In this guide
Why References Matter in EU Grant Proposals
EU grant evaluators are not academic peer reviewers, but they do expect evidence. Every factual claim in your proposal — market size, technology performance, competitive landscape, regulatory timeline — needs a credible source. Unsourced claims weaken your Excellence score because they suggest either lack of rigour or unfamiliarity with the state of the art.
References serve three strategic purposes in an EU proposal. First, they demonstrate your knowledge of the field. The state-of-the-art section is explicitly evaluated on whether you have identified and understood the existing solutions, competing approaches, and relevant prior work. Citing the right papers and patents shows evaluators that you know the landscape.
Second, references establish credibility for your claims. When you state that your target market is worth EUR 4.2 billion by 2028, an evaluator will look for a source — Eurostat, Statista, Grand View Research, a Commission study, or an industry report. An unsourced market size figure is just a number you made up.
Third, references connect your project to the EU policy context. Citing the European Green Deal, the Digital Decade strategy, the EU AI Act, or specific Commission communications shows alignment with EU priorities — which directly affects your Impact score. Evaluators want to see that you understand not just your technology and market, but the policy framework your innovation operates within.
The 2026–2027 Work Programme reduced page limits to 40 pages for RIA/IA proposals. This makes every line count — and forces you to be selective about which references to include. Quality over quantity: 15 well-chosen references that directly support your key claims are better than 50 references padded to look comprehensive.
Include at least one reference to a Commission policy document (e.g. COM(2025)xxx) in your Impact section. This signals to evaluators that you have read and understood the policy context of the call. The EUR-Lex database is the official source for all EU legislation and policy communications.
Reference Formats and Placement
Horizon Europe does not mandate a specific citation format. You can use numbered references [1], author-date (Smith et al., 2024), or footnotes — whichever your team prefers. The key requirement is consistency and traceability: every reference must be complete enough for an evaluator to find the source.
For academic papers, include authors, title, journal, volume, pages, year, and DOI where available. For industry reports, include the publisher, title, year, and a URL if publicly accessible. For patents, include the patent number, title, inventor, and filing date. For EU policy documents, include the document number (e.g. COM(2024) 62 final), title, and date.
Placement matters. The most common approach is inline numbered references with a reference list at the end of the relevant section or at the end of the proposal. Footnotes work well for short asides or policy references that do not merit a full reference list entry. Avoid endnotes at the very end of the proposal — evaluators reviewing the Excellence section should not have to flip 30 pages to verify a citation.
For EIC Accelerator proposals (which are now limited to 20 pages under the 2026 Work Programme), space is precious. Use a compact reference format: number, authors (abbreviated), title, source, year. A typical EIC proposal includes 10–20 references, tightly integrated into the technical narrative.
A practical approach: maintain a running reference list in a shared document throughout the writing process. Every team member adds references as they write. Before submission, deduplicate, standardise format, and verify each entry. This prevents the scramble of assembling references in the final hours before a deadline.
Number your references sequentially as they appear in the text (not alphabetically). This makes it easier for evaluators to check a specific claim without searching through an alphabetical list. Most EU proposal templates use this convention.
Types of Evidence That Strengthen Your Proposal
References to published literature are important, but they are only one type of evidence. The strongest EU proposals layer multiple evidence types to build an unassailable case.
Primary evidence — data you have generated yourself — carries the most weight. This includes laboratory test results, prototype performance measurements, pilot deployment data, user study results, and clinical trial outcomes. Include specific numbers, not just summaries: "Our catalyst achieved 94.3% conversion efficiency at 350°C in a 500-hour continuous run" is far stronger than "Our catalyst showed excellent performance in laboratory tests."
Commercial evidence demonstrates market pull. Letters of Intent from potential customers, signed pilot agreements, pre-order commitments, partnership MOUs, and channel partner agreements all count. Evaluators treat a Letter of Intent from a named Fortune 500 customer as significantly stronger than any market report citation. If possible, include LOIs as annexes and reference them in the main text.
IP evidence demonstrates competitive moat. Filed patents (with application numbers), granted patents, registered designs, trade secrets documented in internal records, and freedom-to-operate analyses from qualified patent attorneys all strengthen the Excellence section.
Policy and regulatory evidence connects your project to EU objectives. Reference specific EU strategies (European Green Deal, Digital Decade, EU AI Act, Critical Raw Materials Act), Commission Communications, Council Conclusions, and Parliamentary Resolutions. For regulated sectors, cite the applicable regulations (MDR for medical devices, EASA for aerospace, REACH for chemicals) and explain how your project addresses compliance.
Third-party validation includes independent test reports, certifications (ISO, CE marking, sector-specific), awards, media coverage in recognised outlets, and endorsements from industry associations or standards bodies.
Create an "evidence matrix" with your key claims in rows and evidence types (primary data, commercial, IP, policy, third-party) in columns. A strong proposal has at least 2–3 evidence types supporting each major claim. Gaps in the matrix reveal where you need to strengthen your case.
Write Your Proposal with EUACC
Our AI application builder is trained on thousands of winning EU proposals. It structures your application, flags compliance issues, and generates publication-ready sections — in hours, not weeks.
Create Free AccountAI-Generated Citations: Rules and Risks in 2026
Since 2025, the Horizon Europe Standard Application Form requires applicants to "provide a list of sources used to generate/rewrite content and citations" when using AI tools. This is not optional — the form explicitly states that applicants must "double-check citations to ensure they are accurate and properly referenced."
The risk is real. A 2025 study across 13 state-of-the-art language models found hallucinated citation rates ranging from 14% to 95%. These are not obviously fake citations — AI models generate plausible author names, realistic journal titles, and convincing abstracts for papers that simply do not exist. GPTZero's analysis of 4,000+ NeurIPS 2025 papers identified hundreds of AI-hallucinated citations across at least 53 published papers.
In a Horizon Europe evaluation, a single fabricated citation can be devastating. If an evaluator is an expert in your subfield — which is likely, since experts are specifically recruited for technical evaluation — they may recognise a fake reference immediately. This destroys credibility not just for that citation but for the entire proposal, because the evaluator will question whether other claims are equally unsupported.
The safe approach: never use AI to generate references. Use AI for structuring arguments, improving clarity, and checking that all evaluation criteria are addressed — but source your references from real databases. Semantic Scholar, Google Scholar, OpenAlex, and PubMed are far more reliable for literature discovery than any generative model.
If you do use AI to find relevant literature, treat every suggested reference as unverified until you have personally confirmed: (1) the paper exists, (2) the authors are correct, (3) the publication details are accurate, and (4) the specific claim you attribute to it actually appears in the paper. Document this verification process in your AI disclosure statement.
Never copy-paste a citation from AI output directly into your proposal. Instead, take the topic or claim, search for it in Google Scholar or Semantic Scholar, find the real paper, verify the content, and create the citation yourself. This takes 2–3 minutes per reference and eliminates hallucination risk entirely.
Open Access and Data Management Requirements
Horizon Europe has mandatory open access requirements that affect how you handle publications and data arising from your project. These are not just administrative boxes to tick — they are evaluated as part of the Impact section and must be budgeted in your financial plan.
All peer-reviewed publications resulting from Horizon Europe projects must be published in open access. The Commission mandates immediate open access with no embargo period, and the publication must use a Creative Commons Attribution licence (CC BY) or equivalent. The costs of open access publication (article processing charges, or APCs) are eligible direct costs in your budget — typically EUR 2,000–5,000 per publication.
Data management follows the FAIR principles: data should be Findable, Accessible, Interoperable, and Reusable. You must submit a Data Management Plan (DMP) within 6 months of project start, and it must be updated throughout the project. The DMP describes what data you will generate, how you will store and share it, what metadata standards you will use, and any restrictions on access (for legitimate reasons such as privacy, IP protection, or security).
In your proposal, include a brief data management outline in the Implementation section. Describe the types of data you will generate, how you plan to make them FAIR, which repositories you will use (Zenodo for publications, domain-specific repositories for datasets), and how you will ensure long-term preservation. Budget for data management: storage, curation, DOI registration, and repository fees.
For EIC Accelerator proposals, the open access requirements apply to any publications arising from the grant-funded R&D activities. The equity component does not carry the same open access obligations, since commercial activities are outside the scope of the grant agreement.
Zenodo (run by CERN, free to use) is the default repository for Horizon Europe publications and datasets. Create a Zenodo community for your project at the start — this simplifies compliance tracking and demonstrates to evaluators that you have a concrete open access plan.
Building a Compelling State-of-the-Art Section
The state-of-the-art (SotA) section is where your references do the heaviest lifting. This section, typically 2–3 pages in a Horizon Europe proposal, must demonstrate that you have a comprehensive understanding of existing solutions, competing approaches, and the scientific frontier — and that your innovation goes beyond it.
Structure your SotA as a funnel: start broad (the general problem domain and why it matters), narrow to the specific technical challenge you are addressing, survey existing approaches and their limitations, and conclude with the gap that your innovation fills. Each step should be supported by references.
Avoid two common pitfalls. First, do not write a literature review. Evaluators are not looking for a comprehensive survey of the field — they want to see that you understand what exists and why it is insufficient. Focus on the 5–10 most relevant competing approaches and explain, with specific technical detail, what each one fails to achieve that your innovation addresses.
Second, do not strawman the competition. If a competing approach has real strengths, acknowledge them. Evaluators are experts — they know the field, and pretending that all existing solutions are terrible while yours is perfect undermines credibility. The strongest SotA sections honestly assess competitors' strengths and then explain, with evidence, why your approach offers a step-change improvement on the specific dimension that matters most for the target application.
End the SotA with a clear "gap statement" — a 2–3 sentence summary of what no existing solution can achieve and how your innovation closes that gap. This gap statement should directly set up the Innovation section that follows, creating a logical narrative flow that evaluators can follow without effort.
Include a comparison table summarising key parameters (performance, cost, scalability, TRL) across your solution and 3–5 main competitors. This table is one of the highest-impact elements in a proposal — evaluators can assess your competitive positioning at a glance.
Include a comparison table in your SotA section with your solution and 3–5 competitors on rows, and 4–6 key performance parameters on columns. Highlight your advantage in bold or colour. This single table often carries more persuasive weight than pages of narrative — and it survives even when evaluators are speed-reading.