The conversation about AI in EU funding proposals has mostly been has mostly been driven by opinion. Should you use it? How much is too much? Will evaluators notice?
Until recently, much of that conversation relied on assumptions. Now, research from the European Commission’s Joint Research Centre gives us a much clearer picture.
And the findings are more nuanced than a simple argument for or against AI.
AI is already part of proposal writing
By the end of 2024, around 40% of Horizon Europe proposal abstracts showed signs of LLM modified content, according to a JRC study published in March 2026.
The research looked at firm applications to Horizon Europe and found a sharp rise in LLM assisted writing following the public release of ChatGPT in late 2022.
In other words, AI in EU proposals is already widespread.
The more interesting question is what happens when applicants rely on it, and how much that reliance matters.
What the research actually found
The JRC study found that proposals with extensive LLM generated content tended to receive lower evaluation scores and were less likely to receive funding. With partial AI assistance, however, the pattern was much weaker.
That distinction matters.
The findings suggest that using AI somewhere in the proposal process and relying heavily on AI generated writing are not the same thing. The negative pattern was strongest among proposals with more extensive AI involvement.
There is also an important caveat.
The study does not show that AI itself causes poorer results. When researchers analysed repeated submissions of the same proposals, they found no clear evidence that using LLM-assisted writing led to worse evaluation results.
Other factors may be at play. For example, the research found that LLM adoption was more common among younger and less innovative firms, as well as firms based in countries with lower English proficiency, economic development and R&D intensity.
So the takeaway is more nuanced than “AI makes proposals worse”.
Extensive AI use and stronger proposal outcomes were not strongly linked in the data. Partial assistance was a different story.
Where AI can add value
A Horizon Europe proposal requires much more than good writing.
Evaluators look at the quality of the idea, the problem being addressed, the credibility of the approach, the team’s capacity to deliver and the expected impact. That substance has to come from the people behind the project.
AI can still play a useful supporting role.
It can help structure information, clarify complex language, improve consistency or support drafting based on ideas and evidence that have already been developed. The key is what comes first.
When the concept, evidence and strategic thinking are already there, AI can help communicate them more clearly. But when AI is asked to create those foundations too, the proposal may sound convincing without having enough substance behind it.
What the Commission requires
Using generative AI to prepare a Horizon Europe proposal is permitted, but applicants remain responsible for how they use it.
Commission guidance requires applicants to carefully review and validate AI generated content, ensure compliance with intellectual property rules, and remain fully responsible for everything included in the proposal.
Applicants must also be transparent about which generative AI tools they used and how they used them.
The principle is straightforward: AI can support the preparation process, but responsibility stays with the applicant.
And AI transparency is also becoming an issue in the evaluation process..
In February 2026, the European Ombudswoman opened an inquiry into the safeguards governing the use of AI by external experts evaluating EU funding proposals. The inquiry is examining areas including existing rules, the use of third party AI tools and whether evaluators should disclose their use of AI.
So the conversation is no longer just about how applicants use AI. It is also about its role in the funding process more broadly.
What this means for your next proposal
The data does not suggest avoiding AI. It suggests being intentional about where it sits in your process.
If you are using AI to sharpen writing that already reflects clear thinking, a well-developed concept, and genuine evidence, it is a useful tool. If you are using AI to generate the thinking itself, or to fill gaps in the underlying thinking, the research suggests that it does not tend to produce better outcomes.
The proposals that consistently perform well in Horizon Europe are the ones where the thinking behind the proposal is clear. Where the problem is framed with precision, the innovation is specific and credible, and the team clearly understands what they are taking on. AI cannot replace that work. But it can help you communicate it more clearly if you have done the harder work first.
At Sploro, we work with startups and project teams at every stage of the EU funding process, from shaping the concept to submitting a proposal that communicates that concept clearly and accurately. If you are working on a Horizon Europe application and want a clearer picture of where to focus your energy, we are happy to talk!
