STM and Ithaka S+R have launched an independent study examining how AI agents discover, access, use and attribute scholarly content.
Publishers’ technical teams, platform providers, AI developers and open repositories are being invited to contribute their experiences to the research.
The study will examine both how AI agents currently identify and interact with scholarly content and what a trusted scholarly endpoint for those systems could look like in the future.
“AI agents are becoming an increasingly important gateway to scholarly content, but the scholarly community has not yet agreed how those systems should discover, access, attribute and use research content,” said Hylke Koers, Chief Information Officer, STM Solutions.
Among the issues being investigated are what constitutes a trusted scholarly endpoint for AI agents; how publishers can identify agent-mediated activity; how provenance and attribution information can travel through AI workflows; and how consumption of scholarly content by AI systems can be measured.
STM and Ithaka S+R have launched an independent study examining how AI agents discover, access, use and attribute scholarly content.
The three-month study, called The Endpoint Question, will explore the technical infrastructure needed as AI-powered research assistants and agentic systems become an increasingly important route through which researchers access scholarly information.
Publishers’ technical teams, platform providers, AI developers and open repositories are being invited to contribute their experiences to the research.
The project forms part of TRACE, a cross-stakeholder initiative involving STM, NISO, COUNTER and other partners examining trusted retrieval and attribution in AI-mediated scholarly communication.
The study will examine both how AI agents currently identify and interact with scholarly content and what a trusted scholarly endpoint for those systems could look like in the future.
“AI agents are becoming an increasingly important gateway to scholarly content, but the scholarly community has not yet agreed how those systems should discover, access, attribute and use research content,” said Hylke Koers, Chief Information Officer, STM Solutions. “That is the question this study sets out to explore. As publishers, repositories and AI developers begin building the next generation of research tools, we need a shared understanding of what a trusted, authorised and provenance-aware interaction should look like.”
The project follows STM’s recent consultation on the responsible use of research content in generative AI. That consultation found broad support for principles including verifiability through attribution and citation, the inclusion of rebuttals, corrections and retractions, transparency, and differentiation between peer-reviewed and non-peer-reviewed content.
The Endpoint Question will consider how those principles can be implemented in practice.
Led independently by Ithaka S+R, the research will investigate how AI agents discover scholarly material, authenticate through users’ existing entitlements, receive information carrying details of its provenance, and create a credible record of content use.
It will bring together publishers, repositories, infrastructure providers and AI developers to identify where technical approaches are converging, where challenges remain and what requirements could underpin future implementation.
Among the issues being investigated are what constitutes a trusted scholarly endpoint for AI agents; how publishers can identify agent-mediated activity; how provenance and attribution information can travel through AI workflows; and how consumption of scholarly content by AI systems can be measured.
The study will also examine which approaches are most likely to be adopted by AI developers and platform providers.
“The technology is moving quickly, but much of the experimentation is happening organisation by organisation,” said Todd Toler, Principal Investigator at Ithaka S+R. “Publishers are already exploring MCP endpoints and other ways of enabling access to content, while new protocols are beginning to shape how agents discover and interact with information. This study gives the community an opportunity to examine the evidence and inform the standards and infrastructure that may follow.”
Ithaka S+R will publish the findings of the study openly.
