Clear English will not rescue weak science, but unclear English can stop good science from being judged fairly. The real question is what kind of help you need, and when.
Work out what kind of help you need
"Editing" covers several different services, and paying for the wrong one wastes money and time. Providers use the labels inconsistently, so judge a service by what it says it will change, not by what it calls itself.
| Type of help | What it changes | Use it when |
|---|---|---|
| Proofreading | Spelling, punctuation, typos, consistency of terms and formatting | The English is sound and you want a final check |
| Language editing | Grammar, word choice, sentence structure, readability | The science is settled but sentences are hard to follow |
| Substantive or scientific editing | Order of the argument, clarity of claims, sometimes consistency with the data | Readers follow each sentence but lose the thread of the paper |
| Translation | Moves text written in another language into English | You write most precisely in your first language |
| Writing assistance | Drafts text from your data, outline or notes | Always acknowledged; never a substitute for the authors' own thinking |
When people talk about English language editing for manuscripts, they usually mean the second row. A simple test tells you whether that is enough. Ask a colleague in your field to read two pages. If they stumble over sentences, you need language editing. If they understand every sentence but cannot tell you what the paper found, no amount of language editing will fix it. That is a structure problem for you and your co-authors to fix first.
When editing is worth it
In a survey of 908 environmental scientists published in PLOS Biology, researchers from countries with moderate or low English proficiency who had published one English-language paper were about two and a half times as likely as native English speakers to report a paper rejected because of English writing. They were also far more likely to say they were often asked to improve their English during revision. The data are self-reported and come from one field, but the mechanism is easy to see: a reviewer who cannot follow your Methods cannot judge them.
That does not mean every manuscript needs a paid edit. Match the help to the situation:
| Situation | Reasonable next step |
|---|---|
| A fluent co-author knows the field and has time | Ask them first; they will catch terminology errors an outside editor may miss |
| Reviewers or the editor asked for language improvement | Get a thorough human edit before resubmitting, and say who did it in your response |
| No one on the team writes fluent English, and the paper has a long Methods or a subtle argument | Pay for language editing by an editor who works in your field |
| The English is good and only small slips remain | Proofread it yourself, then have one colleague read it cold |
| You drafted in another language | Translate, then have a bilingual reader in your field check the meaning against the original |
Timing matters as much as the choice. Edit once the analyses are final, the co-authors agree on the content, and you have chosen the journal, because word limits and spelling conventions (US or UK English) differ. Editing a draft that will later be restructured means paying twice. Afterwards, circulate the edited version to every author: final approval of the version to be published is one of the ICMJE authorship criteria, and an editor's changes are part of that version.
Start with the help you already have
Free help is often better help, because it comes from people who understand the science.
- Co-authors and colleagues. A fluent co-author is the best language editor you will find, since they know what each sentence is supposed to mean. A colleague from a neighbouring field is useful for a different reason: they read the way a reviewer outside your subspecialty will.
- Your institution. Check whether your research office, library or graduate school offers writing support, writing groups or editing for staff and students.
- Free networks. Rising Scholars, the network formerly called AuthorAID, offers free online mentoring, courses and resources for researchers in the Global South.
- Your own final pass. Read the text aloud. Keep one idea per sentence, keep the subject close to its verb, and use one term for each concept: do not alternate between "patients", "participants" and "subjects" for variety, because a reviewer will wonder whether they are different groups.
What AI tools can and cannot fix
The ICMJE Recommendations acknowledge, in their section on editors and AI, that many forms of AI use can be very helpful, especially for authors writing outside their primary language, and that using AI to spell-check and correct grammar is generally viewed as acceptable. The tools are good at the surface of a text: articles and prepositions, verb agreement, awkward word order, and long sentences that need splitting.
They are poor at knowing what you meant. An illustrative example:
Your draft: Mortality was lower in the intervention group, but the difference was not statistically significant (p = 0.08).
A "polished" version: The intervention significantly reduced mortality (p = 0.08).
The second sentence reads better and is wrong twice: it claims statistical significance and it turns a comparison into a causal claim. Tools can make the same kind of change to hedged conclusions, to terms such as incidence and prevalence, and to drug names, units, numbers and reference text. They also cannot fix a missing step in your reasoning, and they should never be used to push copied text below a similarity threshold. Our guide to reading a similarity report explains why rewording tools leave that problem in place.
If you use an AI tool, use it deliberately:
- Check the tool's terms before pasting anything. Elsevier's AI policy, for example, asks authors to check a tool's terms so that unpublished manuscripts stay confidential and the tool gains no rights beyond providing the service, including no right to train on your text. Never paste identifiable patient information.
- Give narrow instructions. Ask it to correct grammar and clarity only, and not to change technical terms, numbers, statistical wording or claims.
- Work one section at a time, so you can review every change.
- Compare versions with your word processor's compare function, and accept each change deliberately.
- Check every number, unit, dose, p value, name and reference against your source files.
- Keep a note of the tool, the date and the sections it touched. You will need it for the disclosure.
How to choose a legitimate editing service
Before you send a full manuscript to a paid service, confirm:
- It tells you the subject background of the editor who will work on your paper
- It states what the service includes: proofreading, language editing or substantive editing
- It returns tracked changes and queries, not only a clean file
- Its confidentiality terms are in writing, including whether AI tools are used on your file and whether your text is kept afterwards
- The price, how words are counted, and the turnaround are fixed before you pay
- It offers a short sample edit or answers questions about its edits
- It makes no promise about acceptance
Some journals point authors to particular editing services. Unless the journal says otherwise, treat that as an option rather than a requirement, and remember that no service, including one a journal lists, can guarantee acceptance.
The red flags are specific. Paper mills produce fraudulent manuscripts for profit, sell authorship positions and manipulate peer review, and they often look like editing or translation companies from the outside. Walk away from any service that:
- guarantees acceptance, or publication in a named or indexed journal by a set date
- offers to write sections, improve results, or supply data or references
- offers to add co-authors, or sells you a place on someone else's paper
- offers to submit on your behalf or asks for your submission system login
- reaches you by unsolicited email and gives no verifiable company name or address
A paper connected to a paper mill can be retracted long after publication, and the authors named on it carry the consequences.
What to tell the journal
Human editing goes in the acknowledgments. The ICMJE Recommendations list language editing and proofreading among contributions that do not qualify for authorship, and say that contributors who are not authors should be acknowledged with their contribution specified. They also advise editors to require written permission from everyone acknowledged, so ask first. A plain sentence works:
We thank [name, affiliation or company] for English-language editing of the manuscript.
If a study sponsor or another third party paid for writing support, report that as well. The ICMJE disclosure form asks authors to list all support for the present manuscript, and its examples include medical writing.
AI use needs a disclosure, and journals put the line in different places.
- The ICMJE Recommendations ask authors to describe AI use both in the cover letter and in the submitted work, with writing assistance reported in the acknowledgments. They warn that nondisclosure may require corrective action and may be construed as misconduct in some circumstances.
- Elsevier asks for a separate AI declaration statement in the manuscript, which appears in the published article. Basic grammar, spelling and punctuation checks do not need a declaration, but substantive changes to sentence structure or organization should be disclosed.
Other publishers use their own categories, and these policies are still changing. Read your target journal's current policy before you write the statement. If it is silent, follow ICMJE and disclose. A statement adapted from Elsevier's recommended format:
During the preparation of this manuscript, the authors used [tool name] to improve the grammar and readability of [sections]. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the article.
Add one matching sentence to your cover letter; our guide to writing a journal cover letter shows where declarations fit. If reviewers asked you to improve the English, say in your response letter who edited the manuscript and how the changes are marked; our guide to responding to peer review comments covers the layout.
Check the edited version before you submit
Whoever did the editing, the authors own the result. Go through the edited file once more with the original beside it:
- Every number, unit, dose and p value matches your analysis output
- Hedged claims are still hedged, and "associated with" has not become "caused"
- Technical terms, drug and gene names, and abbreviations are unchanged
- References and in-text citations are unchanged
- Spelling follows the journal's convention throughout
- Abstract and main text are still within word limits
- Every author has approved the edited version
- The acknowledgment and any AI statement are where the journal asks for them
Clear English is the last step
Language help is worth paying for when unclear sentences would stop a reviewer from judging sound work. It is not worth paying for before the science and structure are settled, and never worth buying from anyone who promises an outcome.
Directive Publications gathers its author guidelines, editorial policies and ethics statements on its For Authors page, so you can check what a journal expects before you pay anyone to edit your manuscript. When it is ready, you can submit your manuscript directly.
Frequently asked questions
Does language editing improve the chance of acceptance?
No editing service can promise acceptance, because editors decide on the question, the methods and the fit with the journal. What editing can do is remove a reason for rejection or an extra round of revision. In a published survey of environmental scientists, non-native English speakers reported having papers rejected because of English writing far more often than native speakers did.
Is it ethical to use a paid editing service for a manuscript?
Yes, as long as the service improves how your work is expressed and does not create content, change results or add authors. The ICMJE Recommendations list language editing and proofreading as contributions that do not qualify for authorship but should be acknowledged. The ethical problem lies with services that write the paper, supply data, sell author places or promise acceptance.
Do I need to disclose using AI tools for language editing?
Often yes, but the threshold and the place differ between journals. The ICMJE Recommendations ask authors to describe AI use in the cover letter and to report AI writing assistance in the acknowledgments, while Elsevier exempts basic grammar, spelling and punctuation checks but requires disclosure when a tool makes substantive changes to sentence structure or organization. Read your target journal's policy, and disclose when in doubt.
How can I tell whether an editing service is legitimate?
A legitimate service tells you what subject background your editor has, returns tracked changes, and states its confidentiality terms and price before you pay. It never promises acceptance. Walk away from any service that offers to write sections, supply data, add authors, submit on your behalf or guarantee publication in a particular journal.
Should a language editor be named in the acknowledgments?
Under the ICMJE Recommendations, yes. Language editing does not meet the authorship criteria, but contributors who are not authors should be acknowledged with their contribution specified. ICMJE also advises editors to require written permission from everyone acknowledged, so ask the editor, or the company they work for, how they agree to be named.