Reuters Institute: how newsrooms are using AI to revive their archives
What the article is about
Published by Nieman Lab on April 7, 2026, and originally produced by the Reuters Institute for the Study of Journalism, this article covers how several news organizations are using AI to turn dormant archives into useful editorial and reader-facing resources. The organizations profiled include The Economist, Charlie Hebdo, RetroNews, Archivi.ng, The Guardian, and L’Eco di Bergamo.
Context: what makes archives a useful AI application
News archives represent decades of structured, indexed content with known publication dates, authors, and topics — properties that make them more tractable for AI applications than unstructured text. The challenge is that this material was typically stored in formats designed for human search rather than machine access, and the sheer volume makes manual curation impractical.
How specific newsrooms approached it
The Guardian built two internal AI tools. The first is a chatbot that lets journalists query the paper’s full archive directly, surfacing relevant prior coverage for a reporter working on a current story. The second generates AI-summarized overviews of recent coverage for category pages, giving readers a narrative summary of ongoing stories drawn from the archives. Both tools keep the AI in a supporting role: journalists query it and editors review the output before publication.
L’Eco di Bergamo, an Italian local newspaper, used AI to process more than 70 years of obituaries and convert them into a searchable database. Readers can look up family members, explore local history, and find connections across generations. This application is less about current journalism production and more about making a historical record accessible to the general public in a format they can navigate without editorial expertise.
Two distinct archive use cases
The article draws a practical distinction between two different archive applications that require different approaches:
The first is journalist-facing: AI tools that help reporters find prior coverage, identify historical patterns, or pull context for an active story. The goal is research acceleration, and the output is used internally.
The second is reader-facing: AI tools that surface archive material for a general audience, typically in summarized or structured form. The goal is expanding the publication’s value for readers, and the output is published.
Both require investment in metadata quality and AI infrastructure, but the editorial risks and success criteria differ. A journalist-facing tool needs to be accurate and fast. A reader-facing tool also needs to be accurate, but must also meet editorial standards for published content.
Who should read this
Editorial directors and news product managers considering how to extract value from existing archives. Also useful for journalists who want to understand how AI research tools are being built in newsrooms that have moved past experimentation into operational use.