What Horizon Europe actually requires of your English

The short answer

There is no rule that a Horizon Europe proposal must be written in English. The official application form assumes it might not be. It tells you that if the proposal is written in another language, you have to supply an English version of the abstract:

"If the proposal is written in a language other than English, please include an English version of this abstract in the Part B (technical description) of the proposal."

That is from the standard application form for Research and Innovation Actions and Innovation Actions, version 5.1 of 22 January 2026, published on the European Commission's Funding and Tenders portal. So the formal bar is low, and almost everyone writes in English anyway, because the evaluation panels are assembled across Europe and English is the language they share.

Which means the real bar is somewhere else entirely. It is an independent expert, part way through a stack of proposals, who has to grasp your impact section on the first pass. Nothing in the rules says your English must be good. Everything in the process says it must be effortless to read.

Check your own call, not this post

One caution before anything else. The templates differ by instrument and they get revised. The version quoted above is specific to Research and Innovation Actions and Innovation Actions, and it is version 5.1; European Research Council grants, Marie Skłodowska-Curie Actions and European Innovation Council instruments each have their own forms and their own quirks.

The template attached to your call on the Funding and Tenders portal governs, and it is the only thing that does. Download it, read its instructions, and treat any blog post, including this one, as background rather than authority.

What the process actually looks like from the evaluator's side

The European Research Executive Agency describes the people who read your proposal as "independent specialists from the relevant field who peer-review and evaluate each proposal against a set of pre-defined criteria", and says the evaluation phase "lasts up to five months".

Three things follow from that, and they matter more than any rule about language.

The reader is expert but not necessarily in your subfield. A panel covers a topic, not a speciality. The person scoring your impact section may know the domain well and your particular method not at all.

They are reading many proposals. Not one, carefully, with time to reconstruct your intent. A stack, against criteria, with a scoring sheet.

The space is fixed and enforced. The same official template sets a hard limit: "The title, list of participants and sections 1, 2 and 3, together, should not be longer than 40 pages," and adds that "the page limit will be applied automatically". It also closes the obvious escape route, instructing that "experts will be instructed to ignore hyperlinks to information that is specifically designed to expand the proposal, thus circumventing the page limit".

Then it says something unusually direct for a form:

"Please, do not consider the page limit as a target! It is in your interest to keep your text as concise as possible, since experts rarely view unnecessarily long proposals in a positive light."

That is the Commission telling applicants, in its own template, that how you write affects how you are received. It is the closest thing to an official statement that language quality carries weight, and it is worth more than any amount of advice from consultants.

Where language actually decides the score

The same template explains that "Part B of the proposal is the narrative part that includes three sections that each correspond to an evaluation criterion". Those sections are Excellence, Impact, and Quality and efficiency of the implementation. Language does different work in each.

The abstract. It is the first thing read, it frames everything after it, and it is the one piece the form requires in English no matter what language the rest is in. If a reader finishes the abstract unsure what you will actually do, every later section is being read sceptically.

Excellence, and the calibration of a claim. This section asks you to describe how the project goes beyond the state of the art. That is a claim, and claims have a register. Understate it and the work reads as incremental. Overstate it and the reader starts hunting for the evidence you did not provide. Getting that dial right, in a second language, under a page limit, is a writing skill rather than a scientific one.

Impact, which is where most proposals lose points to language. This is the section most likely to be read by the panel member furthest from your speciality. Abstraction is fatal here. "The project will contribute to European competitiveness" tells a reader nothing they can score.

Implementation, where vagueness reads as unpreparedness. Work packages, roles, risks and mitigations. Weak verbs and passive constructions here create a specific impression: that the consortium has not decided who does what. The content might be perfectly well planned. The prose is what gets scored.

The register of impact writing

Two habits separate impact sections that score from ones that do not, and both are about language rather than science.

Hedge where the evidence is thin, claim where it is strong. A proposal that hedges everything reads as unsure of itself. One that hedges nothing reads as naive. The skill is doing both in the same paragraph, deliberately.

Before: "The project could potentially contribute to improved efficiency in the sector."

After: "By month 30 we will deliver a validated method and report the efficiency gain we measure, openly and with the dataset. We expect an improvement; the size of it is what the project is designed to establish."

The second version claims the deliverable confidently and hedges the result honestly. The first hedges both and therefore promises nothing.

Make every sentence answer "and therefore?" Impact is a chain: what you produce, what changes because it exists, and who is better off. Most weak impact sections stop at the first link and assume the reader will build the rest.

Before: "We will publish our results in open-access journals and present at conferences."

After: "We will publish openly so that the three standards bodies named above can cite the method in their revision cycle, which is the route by which a laboratory result becomes an industrial practice."

The first sentence describes an activity. The second describes a mechanism. Only one of them is impact.

Three things that help, and one that does not

Read the impact section aloud to someone outside your field. Mark every place they slow down or ask a question. That list is your edit list, and it is more reliable than any checklist.

Cut sentences that exist only because they sound like grant prose. Every proposal contains a few. They cost space that the page limit will not give back, and the Commission's own template says conciseness is in your interest.

Rehearse the spoken versions. Consortium calls, defending a method choice to a partner who wants their own in its place, and hearings where they occur. Written and spoken register pull from the same place, and practising one improves the other.

What does not help: translating at the end. A proposal written in one language and translated afterwards keeps the sentence shapes of the original. Evaluators feel the friction without being able to name it, and it reads as imprecision rather than as translation. If the proposal will be read in English, draft it in English, even if that is slower.

Where practice fits

This is the gap I built nicheAce: language for. Not general English, and not a writing course, but the specific register of one professional world at a time. You describe the work you actually do, and the app generates a fifteen-week journey through the vocabulary and the situations of that world: five exercises a day of about fifteen minutes, five days a week, including role-play with a character who pushes back the way a sceptical partner does. For a researcher, that means the language of call documents, hedging and claiming, reviewer comments and consortium calls, rather than the language of business in general. The language page shows how a journey is put together.

The honest version of the offer: it takes about seventy-five minutes a day for fifteen weeks, it runs on Android in the EU today, and hands-free Audio Mode is live in English, French, German, Spanish and Italian. It is also not a substitute for having something worth funding.

The rule and the bar

The rule is that you may write in another official language, provided the abstract is in English. Almost nobody uses that freedom, and the reason is not the rule. It is that the panel is European, the reading is fast, the space is capped at forty pages, and the template itself warns that experts rarely reward length.

None of that appears in a language requirement, and all of it decides scores. If you want to know how your English is doing, do not check whether it clears a threshold. Hand your impact section to a colleague in another department and watch where they stop.

If you are curious why the whole system settled on English in the first place, and what that costs the people it was not written for, that history is worth knowing too.