Why AI Is Secretly Killing Scientific Research Funding

The Lottery Trap No One Saw Coming

Six months on a proposal. You check every comma twice, you lose sleep over the methodology section — and then you submit it into a system where your odds are worse than picking the right lottery number.

That's the new normal for thousands of scientists in Europe. And the numbers really are staggering: for the European Research Council's Starting Grant, only out of , applicants will get funded. A success rate below five percent.

But here's the twist most headlines miss. It isn't just that there are more scientists than ever. The real culprit is hiding in plain sight. Artificial intelligence has turned grant writing into a commodity, and the pipeline is now flooded with polished but hollow applications.

The Flood of Polished Nonsense

Serge Horbach from Radboud University has watched this up close. He regularly assesses applications for the Flemish provider of scientific grants, FWP. His observation is blunt, and honestly a little disturbing.

"In the past you still came across submissions in which insufficient attention had been paid to content or language use," Horbach told Trouw. "Those were easy to filter out. Now almost everything is tight and polished." Surface quality has skyrocketed. So have the numbers.

Which creates a paradox. Reviewers sift through thousands of applications that all look professional on the surface. The low-hanging fruit is gone. Every proposal reads like it came from a seasoned expert, and that makes telling true brilliance from AI-assisted mediocrity incredibly difficult.

The EU's Marie Curie program saw , submissions this year. Nearly percent more than last year. That volume surge isn't just a statistic — it's a physical burden on a review process that threatens to collapse under its own weight.

The Interview Where The Mask Slips

And this is where the system breaks. Horbach notes that while the written applications are flawless, the reality check comes at the interview. "As soon as we invite researchers for an interview, some of them fall rather short," he says.

Candidates who leaned heavily on AI to structure their arguments often can't explain the core mechanics of their own proposed research when you ask them directly. Perfect document. No deep conceptual understanding behind it.

The cost of that mismatch is high. These candidates take up interview slots that should have gone to stronger applicants who maybe didn't have the time or the skill to polish their writing as effectively. The system burns resources on false positives.

The Dutch Perspective and the NWO Warning

In the Netherlands, the National Organisation for Scientific Research (NWO) is watching closely. A spokesperson confirmed that success rates are falling sharply across various programs.

"We aim for an award rate of to percent," the spokesperson told Trouw. "But we will probably soon have to announce that for one grant the success rate has fallen below percent." That's a direct admission the system is becoming unsustainable.

Why does it matter? Because many early-career scientists depend on these grants to survive. Drop below ten percent and nine out of ten dedicated researchers are working in vain for months. The psychological and financial toll is immense.

Why This Changes More Than You Think

Most people think this is just a problem for grant reviewers. It isn't. The damage reaches much further into the fabric of scientific progress itself.

When funding becomes a lottery based on writing polish rather than research depth, the wrong projects get funded. We risk pouring billions into studies that are well-written but scientifically weak.

And that creates a perverse incentive structure. Scientists spend more time learning to prompt AI for better prose than they do on actual experimental design or data analysis. The craft of science gets replaced by the craft of marketing.

If you're interested in how data integrity issues affect scientific outcomes, take a look at how biased datasets can skew entire fields. Distorted inputs are not a problem unique to grant writing.

The Hidden Cost for Your Research Career

For individual researchers, the stakes are personal. You might have a brilliant idea that could change our understanding of natural products or polymer chemistry.

But if your proposal isn't polished to the standard set by AI tools, you might be rejected before a reviewer even reads the science. Your idea dies in a pile of thousands of similarly formatted documents.

So the filter now favors those with access to advanced AI tools, or the time to master them. It does not necessarily favor the best scientists.

We are seeing a quiet exclusion of talented researchers who are better at lab work than digital marketing. That's a loss for the entire scientific community.

What Can Be Done About This Crisis

There are no easy fixes. Grant committees can't simply ban AI usage — it's nearly impossible to detect reliably.

But we can shift the evaluation criteria. Instead of rewarding polished prose, reviewers should focus more on clarity of thought and logical structure.

Interviews need to become a larger part of the process. Face-to-face interaction reveals who truly understands their work and who is just reciting generated text.

Funding bodies might also need to increase their budgets. If applications are rising because barriers to entry are lower, the pool of winners has to grow too.

The Future of Scientific Integrity

We're at a crossroads. Let the grant system become a lottery of words rather than ideas, and we risk stagnation in the very fields that depend on innovation.

The rise of AI is not inherently bad. It can help us analyze complex data and simulate molecular interactions more efficiently.

But when it distorts the allocation of resources, we have to act. The health and well-being of our scientific community depend on a fair funding system.

If you're curious about the broader implications of AI in scientific reasoning, read how top mathematicians view AI limitations. The consensus is that human oversight remains critical.

The bottom line is simple. We need to value substance over style in scientific funding.

Until we do, the best minds might be stuck writing better prompts instead of doing better science.