Oxford University Press (OUP) has announced the start of a three-month pilot scheme which will allow Oxford students to search licensed academic content from OUP inside ChatGPT Edu. While current ChatGPT Edu is generally trained on freely available information on the internet, this project will have access to academic scholarship.
Students on the pilot project using ChatGPT Edu will receive responses informed by peer-reviewed books, journals, and reference content published by the world’s largest university press and its partners. OUP’s press release states that students in the project will receive AI outputs “grounded in trusted, cited scholarship with links to content hosted on the Oxford Academic platform”.
In emphasising trusted content, they highlight the results of a recent survey carried out by the Press which showed that 64% of researchers report benefit from using AI but only 20% trust its outputs. The pilot aims to provide a solution to this problem.
This pilot period also involves two other academic institutions: the Wharton School of the University of Pennsylvania and Harvard University. While the Oxford trial is limited to 30 students, Wharton and Harvard have opened the trial up to all students and staff.
It will last for a three-month pilot period from October to December, with the intention to gather evidence on the use of Model Context Protocol (MCP) tools that link AI models with OUP content in the real world.
In their pilot announcement last week, OUP stated that their aim is to use “both quantitative and qualitative data from the pilot to gain insight into researchers’ behaviours within AI research tools”. Oxford students involved in the project are keeping a diary to log their experiences.
OUP, which is the world’s largest university press, is especially focused on how motivated researchers are to use such tools, and how large language models (LLMs) improve their outputs when connected to OUP.
Richard Ovenden, the current Bodley Librarian at the University of Oxford, highlighted the need for collaboration between libraries and publishers to make studies as reliable as possible in the age of AI.
In OUP’s press release, Ovenden said: “This collaboration will allow us to review how readers use AI technology in their studies and to help develop approaches to ensure that students and researchers can confidently rely on the information they discover.”
The decision by publishers and libraries, like OUP and Bodleian Library, to explore AI tools comes at a time in which AI usage in research is sky-high. The prevalence of AI in student spaces has never been higher, with a Higher Education Policy Institute study in February 2025 finding that 92% of students used AI in some form, up from 66% in 2024.
This pilot follows Oxford’s recent partnership with OpenAI that allowed rare Bodleian texts to be used to train ChatGPT. The pilot scheme allows Bodleian texts, chiefly 3,500 dissertations from 1498-1884, to be digitised through OpenAI software.
The trial is made possible by a Model Context Protocol (MCP) server produced by Silverchair, an American tech company that provides digital services for publishers. Silverchair have previously collaborated with OUP to create Oxford Law Pro, an AI research assistant designed as a “specialised legal research product” based on over 9000 journal articles and 600 books provided by OUP.
After undergoing a pilot period, Oxford Law Pro is now a fully established AI tool that is available for organisations and libraries by subscription.
Silverchair told Cherwell: “Our goal is to help publishers serve peer-reviewed research through AI channels on their own terms, maintaining the integrity of the scholarly record and supporting students and other users in accessing world’s best research.”
