A new national report led by the University of Bristol reveals that generative AI (GenAI) is already being used by some UK universities to evaluate research quality, with potential to significantly reduce the time and cost of the Research Excellence Framework (REF). However, the study also uncovers deep skepticism among academics and professionals about the use of AI in this context, underscoring the urgent need for national governance and clear ethical guidelines. The REF is the UK’s primary system for assessing research quality in higher education institutions (HEIs), directly influencing the distribution of around £2 billion in public funding annually. The last assessment, REF2021, cost an estimated £471 million, with each participating university spending an average of £3 million. With REF2029 expected to be even more expensive, the report arrives at a critical moment. Lead author Professor Richard Watermeyer from the University of Bristol noted that while there is strong resistance to AI integration in some academic circles, GenAI tools are already being used—often quietly—to support REF submissions. These uses include gathering evidence of research impact, drafting impact case studies, and even reviewing or scoring research outputs. Some institutions have developed in-house AI tools, while others rely on public platforms. The report surveyed nearly 400 academics and professional services staff across 16 HEIs, including Russell Group and post-92 universities. Results showed widespread opposition, with between 54% and 75% of respondents strongly disagreeing with the use of GenAI in various aspects of the REF process. The only area with notable support—23%—was using AI to assist in developing impact case studies. Interviews with 16 university Pro Vice-Chancellors revealed mixed views. Some welcomed AI as inevitable and necessary for future readiness, urging institutions to experiment and lead. Others expressed concern about overhype, limitations in AI capabilities, and a lack of experience among staff. A recurring theme was the lack of trust in AI systems, especially among those without prior exposure. The report found that opposition is strongest in arts, humanities, and social sciences, as well as among non-users. In contrast, professional services staff and staff at less-resourced institutions tend to be more open to AI’s potential, particularly given the heavy administrative burden of REF. Co-author Professor Lawrie Phipps from Jisc stressed the need for sector-wide standards and governance. Without clear policies, institutions risk unequal access, fragmented practices, and growing disparities between well-resourced and under-resourced universities. Key recommendations include establishing and publishing institutional AI policies for research, providing comprehensive training for staff, implementing security and risk controls, and creating a national REF AI Governance Framework. The report also suggests developing a shared, high-quality AI platform accessible to all HEIs to promote fairness and reduce inequity. Professor Tom Crick from Swansea University emphasized the need for coordinated, transparent, and responsible AI use across the sector to prevent uneven advantages and ensure the REF remains credible and equitable. Globally, similar research assessment systems in Australia and New Zealand have been discontinued, highlighting the need for reform. Professor Watermeyer argued that while GenAI is not a complete solution, it cannot be ignored in the effort to modernize the REF for the data-driven age. Dr. Steven Hill from Research England called the findings a call to action—urging the sector to act with principle, collaboration, and informed critique. Professor Guy Poppy from the University of Bristol affirmed the report’s importance, noting that the UK, as a global leader in AI, has a unique opportunity to lead in shaping responsible, transparent, and rigorous research assessment.