From Finished Draft to Accepted Paper: The Journal Submission Process

Your manuscript leaves your hands the moment you click submit inside Editorial Manager or ScholarOne, and for a first-time corresponding author that is usually the last concrete thing about the journal submission process for weeks. Nobody emails to say an editor is reading your cover letter that same afternoon, weighing scope against fit before your paper gets anywhere near a reviewer. That first read, not the peer review everyone worries about, is the step most submissions do not survive.

Knowing what happens on the editor's side changes how you read every week of silence that follows. You have to wait longer than you think, but not as long as you fear. An editorial assistant checks that your files are complete, an editor reads your title, abstract and cover letter to judge fit, and only a fraction of what lands in that queue ever reaches a reviewer's inbox. Each stage runs on its own timeline, and the two that matter most for planning, time to first decision and time to revision, vary by field more than most guides admit.

This piece maps that sequence from the editor's side: what a screening editor actually checks, how long a first decision really takes in two different fields, the four outcomes an editor can hand back, and what changed about disclosure now that most submission systems ask directly whether you used generative AI. Narrower questions, like choosing a journal or writing a cover letter, get their own space elsewhere. What follows stays at the level of the whole process, stage by stage.

Before you submit: the journal submission checklist

Most of what determines whether your paper survives screening was decided before you opened the submission portal. A mismatched journal, chosen without cross-checking it against a journal finder for scope and typical turnaround, is not something a strong methodology section can rescue later. An abstract that buries your actual finding, or a cover letter with nobody's name on it, does similar damage before a human editor reads a word of your results. Five checks catch most of what trips up a first submission.

• Scope and fit, checked against a journal's last few issues rather than its aims and scope page alone
• Predatory warning signs: an editorial board you cannot verify, a promised decision in days, or a fee that only appears after acceptance
• A cover letter addressed to a named editor that states what is new about the paper and why this journal specifically
• An abstract that states the actual finding inside the journal's word limit, not a description of the topic
• Keywords chosen for how other researchers actually search, not restated words from the title

This isn't fancy stuff, but missing any of it means losing two or three weeks that you could have spent on your work. You can get a head start by running your draft letter through a cover letter generator built around what editors actually look for. A fill-in-the-blanks problem beats a blank-page problem at eleven at night before a deadline. It might not matter so much otherwise.

Declaring AI use: what publishers now require

Most major submission systems now ask directly whether generative AI played any role in preparing your manuscript, and that question is no longer optional guidance. It is a required field you cannot skip before the system lets you finish submitting. Editors are not asking to catch you out. They are asking because their publisher now requires a specific declaration, worded a specific way, filed in a specific place in the manuscript.

This is exactly the point where the tool you used starts to matter for what you write in that box. An AI humanizer for academic writing that rewrites sentences you already drafted, preserving your citations and your argument, is a different disclosure than a tool that generated paragraphs from a prompt, even though both are technically generative AI. The whole requirement is to name the tool accurately in the declaration, not a judgment on whether you were allowed to use it.

The wording differs enough between publishers that copying a declaration from one submission to the next can put it in the wrong section, or leave out a required phrase entirely. Three of the largest publishers make the differences easy to see side by side.

Publisher	What must be disclosed	Where it goes	What is exempt
Elsevier	Any generative AI or AI-assisted tool used in preparing the manuscript, named and described	A declaration statement placed immediately before the reference list	Basic grammar, spelling and punctuation checks
Springer Nature	AI tools with any generative, drafting or editorial role	The Methods section, or an equivalent section if the paper has none	AI-assisted copy editing, such as wording or formatting fixes
IEEE	The AI system used, which sections it touched, and how it was used	The acknowledgments section	Editing and grammar tools, though disclosing them is still recommended

None of the three publishers allows a generative AI to be listed as an author, and all three maintain that the human author is fully responsible for everything claimed in the paper, whether it's disclosed or not. This is also the position we take in our piece on where AI humanizing crosses an ethical line: editing your own argument into your own voice is normal practice. Generating the argument itself is a different act entirely, whatever the disclosure box says.

How long does the journal submission process really take?

There is no single answer, but there are real numbers by field. PLOS ONE, which reviews for technical soundness rather than novelty, reported a median time to first decision of 32 days in 2025. PLOS Medicine, reviewing clinical research, reported 67 days over the same year, more than twice as long, with PLOS Biology in between at 48 days. Field is doing real work in that gap, not just journal prestige or backlog.

Journal	Field	Median time to first decision (2025)	2025 acceptance rate
PLOS ONE	Multidisciplinary science	32 days	26%
PLOS Biology	Biology	48 days	11%
PLOS Medicine	Clinical medicine	67 days	4%

Treat any single figure you read elsewhere, including these, as a starting estimate rather than a promise. A journal's own website usually states its current average under an editorial policies or journal metrics page, and it is worth checking before you submit, not after you have been waiting eight weeks.

What happens during editorial screening?

Screening starts with an administrative check: word count, required sections, ethics statements, competing interest forms, all confirmed present before an editor even opens the file. That part is mechanical and fast. What follows is not: an editor with subject expertise reads your title, abstract and cover letter and decides whether your paper is even a candidate for review at this journal, a judgment that has nothing to do with whether your methodology is sound.

This is the step most manuscripts fail. A study of more than 128,000 submissions to 25 Nature-branded journals between 2015 and 2017 found that the big majority, between 77 and 92 percent depending on the review model, never reached a reviewer at all. eLife's own 2025 data show a similar pattern in a different way: editors sent just 35 percent of what they received out for peer review, up from 27 percent in 2024, meaning most submissions were assessed and set aside before an outside reviewer ever saw them.

The most common reasons editors cite for not sending your paper out for peer review include weak novelty, non-compliance with guidelines, scope mismatch and duplication (see UC Berkeley's own guide to the editorial process). You can control the scope of your paper more than anything else before you submit it anywhere. The first step in fixing all the things the checklist above asks about is to fix the scope.

What are reviewers actually asked to do?

Once a paper clears screening, an editor invites reviewers, usually two to three, chosen for expertise in your specific method or subfield rather than your topic broadly. Reviewers are not asked whether they like the paper. They are asked whether the question matters, whether the method actually answers it, whether the conclusions follow from the results shown, and whether anything necessary to replicate the work is missing.

PLOS gives invited reviewers two weeks to submit comments, extendable on request. It is also common for at least one invited reviewer to decline or go quiet, which is part of why first decisions stretch past the fastest realistic timeline even at journals that move quickly. None of this happens in a single block of time from your side. It happens in parallel, in inboxes you cannot see, at a pace mostly outside your control.

The four decisions an editor can return, and what happens next

An editor weighing reviewer reports has four real options: accept as is, which is rare on a first decision, invite minor revisions, invite major revisions, or reject. Minor revisions usually mean the core claim stands and reviewers want clarifications, additional citations, or small analyses added, often with no second round of external review. Major revisions mean at least one reviewer questioned something structural, a control condition, a confound, an alternative explanation, and the revised paper typically goes back to the same reviewers before anyone decides again.

Either kind of revision comes back to you as a document you did not write: a list of reviewer comments, sometimes generous, sometimes blunt, always demanding a reply to every point whether you agree with it or not. Starting from a response to reviewers generator that works directly off those comments, rather than a blank page, is the fastest way into a task most first-time authors find harder than the original writing.

A reject decision closes this submission, not your paper. Most rejected manuscripts are publishable somewhere, just not here, and the fastest path back to review is rarely the one that changes the least. Reformatting for a new journal's template is not the same task as addressing why the paper was rejected in the first place, and skipping that second part is the most common reason a resubmitted paper collects the same objections twice.

• If reviewers raised a fixable problem, fix it before you resubmit anywhere, even though nothing forces you to
• If the objection was scope, a less selective or more specialized journal in the same field is usually the faster route back
• Ask a colleague outside the project to read the reviewer comments cold, since they will spot the pattern faster than you can
• Keep the original reviewer comments on file. The next set of reviewers sometimes raises the same point, and you will already have an answer

A second rejection on the same paper is a signal worth taking seriously before a third attempt, not a reason to stop. The paper that gets published is very often not the paper that was originally submitted, and that is closer to the norm than the exception.

After acceptance: proofs, indexing, and being found

If acceptance is reached, then something else happens: copyediting, typesetting, and a set of proofs you get a short window to check, typically days rather than weeks. This is the last point where you control the exact wording of your own paper, and it's worth reading the proof against your submitted file line by line rather than skimming for typos.

None of that effort matters if nobody finds the paper afterward. Indexing in the databases your field actually searches, not just appearing on the journal's own website, depends on the same title, abstract and keywords you chose weeks earlier under submission-portal time pressure. Ten minutes spent checking that keyword list against how researchers in your field actually search is the cheapest discoverability work you will do on this paper.

Some journals now ask for a plain language summary alongside the technical abstract, aimed at readers outside your subfield entirely: funders, clinicians, patients, journalists. A plain language summary generator built for exactly that gap is worth using even when the journal does not require one, since the same summary tends to become your conference blurb, your grant impact statement, and your department newsletter entry with almost no editing.

The paper that gets cited five years from now is rarely the one with the cleanest methods section. It is the one a stranger in a different subfield can find, understand the point of within thirty seconds, and decide to read the rest of. That work starts at submission, not after acceptance, which is worth remembering the next time a blank cover letter field feels like the least important part of the form.