- NIH’s early-career applicant pool rose 11 percent in 2025 alone, while a European study found grant application volumes rose by an average of 57 percent since 2022
- The overload extends to journal submissions, corporate hiring and policy consultations
- Elsevier traced the increase to AI-assisted submissions
Grant application volumes have risen by an average of 57 percent since 2022, straining research funders’ ability to evaluate the applications they receive, according to the latest issue of the Funding Forward newsletter from Elsevier, posted on LinkedIn Sept. 1.
The analysis argued the increase is related to the rise of AI-assisted submissions, citing a multinational European study.
The post said a similar overload has hit university journal submissions, corporate hiring and government policy consultations, where AI has lowered the barrier to producing long, technically sophisticated documents.
How Are Funders Restricting AI Use?
The first response has been to curb submissions outright. The National Institutes of Health and China’s Ministry of Science and Technology have adopted near-comprehensive bans on applications substantially developed by AI, according to the analysis.
The European Research Council and UK Research and Innovation permit the use of AI for early-stage work, such as brainstorming and drafting, provided applicants disclose its use.
The post characterized most such measures as stopgaps intended to slow the flow while longer-term approaches are developed. Many proposals involve using AI against AI, either to detect non-compliant submissions or to triage applications for human review.
What Other Approaches Are Funders Weighing?
Beyond the immediate curbs, the analysis described two further responses. One directs attention away from the application itself toward an applicant’s track record and institutional setting — publication impact, collaborative reach, career trajectory, the standing of a research team or institution. Elsevier noted this tends to reinforce existing structures, citing Spain’s La Caixa Foundation, whose models scored submissions in part by matching large language models against previously successful proposals.
The other treats AI as a way to open a system that is historically biased toward senior investigators over early-career researchers. That approach targets language barriers, regional bias and the weight given to prior grant success. Elsevier pointed to Horizon Europe’s blind first-stage evaluation, where reviewers see only the methodology and research plan, without the applicant’s name, gender or host institution.
The analysis said both have drawbacks and that AI triage systems introduce biases from their algorithms and training data, even when trained on peer-reviewed literature.
The post framed the three funder responses as sequential stages rather than competing options. What matters more, it argued, is whether the research community can agree on what good science funding should achieve and build a system around that answer.
