If you are an early-career researcher currently sweating over your next grant application, brace yourself. The system is buckling under the weight of AI-generated proposals, pushing success rates to historic lows. According to reports from Trouw, the European Research Councilโ€™s 2027 Starting Grant saw 7,710 applications from scientists, but only 373 will actually receive funding. That drops the success rate below five percent, a brutal statistic for anyone trying to launch a career in academia.

The Volume Surge

The influx isn't just a minor uptick; it is a flood. The EUโ€™s Marie Curie program recorded 21,627 submissions this year, marking a nearly 27 percent increase compared to last year. With the number of available grants remaining fixed, this surge in volume is mathematically guaranteeing that the vast majority of applicants will be rejected. Serge Horbach, a researcher at Radboud University who studies the scientific system, points directly to artificial intelligence as a primary driver. He notes that while journals and conferences have already been flooded with AI-written submissions, grant providers are now facing the same crisis.

Quality Illusion

The problem isn't just quantity; it's the deceptive quality of the applications. Horbach, who assesses applications for the Flemish provider FWP, explains that the bar for submission has shifted. In the past, reviewers could easily filter out submissions with poor language or insufficient content. Today, almost every application arrives looking tight and polished, masking the underlying lack of depth. This polish makes it significantly harder for reviewers to distinguish between a genuinely well-prepared scientist and someone who simply prompted an LLM to generate a convincing narrative.

The Interview Bottleneck

The real pain hits during the interview stage. Horbach reveals that when researchers are invited for interviews, some fall short because they cannot properly explain how they intend to carry out their research. These candidates have effectively taken up valuable interview slots that should have gone to stronger, more prepared applicants. The Dutch NWO spokesperson echoed these concerns, stating that while they aim for a 20 to 25 percent award rate, they expect to announce success rates falling below 10 percent for certain grants soon. The system is becoming unsustainable for everyone involved.

Key Takeaways

  • AI tools are enabling researchers to submit polished proposals at a volume that fixed funding pools cannot support.
  • The ERC 2027 Starting Grant success rate has dropped below five percent, with only 373 of 7,710 applicants funded.
  • Reviewers struggle to filter out low-effort applications because AI ensures most submissions look linguistically perfect.
  • Interview stages are becoming bottlenecks, as AI-assisted applicants often lack the deep understanding needed to defend their proposals.

The Bottom Line

We are watching the industrialization of academic bureaucracy, where AI optimizes for submission volume rather than scientific merit, leaving genuine talent stranded in a sea of synthetic polish.