Free scoping read. A written assessment of your manuscript in two business days, at no cost.

Start one
Contents

Academy · Publication support · Online proofing

Online Proofing Systems: The Purpose and Scope of the Proof Stage

For most journals there is one round of proofs and no second proof. The changes permitted at proof are narrower than authors commonly expect, and those permitted after publication are narrower still.

Pubrica AcademyWritten by a research writing specialist Reviewed bya research methodology reviewer Published15 October 2026 Last checked15 October 2026 Reading time14 minutes Sources13 primary

The proof is the typeset version of the accepted article, sent to the authors for checking before publication. It is a quality control step — in Elsevier’s words, “Proofing is a quality control step that involves reviewing your content to ensure it is presented exactly as intended.”3 — and for most journals it is the only such step offered to authors.

Definition of the proof stage

An online proofing system is a web application that delivers the proof and collects your corrections. Elsevier's description of its own is a fair account of what these tools do: they let authors “apply corrections, comments and change requests in your proof and submit them directly for correction”4 They also carry the typesetter's queries, and generally will not let you submit until you have dealt with them — “Corrections cannot be submitted until all queries have been answered.”4

Two practical details of such systems that catch authors out. They often disable pasting — “Proof Central does not allow for the use of copy and paste.”4 — so long corrections have to be typed or entered as comments. And submission is a one-way gate: “After this, you can no longer provide further corrections.”4

There is usually no second proof. Elsevier states that “After you have received your proof and submitted your corrections, you will not receive a further revised proof.”5 and that “When you have submitted your corrections, your paper will be finalised.”5 After that, “it will only be possible to make further changes to your paper if there is a critical error that must be corrected”5 PLOS similarly states that “There is one round of proof review”2 This is the single most important fact about the stage.

Permissible and impermissible changes at proof

The governing principle is stated most clearly by Taylor & Francis: “If there is a factual or typographical error in the proofs, please correct this but do not rewrite the text.”1 Your job is “Check your proofs against the original text for accuracy, and check that all figures, tables, etc. are correct.”1

Table 1. What may be corrected at proof stage

ChangePermitted?Notes
Typographical errorsYesThe core purpose of the stage. Taylor & Francis asks authors to correct a factual or typographical error.
Errors of scientific factYesPLOS lists errors of fact that would cause misunderstandings among permitted corrections.
Printer-introduced errorsYesErrors the production process introduced are the publisher’s responsibility to correct, and are what an erratum covers after publication.
Misspelled author names or wrong affiliationsYesPLOS lists these explicitly. Note this is correcting an error in a name, which is not the same as changing who the authors are.
Copy-editor changes you disagree withUsuallyOnline proofing systems generally allow authors to reject them, though house style on spelling and punctuation is the publisher’s.
Rewriting the textNoTaylor & Francis: correct the error but do not rewrite. This is the limit authors most often exceed.
New data, new analyses, new sectionsNoBeyond the scope of a proof. Requests beyond the permitted list cause delay, and PLOS warns of a week’s delay minimum.

PLOS's enumerated list is the most explicit published version of that table: it “PLOS permits the following types of corrections to the PDF proof:”2 “Errors in author names or affiliation”2, “Errors of scientific fact that would cause misunderstandings to readers”2 and “Printer's (introduced) errors”2 Anything further and “Correction requests beyond these items can result in a publication delay of a week minimum”2

On copy-editing. Expect the text to have changed before it reached you: “the copy editor may have altered the spelling and punctuation in your article to match the journal style”1 House style is the publisher's to set, but a copy-edit that has changed your meaning is exactly the kind of error the proof exists to catch. And responsibility sits with you — “It is your responsibility to check the proofs.”1 “We can't be held responsible for errors that remain in the published version.”1

Authorship at proof stage

Correction of a common assumption about authorship changes

The following corrects a widely held assumption, which this page also held before verification. Authorship changes are not forbidden at proof stage. Elsevier's own authorship change request form covers requests made during proofing and production and routes them to the journal manager. The bar is procedural rather than prohibitive: the form “must be signed individually by each author, including any added/removed authors”, with a stated reason and editor approval. What is in fact closed is the post-publication route — the form “should not be used for changes requested after publication”. The proof is therefore the last straightforward opportunity, not an impossible one. Correction of a misspelled author name is, moreover, distinct from a change in authorship; PLOS permits the first at proof as a matter of routine.

The procedural bar is real, though. The form “The form must be signed individually by each author, including any added/removed authors.”6 and where there is disagreement the publisher “will not proceed with the publication of your manuscript until all outstanding authorship disputes are resolved”6. Note also who will not help you resolve it: “The publisher and editor cannot investigate or mediate any authorship disputes.”6 Our guide to high-impact papers covers the four ICMJE authorship criteria that should have settled this long before the proof arrived.

Deadlines for returning proofs

Table 2. Stated proof deadlines

PublisherDeadline
Elsevier, general guidance48 hours to one week, set per journal
Elsevier Proof Central2 days
Taylor & FrancisNo figure published on its proof-checking page

There is no universal deadline, and any single figure presented as universal is unsupported. Elsevier presents 48 hours as the bottom of a range and says the deadline is set by the journal; its own proofing system asks for two days; Taylor & Francis publishes no number at all. Authors should follow their journal’s own instruction and assume that days rather than weeks are available.341

Worth planning for: Elsevier notes that most journals deliver proofs “with most delivering proofs within two weeks of the acknowledgement letter notifying final acceptance.”3 So the proof arrives at a predictable point, and the deadline is short. If you will be unreachable, say so when the article is accepted rather than when the proof lands.

Article versions and the significance of the proof stage

The standard vocabulary for article versions comes from NISO. An Accepted Manuscript is “The version of a journal article that has been accepted for publication in a journal.”7, at which point “A second party (the "publisher") takes permanent responsibility for the article.”7 A Version of Record is “A fixed version of a journal article that has been made available by any organization that acts as a publisher”7 “by formally and exclusively declaring the article "published"”7

The proof is the transition between those two. Before it, the article is yours to change through the normal editorial process. After it, the Version of Record exists, and changing it means publishing a notice.

A currency note. NISO announced in November 2020 that it had approved a work item to update this recommended practice. It could not be confirmed whether a revised version has since been published, and the 2008 document's own landing page carries no superseded notice. It is cited here as the April 2008 recommended practice, which is the version verified, rather than as the current standard.

After publication: corrections and retractions

Why a published article cannot simply be edited, in the most practical terms available: Springer Nature states that post-publication changes “changes that are made post-publication must be made through a formal correction process”9, because “the published version becomes the official record, and it is then pushed to more than 100 distributing partners”9 The record has already been copied downstream; there is no single file to fix.

ICMJE sets out what a journal should do instead. “Honest errors are a part of science and publishing and require publication of a correction when they are detected.”8 The journal should “publish a correction notice as soon as possible”8, and “The journal also should post a new article version with details of the changes from the original version”8 “Previous electronic versions should prominently note that there are more recent versions of the article.”8 and “The journal should archive all prior versions of the article.”8

Which notice. Elsevier distinguishes them by origin: “An erratum refers to a correction of errors introduced to the article by the publisher.”10 while “A corrigendum refers to a change to their article that the author wishes to publish at any time after acceptance.”10

Correction or retraction. The boundary is whether the findings survive. COPE's guidance says editors should consider retraction in defined circumstances11, and that retraction is not usually appropriate where “The main findings of the work are still reliable and correction could sufficiently address errors or concerns”11 A retraction notice should “Be clearly identified as a retraction (ie, distinct from other types of correction or comment)”11 and “Be linked to the retracted article wherever possible (ie, in all online versions)”11

A sourcing note. COPE's canonical retraction guidance page could not be accessed during verification, so the wording quoted here is from an archived revision dated 19 November 2020, which states that a newer version exists. The correction-versus-retraction principle it sets out is stable, but the live guidance should be checked before it is quoted in a submission or a policy document.

Information conveyed by CrossMark

Correction of a claim contradicted by Crossref

A widely repeated claim is contradicted by Crossref itself. The CrossMark button does not show a reader that they are looking at the current version: asked that question directly, Crossref's FAQ answers “No — it doesn't.” What the button indicates is that the publisher is maintaining the document somewhere, and the reader must click it to see the status. On a copy hosted somewhere other than the publisher, the box may warn that the copy is not being maintained and may be out of date. The indicator is useful but not self-evident at a glance.

In Crossref's own words, CrossMark “Crossmark is geared towards letting readers know about significant changes to the published literature.”12 and “A Crossmark button indicates that the publisher is maintaining the document somewhere.”12 Where something has changed, “the Crossmark box shows that an update is available”12 — but only to someone who clicks it.

Gaps in the evidence

An unquantified error rate

This page does not report how often errors are introduced or missed at the proof stage, because no source for such a figure could be identified. A search for a study measuring it found none, and any percentage quoted for it lacks a measured basis. A documented instance does, however, exist: in March 2024 it was reported that PLOS ONE “is in the process of implementing an author proof step” and “will be correcting about 1,000 papers over the next few months”, for errors in author names, affiliations, titles and references. This report concerns one journal and gives no denominator; it illustrates why the stage exists but supports no rate.

The documented instance is worth reading for what it implies about where errors accumulate: the corrections concerned “errors in author names, affiliations, titles and references”13 — metadata and references rather than results. Those are precisely the parts of a proof authors skim, because they are not the argument. They are also the parts that machines index, which is why a misspelled author name or a broken reference does more damage than it looks.

Systematic checking of a proof

Practical application

  1. Check the proof against the accepted manuscript rather than from memory, in order to identify differences introduced in production.
  2. Read the typesetter’s queries first. Most systems do not permit submission until they are answered, and they identify the substantive ambiguities.
  3. Check every author name, initial and affiliation character by character. These are the errors most often corrected after publication, and they are most easily corrected at this stage.
  4. Check the references against the authors’ own list, not only for formatting. A reference corrupted in production is easily overlooked by the authors and misleading to readers.
  5. Check numbers in tables against the source data, and confirm that each figure is the correct figure, correctly oriented, with labels legible at print size.
  6. Check special characters, units, equations, subscripts and superscripts. These elements are the most frequently corrupted in typesetting.
  7. Correct errors rather than rewriting. A request to improve the text causes delay and may be refused.
  8. If the author list requires change, initiate the publisher’s authorship change process immediately; every author must sign, and the route closes after publication.
  9. Assume that this is the only proof, and allocate time before it arrives rather than after.

Key points

The proof is a quality control step, not an opportunity for revision: authors should correct factual and typographical errors and should not rewrite. For most journals there is one round and no second proof. Deadlines are journal-specific and short; 48 hours is a minimum, not a standard. Authorship may still be changed at this stage with every author’s signature, but not after publication. Once the Version of Record exists, a change requires a published correction notice and a new version, because the article has already been distributed. The CrossMark button does not indicate to readers that they hold the current version; it directs them to check.

A web application that a publisher uses to send authors the typeset proof of their accepted article and to collect their corrections. Elsevier describes its own as letting authors “apply corrections, comments and change requests in your proof and submit them directly for correction”. It replaces annotated PDFs or marked-up paper, and it usually also carries the typesetter’s queries for authors to answer.

Quality control before publication. Elsevier describes proofing as “a quality control step that involves reviewing your content to ensure it is presented exactly as intended”. The task is to verify that the typeset article says what the accepted manuscript said, not to improve the article.

No. Taylor & Francis states this explicitly: “If there is a factual or typographical error in the proofs, please correct this but do not rewrite the text.” PLOS permits a defined list — errors in author names or affiliations, errors of scientific fact that would mislead readers, and printer-introduced errors — and warns that requests beyond those “can result in a publication delay of a week minimum”.

Yes, under certain conditions; although commonly assumed to be forbidden, it is not. Elsevier’s authorship change request form covers requests made during proofing and production, and requires that the form “be signed individually by each author, including any added/removed authors”, with a reason and editor approval. What it cannot be used for is changes “requested after publication”.

Days rather than weeks, and the figure is set by the journal. Elsevier says deadlines “typically ranging from 48 hours to one week”, and its own system asks for corrections “within 2 days”. Taylor & Francis publishes no number on its proof-checking page. A period of 48 hours should be treated as a minimum rather than a rule.

Normally not. Elsevier states that “after you have received your proof and submitted your corrections, you will not receive a further revised proof”, and that further changes are possible only “if there is a critical error that must be corrected”. PLOS states there is “one round of proof review”. For this reason the stage warrants careful attention.

The distinction lies in who introduced the error. Elsevier defines an erratum as “a correction of errors introduced to the article by the publisher” and a corrigendum as “a change to their article that the author wishes to publish at any time after acceptance”. Both are formal published notices; neither is a silent edit to the article.

No. Springer Nature states that post-publication changes “must be made through a formal correction process”, because “the published version becomes the official record, and it is then pushed to more than 100 distributing partners”. ICMJE asks journals to publish a correction notice, post a new version, flag previous versions and archive them all.

No, and Crossref states this directly. The button “indicates that the publisher is maintaining the document somewhere”, and to learn the status “a user will need to click on the button and read the status information”. It is a prompt to check, not an assurance.

References

  1. Taylor & Francis. What to look for when checking proofs. Author Services. read 15 October 2026. View on Taylor & Francis Author ServicesThe clearest statement anywhere of the line between a permitted correction and a rewrite. Also the source for where responsibility sits.
  2. PLOS. Accepted manuscripts. PLOS Biology journal information. read 15 October 2026. View on PLOSThe source for one round of review and for an enumerated list of what may be corrected. Note PLOS says final article rather than version of record.
  3. Elsevier. Proofing and publication: ensuring accuracy in your manuscript. read 15 October 2026. View on ElsevierCited for what proofing is and for the deadline range. Note this page does not state which changes are permitted, so no permitted-or-forbidden claim rests on it.
  4. Elsevier. Proof Central. Author tools and resources. read 15 October 2026. View on ElsevierA description of what an online proofing system actually does, from the publisher that runs one. The source for the two-day request and for the submission gate.
  5. Elsevier. Will I get to check my proofs again? Support Center. read 15 October 2026. View on Elsevier SupportThe source for there being no second proof. This is the fact that makes the proof stage consequential rather than routine.
  6. Elsevier. Authorship change request form. Support Hub (Journals). read 15 October 2026. View on Elsevier Support HubThe source for the correction below: authorship changes are permitted during proofing and production, under a procedural bar, and the hard cutoff is publication rather than the proof.
  7. National Information Standards Organization. Journal Article Versions (JAV): Recommendations of the NISO/ALPSP JAV Technical Working Group. NISO-RP-8-2008. Baltimore, MD: NISO, April 2008. read 15 October 2026. View on NISOThe standard vocabulary for article versions. NISO approved a work item to update it in November 2020 and we could not confirm whether a successor was published, so it is cited as the 2008 recommended practice rather than as current.
  8. International Committee of Medical Journal Editors. Corrections, Retractions, Republications and Version Control. read 15 October 2026. View on ICMJEThe strongest statement of why a published article cannot be edited silently: a correction notice, a new version, prior versions archived and flagged.
  9. Springer Nature. Post-publication journal corrections. Springer Nature Support. read 15 October 2026. View on Springer Nature SupportThe practical reason silent edits are impossible: the published version has already been distributed downstream. The source for the distributing-partners figure.
  10. Elsevier. Policy and best practice: errata and corrigenda. Publishing Ethics Resource Kit. read 15 October 2026. View on ElsevierCited only for the distinction between the two notices. It does not state a no-silent-edits rationale, so that claim rests on ICMJE and Springer Nature instead.
  11. Committee on Publication Ethics. COPE Retraction Guidelines. Revision of 19 November 2020. read 15 October 2026. View on COPE (archived revision)The canonical page could not be accessed, so this citation refers to an archived revision dated 19 November 2020, which itself notes that a newer version exists. It is cited for the boundary between correction and retraction; the current version should be checked before relying on the wording.
  12. Crossref. Crossmark FAQ. Crossref Support. read 15 October 2026. View on Crossref SupportCrossref's own FAQ. The source for the correction that the button does not by itself tell a reader they have the current version.
  13. Oransky I. Exclusive: PLOS ONE to correct 1,000 papers, add author proof step. Retraction Watch, 21 March 2024. read 15 October 2026. View on Retraction WatchJournalism rather than a study, and labelled as such on the page. Cited as a documented instance of a journal adding an author proof step because errors were reaching publication without one. It reports no denominator, so no rate is derived from it.

Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.

Related: writing an original article · what editors reject papers for.

How Pubrica can help

Pubrica’s editorial team checks proofs against the accepted manuscript — author names and affiliations, reference integrity, table figures against source data, equations and special characters — and returns a consolidated correction list within the journal’s deadline. Where a published article requires a correction notice, the team can draft it as a separate, defined task.

Get a free quote

Regulations

Check your project against this rule

  1. 1Your work
  2. 2Your details
  3. 3Send
What do you need?

Material you send is held under signed confidentiality, with access limited to the assigned team and encrypted transfer. It is not used to train anything and is not passed to third-party tools. Ask for the agreement before you send anything you would rather cover first.