The Growing Tide of AI in UK Grant Proposals

In the competitive world of UK higher education research funding, a new force is reshaping how academics craft their bids: artificial intelligence. Tools like ChatGPT and advanced large language models are enabling researchers at universities across the country to produce grant applications faster and in greater numbers. UK Research and Innovation (UKRI), the primary public funder of research and innovation, along with councils such as the Engineering and Physical Sciences Research Council (EPSRC), Medical Research Council (MRC), and Biotechnology and Biological Sciences Research Council (BBSRC), are reporting surges in submissions. This influx is straining peer review processes and prompting urgent adaptations in how proposals are evaluated.

Over the past three years, grant application volumes have risen by an average of 57 per cent globally, including at UK funders, according to recent analyses. For UKRI specifically, applications have increased by more than 80 per cent in the last seven years, while success rates have halved to around 26 per cent. This perfect storm of volume and quality improvements—many attributed to AI assistance—means reviewers are overwhelmed, leading to calls for innovative solutions.

UKRI's Response to the Application Surge

UKRI, which distributes over £8 billion annually, has taken proactive steps. In late 2025, it released anonymised data from thousands of past grant proposals to a team at the University of Sheffield, led by data scientist Mike Thelwall. The goal? To test if AI can predict peer review outcomes, potentially speeding up triage or acting as a supportive tool for human reviewers. Around 90 per cent of applications still require full expert review, but AI could handle initial desk rejections or provide tiebreaker insights.

Complementing this, the Open University secured £106,000 from UKRI's Metascience programme for a 12-month project starting October 2025. Collaborating with Sheffield and Salford universities, the initiative explores large language models (LLMs) as 'third reviewers' or meta-reviewers, assessing rigour, originality, and feasibility. Expected outcomes include a blueprint for AI-enhanced peer review that maintains fairness while cutting workloads—a critical need as application numbers continue climbing into 2026.

These efforts reflect broader pressures. Success rates dipping below 20 per cent in some schemes underscore the 'grant lottery' dynamic, where even strong proposals struggle. AI's role in boosting submission quality is double-edged: proposals score higher but risk homogenisation, as models optimise to funder templates.

Funder Policies on AI Use in Applications

Leading UK funders have coalesced around clear guidelines. A joint statement from NIHR, UKRI, Wellcome Trust, Cancer Research UK, British Heart Foundation, Royal Society, and others permits generative AI in application preparation—with mandatory disclosure—but bans it outright in peer review. Transparency is key: applicants must declare AI use, mitigating risks like bias, fabrication, or IP breaches.

UKRI's detailed policy emphasises research integrity principles: honesty, rigour, and accountability. Sensitive data must never feed AI tools without consent, and outputs require human verification. Breaches could lead to application rejection, funding clawback, or researcher blacklisting. This balanced approach encourages innovation while safeguarding trust. For more on these rules, see the UKRI generative AI policy.

Perspectives from UK University Researchers

At institutions like University College London (UCL) and the University of Bath, researchers report widespread AI adoption for drafting summaries, literature reviews, and impact statements. Geraint Rees, UCL's pro-vice-chancellor for research, warns that 'agentic AI' erodes distinctions between excellent and good ideas, hollowing out peer review's value. Yet, early-career researchers praise AI for levelling the playing field, allowing non-native speakers or those with heavy teaching loads to compete.

Surveys reveal near-universal student AI use (95 per cent in 2026), spilling into faculty practices. While direct stats on faculty grant AI use are emerging, the trend mirrors global patterns: faster drafting correlates with higher volumes. Universities like Sheffield are piloting AI literacy training to ensure ethical integration.

Benefits: Democratising Access to Funding

AI lowers barriers. Step-by-step: tools scan funder priorities, generate tailored narratives, refine language, and simulate reviewer critiques. For underfunded fields or regional universities, this means more competitive bids without dedicated grant writers. EU data shows Marie Curie applications up 142 per cent with fewer subpar submissions (5 per cent below threshold in 2025 vs 20 per cent in 2018). UK parallels suggest similar gains, potentially diversifying funded research.

• Time savings: Drafting cut from weeks to days.
• Bias mitigation: AI prompts inclusive language.
• Broader participation: Junior researchers submit more.

Challenges: Homogeneity, Integrity, and Detection

Drawbacks loom large. AI-optimised proposals converge on 'safe' phrasing, risking blandness. A Nature study flags AI bids as less distinct yet more fundable, potentially stifling breakthroughs. Detection lags: LLMs evade plagiarism checkers, demanding human scrutiny.

Ethical pitfalls include hallucinated data or undisclosed use, eroding trust. Funders like NIH ban AI-generated apps outright; UK opts for disclosure. Universities must train on risks, as seen in rising misconduct cases (thousands student AI cheats in 2025).

Explore the debate in this Times Higher Education analysis.

Initiatives to Harness AI in Peer Review

UKRI's projects pioneer AI support: LLMs for quality scoring, originality checks, or workload triage. Sheffield's work aims 72-95 per cent accuracy; OU tests agentic AI. La Caixa Foundation's prescreening model—90 per cent full review—offers a template.

Long-term, shift evaluation to PI track records, teams, and outputs via AI analytics. This preserves human insight for innovation while scaling reviews.

Case Studies from UK Universities

The Open University's metascience project exemplifies adaptation. At Bath, vice-chancellors urge integrity in AI use. UCL debates panel evolution. EPSRC-backed AI hubs (£60m) integrate tools ethically, funding 48 proof-of-concepts in 2025.

BBSRC and MRC see bioscience AI proposals rise, with policies mirroring UKRI. Read UKRI's peer review AI exploration.

Implications for UK Higher Education

Universities face dual pressures: more rejections amid budget squeezes (£9.2bn UKRI 2026-27, AI up to £397m by 2030), yet AI boosts competitiveness. Regional impacts vary—Russell Group leverages expertise, post-92s gain from democratisation.

Stakeholders: VCs hail efficiency; purists fear creativity loss. Cultural shift: AI literacy as core skill.

Future Outlook and Actionable Advice

By 2027, AI-peer hybrids could standardise, with £1.6bn UKRI AI investment accelerating. Recommendations:

• Disclose AI transparently.
• Human-edit for originality.
• Train on ethics/tools.
• Advocate policy evolution.

UK higher ed must embrace AI thoughtfully, ensuring innovation thrives. For grant tips, check our career advice.

Photo by Markus Winkler on Unsplash