Oxford University Press (OUP) has launched a pilot program with three academic institutions to connect their ChatGPT Edu accounts with its scholarly content. Running from October to December 2026, the initiative will allow selected students and researchers to access subscribed and purchased academic publications through AI-assisted research workflows. The participating institutions are Harvard Library, the Wharton School of the University of Pennsylvania, and the University of Oxford.
As researchers increasingly incorporate artificial intelligence (AI) tools into their academic workflows, concerns remain about the reliability of AI-generated information. A recent survey conducted by OUP found that researchers continue to express limited trust in AI outputs. The pilot aims to examine how access to authoritative scholarly content through AI tools can support research and respond to changing information discovery patterns, including the growing prevalence of zero-click searches.
The pilot uses a Model Context Protocol (MCP) server developed in collaboration with Silverchair, OUP's platform provider. The server acts as an intermediary between AI assistants, such as ChatGPT Edu, and an institution's subscribed content. When a user submits a query, the MCP server searches the available content, retrieves relevant sources, and uses them to inform the AI-generated response.
During the pilot, selected researchers and students at the three institutions will be able to use ChatGPT Edu to search and query peer-reviewed publications that their institutions have purchased or subscribed to. The resulting AI-generated responses will draw on scholarly sources and include citations and links to the corresponding content hosted on OUP's Oxford Academic platform.
The participating institutions and OUP will use quantitative and qualitative data collected during the pilot to examine how researchers interact with AI-powered research tools. The assessment will consider whether researchers are willing to use MCP connections to access trusted sources, how access to authoritative content influences the quality of large language model (LLM) outputs, and whether citations from established scholarly publishers affect how users verify research information.
Usage data will also be monitored and analyzed to assess the adoption and long-term viability of publisher MCP connectors. This information, together with feedback from faculty members and students, will help OUP and the participating institutions evaluate the potential role of MCP technology in future AI-assisted research workflows.
Click here to read the original press release.
