, rather than resolve, some of those questions as we enter an Age of Agentic AI. 

Bio: Leif Isaksen is Professor of Ancient History and Digital Humanities at the University of Exeter. His main interests are in spatial and temporal representation in the humanities – both in the ancient world and the modern one – and the use of Intelligent Systems to relate and interrogate online resources about the past. This is most notably as Director of several projects associated with the Pelagios Network, including the development of the Recogito annotation platform. He is also engaged with development of digital skills in the Arts and Humanities, specifically through the Hot Source! and DISKAH projects. Leif has been involved as an Officer, Trustee or Steering Group member of a number of Humanities and Digital Humanities activities and organisations worldwide, including CAA, EADH and ISHMap, and was Executive Board Chair of the Alliance of Digital Humanities Organisations (ADHO) from 2019 to 2021. 

Speaker: Dr Trevor Owens, American Institute of Physics 

Abstract: Integration of Generative AI tools into a wide range of consumer products is rapidly reshaping expectations for how people use search and discovery systems. At the same time, archives face daunting growing volumes of born-digital records which come with significant access challenges. Drawing on examples from scientific archives, digital preservation, and web archives, this talk offers a series of provocations about the future of digital archival access at scale. It explores how archives could support AI-assisted discovery while maintaining the transparency, provenance, and trust that are essential to public accountability and meaningful engagement with the historical record. 

Bio: Dr. Trevor Owens is Chief Research Officer at the American Institute of Physics, where he leads research programs in science policy, the history and culture of science, and the stewardship of the Niels Bohr Library & Archives. Previously, he served in leadership roles at the Library of Congress and the Institute of Museum and Library Services, supporting digital preservation, web archiving, and access to born-digital collections. He is the author of The Theory and Craft of Digital Preservation and After Disruption: A Future for Cultural Memory. 

Speaker: Neil Jakeman, King’s College London 

Abstract: Data federation challenges in the A&H research sector are largely tractable problems of coordination at scale. Through design approaches built on validated requirements from data providers and the researcher community we place the challenge of skills and training on an equal footing with the infrastructure challenges. A proof-of-concept application offers an exploratory sandbox for how to connect data services to computing power and provide enrichment pipelines in shareable workflows. 

Bio: Neil Jakeman is a Research Software Analyst at King’s Digital Lab, currently working with colleagues in Uni of Brighton to surface the needs of the A&H research community in the national roadmap for federated computing https://nfcs-networkplus.ac.uk/ 

Speaker: Prof Glen O’Hara, Oxford Brookes University

Abstract: Contemporary historians of the period since the 1990s are met with a dilemma. On the one hand, there is a wealth of material about almost everything available at their fingertips: online statistics, stored government websites, newspaper archives, opinion polling numbers, photograph libraries and the like. But on the other hand, this means that they are faced with trawling through huge amounts of information if they are to feel as if they have thoroughly covered any topic. How do we tackle this paradox? This short paper will discuss the author’s own experience in writing the history of New Labour, and consider some future remedies for information overload.  

Bio:  Glen O’Hara is Professor of Modern and Contemporary History at Oxford Brookes University. He is the author of a number of books about contemporary Britain, including The Politics of Water in Post-War Britain (2017) and most recently New Labour, New Britain? How the Blair Governments Reshaped the Country (2026). He is now working on his next book, Kingdom Come: A History of the Future in Modern Britain.

Speaker: Prof Claire Warwick, Durham University 

Abstract: Narratives about AI tend to be expressed in extremes. Overblown hype generated by tech companies is balanced by narratives of doom in the media. The policy narrative seems polarised between a desire to drive growth through ubiquitous AI, while at the same time portraying the most ubiquitous form of digital technology, social media, as a medium whose potential for harm rivals that of tobacco. 

But what do we know about how people use AI and what they think of it in a professional or scholarly context? A well-publicised study from the Harvard Business school suggests that its benefits in the workplace have not yet been proven. A study of readers of poetry suggests that participants could not tell the difference between poems written by humans and by AI. But they liked them less if they thought they were written by AI. Meanwhile social media users complain about the dominance of AI slop, suggesting that they have imbibed the negative messages about AI that are so prevalent on the press. 

Yet, at the same time, stories emerge of the way in which humans are interacting with chatbots. We ask them for advice on dating and health, use them as digital therapists, wedding planners or tour guides. Sometimes the results of this can be tragic. Yet this behaviour also speaks to a human propensity to create relationships with anything that appears to be able to communicate. Intriguingly, one of the very few studies about how humanities scholars use GenAI suggests that one of the most common uses is to test out new research ideas, in the digital equivalent of bouncing ideas off a colleague. The level of trust and relationship building here is remarkably high. This should not surprise us, since one of the things we are sure of is that users will adopt technology if it suits their needs. They also don’t care how the technology works, as long as the interface is simple and useful. Hence the popularity of GenAI. 

In this short presentation, therefore, I will be discussing what we know about the users of AI, in particular GenAI and how this relates to older knowledge about users of digital technologies. I will also consider the way in which humans tend to create relationships with non-human entities, and what this can tell us about potential futures for AI adoption. 

Bio: Claire Warwick is a Professor of Digital Humanities in the Department of English at Durham University. Her research is concerned with the way that digital resources, including artificial intelligence techniques, are used in the humanities and cultural heritage and in reading behaviour in physical and digital spaces. She has recently completed a monograph: Digital Humanities and the Cyberspace Decade: A World Elsewhere. She has led and co-investigated several digital humanities research projects, including the AEOLIAN network, which considered the potential for the application of AI to archives and cultural heritage. She collaborates widely, especially with researchers in Canada and the USA and gave the closing plenary lecture for the DH2016 conference. She has served on various advisory boards in digital humanities, for example the British Library’s BL Labs and was a member of the Conseil Scientifique du Campus Condorcet in Paris. 

She was previously Pro-Vice-Chancellor: Research at Durham from 2014- 2019 and served as chair of the Russell Group PVCRs working group. Before coming to Durham, she was Head of UCL Department of Information Studies, and director of the UCL Centre for Digital Humanities, and began her academic career as a lecturer at Sheffield University’s iSchool. Her PhD, from Cambridge, was in English Literature, followed by a postdoctoral position at Oxford University’s Humanities Computing Unit and Faculty of English. 

She has supervised PhDs on a range of topics in digital humanities, including web comics, digital newspapers, digital novels and digital publishing more generally, and the use of data science techniques for museum engagement. She would welcome enquiries from potential PhD students or visiting scholars wishing to work on a topics relevant to the application of digital techniques to literary studies, and to the humanities and cultural heritage more broadly, or to the study of digital phenomena, such as social media or AI. 

Speaker: Prof Ian Milligan, University of Waterloo 

Abstract: Each new layer of mediation between us and the historical record, from the finding aid to the digitized database to the language model, provides seemingly more comprehensive answers than the one before it. In so doing, however, it has made the limits of the record harder to see. Tools have struggled to reveal what information they have not retrieved. 

Those gaps are rarely neutral, and AI deepens them. A source is silently refused by a safety filter, and nothing marks the gap. A confident answer fills a silence the record never actually contained. The real risk is not the error you can catch but the absence you never see. A trustworthy discovery system, this keynote argues, is not the one that answers most confidently, but rather the one that can tell you what it did not retrieve, and why. 

Bio: Ian Milligan is a Professor of History at the University of Waterloo and a Fellow of the Royal Historical Society. He is the author or co-author of six books, most recently Averting the Digital Dark Age (Johns Hopkins, 2024), and co-editor of the journal Internet Histories. As PI of the Archives Unleashed project (2017–2023), he led an interdisciplinary, multi-institution team whose work is now offered as a service by the Internet Archive. On 1 July 2026 he becomes the inaugural Director of the School of Social, Political, and Historical Research at Waterloo, a new school bringing together seven departments across the humanities and social sciences. 

Speaker: Prof Paul Gooding, University of Glasgow 

Abstract: As Artificial Intelligence becomes increasingly embedded in archival practices, a growing body of scholarship has emerged to address issues such as infrastructure, governance, ethics, and the impact of AI upon archival work. However, there remains a striking absence of user studies research that seeks to understand the AI-related information behaviours and needs of both human and agentic users of archival materials. This talk will identify the implications of this ‘user studies gap’, in order to provoke discussion about the following question: who, or what, are our users in the age of AI, and what might user studies contribute to emergent debates on AI integration in archives? 

Bio: Paul Gooding is Professor of Library Studies and Digital Scholarship at the University of Glasgow. He is interested in the impact of digital collections and emerging technologies upon libraries and their users. Recently his research has explored responsible AI adoption and deployment in libraries, including topics such as inclusive requirements elicitation, understanding of transparency and openness in AI usage, and the impact of AI on library workflows and practices. His recent publications include Navigating Artificial Intelligence for Cultural Heritage Organisations (2025, UCL Press) and Library Catalogues as Data: Research, Practice and Usage (2025, Facet Publishing). 

Speaker: Dr Christopher R. Roast, Sheffield Hallam University 

Abstract: How humans interact with large scale data has many complex facets. Employing classical HCI concepts to understand interaction with large scale data, is a method of clearly examining that interaction. It provides a foundation for revealing the nature of the risks involved with the adoption of AI-based tools. By considering the opportunities offered by AI, the analysis focuses upon the user need to understand and trust the tools available. Reflections arising from this analysis include questions about human facing support rigour of process and integrity of results. We consider the specific features of some tools and suggest that usage support is of key importance. Specifically, workflow management and support become a design focus for interacting with large scale data resources. While intelligence systems intend to be just that, intelligent, their value comes from enabling transparency and empowering users in the activity of using them. These observations are relevant to not only the data and information accessed by tool users but also the broader context of the demand for data that determines role of specific users’ work. 

Bio: Dr Chris Roast has worked for over 30 years examining human factors linked with systems that empower end users. This has ranged from the analysis of the effective use of programming languages, a wide variety of systems and tools including: assistive technology; digital archives; library support systems; highly configurable tools. His work is frequently multi-disciplinary, benefiting from working with industry and across sectors and academic disciplines. He is a champion of knowledge exchange as a means of creating and articulating academic expertise and its real world relevance.