Generative artificial intelligence (GenAI) is becoming a common tool for researchers applying to highly competitive funding programmes such as Horizon Europe. But could increasingly polished proposals make it harder to identify the most promising projects? Kamila Kozirog of the European University Association argues that evaluation systems already under pressure from overwhelming demand may soon face an even greater challenge.
While the scale of genAI use in proposal writing remains unclear, its role is growing. In particular, institutions with less experience in EU programmes often lack the support offices, networks and administrative capacity needed to compete successfully for grants and turn to AI to do the heavy lifting.
This could further strain evaluation systems already struggling with high application volumes. The influx of highly refined proposals may make it harder to identify originality and scientific merit, raising the risk that strong projects are overlooked.
Unless the rules change, maintaining high-quality evaluations may require more reviewers, greater coordination and larger administrative budgets. Discussions on the next framework programme (FP10), covering 2028–34, are already underway.
You might be interested
National funding organisations are already considering solutions such as implementing two-step procedures, shortening pre-proposals and introducing interviews to ease the evaluation pressure while upholding assessment standards. Kamila Kozirog, Deputy Director of Research and Innovation at the European University Association (EUA) spoke to EU Perspectives and outlined some of the major problems.
How prevalent is the occurrence of genAI in writing proposals?
It is still difficult to measure precisely, because applicants are generally not required to disclose how they use generative AI tools during proposal preparation. However, it is becoming increasingly clear across the sector that these tools are now part of the proposal-writing environment, particularly in highly competitive programmes where researchers face strong pressure to secure funding.
The issue is probably less about fully AI-generated applications and more about the fact that these tools make it much easier to draft, revise and adapt proposals quickly. This changes the overall dynamics of competition, especially in programmes such as Horizon Europe where submission pressure is already very high.
What additional challenges does this create?
The main concern is the growing pressure this may place on evaluation processes that already operate under significant strain.
Programmes like Horizon Europe depend heavily on expert peer review. Evaluators are asked to assess scientific quality, implementation plans, expected impact and consortium credibility within relatively limited timeframes. If application numbers continue increasing substantially, this has direct operational consequences.
More proposals require more evaluators, more coordination capacity and ultimately larger administrative budgets for the European Commission and the executive agencies managing the programme. But beyond the financial aspect, there is also a quality issue. If reviewers are expected to process growing volumes of increasingly polished applications under the same conditions, maintaining the same depth and consistency of evaluation becomes more difficult.
What does this mean in terms of how to differentiate? Is there a risk worthy proposals will be overlooked?
There is certainly a risk that differentiation becomes more difficult when many applications are produced using genAI. Peer review systems work best when evaluators have sufficient time to identify originality and scientific ambition. If reviewers are confronted with very large volumes of technically strong and highly refined proposals, it becomes harder to make nuanced distinctions consistently, especially in very oversubscribed calls.
That does not mean the system should move away from excellence-based competition. But it does suggest that procedures may need to evolve in order to preserve the quality of evaluation.
What could be done in the Horizon Europe 2028-34 plans to mitigate these problems?
What FP10 will likely need to address more directly is whether the current system remains sustainable if application pressure continues to increase significantly. This is not only a question about researchers using new digital tools, but also about the overall balance between demand, success rates, evaluation capacity and available funding.
One important aspect will be ensuring that the programme has sufficient operational capacity and resources to maintain high-quality evaluation processes. If application numbers continue rising sharply, this has implications not only for reviewers, but also for the administrative and financial capacity required to run the programme effectively.
At the same time, we are also seeing that national funding organisations are beginning to reflect on similar challenges. Discussions are emerging around approaches such as stronger two-step procedures, shorter pre-proposals, interviews or other mechanisms that could help reduce pressure on evaluation systems while preserving robust assessment standards. It will be particularly interesting to see how some national agencies experiment with possible adaptations in practice, as these experiences could also inform discussions at European level over time.
Another important issue is strengthening complementary funding opportunities outside the framework programme, particularly at national level. Otherwise, excellent but unsuccessful proposals risk being continuously recycled back into the same already highly oversubscribed European competitions, further increasing pressure on the system.
