ICMJE authorship criteria are the four requirements the International Committee of Medical Journal Editors uses to define who qualifies as an author: substantial contribution to the work, drafting or critically revising it, final approval of the published version, and accountability for its accuracy. All four must be met together; meeting only some qualifies a contributor for acknowledgment, not authorship. ICMJE published a new edition of the Recommendations in January 2026, which moved artificial intelligence out from under the authorship section into a standalone Section V and added a provision on authors’ access to the underlying data.
ICMJE’s recommendations were developed for biomedical journals but are widely adopted well outside medicine as the reference definition of authorship — most journals that publish an authorship policy at all either cite ICMJE directly or restate its four criteria near-verbatim.
The four criteria, verbatim
ICMJE states the requirement as a single connected test, not four independent options:
“The ICMJE recommends that authorship be based on the following 4 criteria:
- Substantial contributions to the conception or design of the work; or the acquisition, analysis, or interpretation of data for the work; AND
- Drafting the work or reviewing it critically for important intellectual content; AND
- Final approval of the version to be published; AND
- Agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.”
The connective is AND, not OR — that single word is where most misapplications of the criteria start. Meeting three of the four does not confer authorship; it confers something else (see below).
1. Substantial contribution to conception/design, or to acquisition, analysis, or interpretation of data
This is the only one of the four criteria with an “or” inside it, and it’s deliberately broad: someone can qualify through study design, through generating the data, or through analyzing/interpreting it — they don’t need to have done all three. What ICMJE does not do is define “substantial” with a bright line (a percentage of effort, a minimum time commitment); it leaves that judgment to the author group. That ambiguity is intentional but is also the most common source of disagreement — see the CASRAI guide on resolving authorship order disputes for what to do when the author group itself can’t agree on who cleared this bar.
2. Drafting the work, or reviewing it critically for important intellectual content
ICMJE updated the wording of this criterion in its May 2023 recommendations update, changing “revising it critically” to “reviewing it critically.” The committee’s stated reason: “revising” had been read by some as requiring an author to have physically edited the draft, when the actual intent is that the person engaged substantively with the manuscript’s intellectual content and was in a position to request changes — not that a change necessarily happened. Like criterion 1, this criterion has an internal “or”: drafting the initial manuscript satisfies it on its own, as does substantive critical review of someone else’s draft.
3. Final approval of the version to be published
Every author must have seen and explicitly approved the exact version that goes to press — not an earlier draft, not the general idea of the paper. This is why authorship lists can’t reasonably be finalized before a manuscript is complete, and why adding an author late (after everyone else has already “approved” a version they haven’t seen) is a process problem, not just a formality.
4. Agreement to be accountable for all aspects of the work
This is the criterion that most clearly distinguishes authorship from a contribution, however large. It requires each author to accept responsibility for the integrity of the entire work — including parts they didn’t personally produce — and to participate in investigating and resolving any question raised about accuracy or integrity after publication. It is also the criterion most frequently cited as the reason a non-human contributor cannot be an author; see CASRAI’s guide on whether AI tools can be listed as authors for that argument in full.
All four, not a menu
A recurring misreading treats the four criteria as a checklist where meeting a majority is enough. It isn’t. ICMJE is explicit that a contributor who satisfies fewer than all four is not an author under its definition, regardless of how significant that contribution was. Someone can, for example, have generated all the data (criterion 1) and never see the manuscript again before submission — they have not met criteria 2, 3, or 4, and do not qualify as an author on ICMJE’s terms no matter how essential their data was to the paper.
What happens to a contribution that doesn’t meet all four: acknowledgment, not authorship
ICMJE addresses this directly: “Contributors who meet fewer than all 4 of the above criteria for authorship should not be listed as authors, but they should be acknowledged.” The recommendations give examples of contributions that, on their own, typically fall short of authorship: acquiring funding, general supervision of a research group, and technical writing or editing assistance. These are real, creditable contributions — they just aren’t authorship, and ICMJE’s answer is to name them in an acknowledgments section rather than the byline.
What typically does and does not clear the bar
- Typically does not qualify alone: securing grant funding for the study; general oversight of a lab or department without direct intellectual input into this specific piece of work; supplying reagents, samples, or dataset access without contributing to its design, analysis, or interpretation; copyediting or formatting a manuscript for submission; running an assay or procedure exactly to protocol with no input into study design or interpretation.
- Typically does qualify, if all four criteria are otherwise met: designing the study’s methodology; performing the statistical analysis and interpreting what it shows; writing the first draft; substantively revising a colleague’s draft in a way that shapes its intellectual content, then approving the final version and standing behind it.
The distinguishing question is never “did this person help?” — nearly everyone named in an acknowledgments section helped. It’s whether they cleared all four criteria together, including the last two: approving the exact published version and agreeing to be accountable for the whole work’s integrity, not just their own piece of it.
This binary — author or acknowledged, with nothing in between — is exactly the gap CASRAI’s CRediT taxonomy was built to close. CRediT doesn’t replace ICMJE’s authorship threshold; it sits downstream of it, letting a paper record which of the 14 contributor roles each listed author performed (and, optionally, at what degree — lead, equal, or supporting). Where ICMJE gives a yes/no gate for the byline, CRediT gives the granularity to document a contribution precisely once someone is on that byline — or, via an acknowledgments-style statement, to record a specific role for someone who isn’t. CASRAI’s CRediT guidance for authors covers how to write that statement in practice. For a direct side-by-side, see CASRAI’s CRediT vs. ICMJE comparison.
Common misapplications: honorary, guest, and ghost authorship
Because the four criteria are demanding and their application is left to author judgment, two opposite failure modes recur across the literature on authorship practice:
- Honorary, guest, or gift authorship — naming someone who did not meet the substantive criteria, typically a senior colleague, department head, or collaborator added as a courtesy or in exchange for resources or prestige. CASRAI’s Dictionary defines this precisely: an authorship is a gift when removing the named person would require no change to the manuscript — a useful, concrete test, since it doesn’t depend on knowing anyone’s intent.
- Ghost authorship — the reverse problem: someone who did make a substantial, undisclosed contribution — commonly professional medical writers on industry-sponsored trials — but is not named as an author or even in the acknowledgments. CASRAI’s Dictionary entry on ghost authorship notes the contribution would have met ICMJE’s criteria had it been disclosed; the problem is the concealment, not the underlying work.
Both are ICMJE-recognized categories of research misconduct in authorship practice, and both stem from the same root cause: treating the byline as a courtesy or a marketing decision rather than an accurate record of who did what and who is accountable for it.
Group and collaborative authorship
ICMJE’s four-criteria test doesn’t change for a three-person paper versus a three-hundred-person consortium, but the recommendations add specific guidance for large multi-author groups: “When a large multi-author group has conducted the work, the group ideally should decide who will be an author before the work is started and confirm who is an author before submitting the manuscript for publication.” Waiting until submission to sort out who qualifies is exactly the scenario this guidance is meant to prevent — see CASRAI’s guide on how large genomics consortia such as UK Biobank credit hundreds of contributors for how this plays out at scale.
Some large groups designate authorship by a group name rather than, or in addition to, an individual byline: “Some large multi-author groups designate authorship by a group name, with or without the names of individuals.” When that happens, the corresponding author is still required to identify precisely which named individuals meet all four criteria and can “take credit and responsibility for the work as authors,” as distinct from non-author contributors — sometimes called collaborators — who are named elsewhere in the paper. ICMJE ties this directly to how the paper gets indexed: “If the byline includes a group name, MEDLINE will list the names of individual group members who are authors or who are collaborators…if there is a note associated with the byline clearly stating that the individual names are elsewhere in the paper.” In practice this means a consortium paper typically needs two distinct lists — authors, who meet all four criteria and appear as such in the citation, and collaborators or consortium members, who contributed without meeting all four and are named in an appendix or supplementary note — sorted out by the group before submission, not after.
Where ICMJE’s role ends and a journal’s own policy begins
ICMJE is explicit that applying its criteria is the author group’s job, not the editor’s: “it is the collective responsibility of the authors, not the journal to which the work is submitted, to determine that all people named as authors meet all four criteria; it is not the role of journal editors to determine who qualifies or does not qualify for authorship.” ICMJE separately encourages journals to publish their own contributorship policy — which is why you’ll see individual journals ask for a CRediT statement, a specific disclosure form, or additional documentation at submission. Those are implementation details layered on top of ICMJE’s baseline definition, not a competing definition of authorship. A journal can require more disclosure than ICMJE mandates; it cannot, per ICMJE’s own framing, substitute editorial judgment for the four criteria themselves.
ICMJE addresses authorship disputes specifically, not just day-to-day determination, and its answer is the same either way: not the editor. “It is not the role of journal editors to determine who qualifies or does not qualify for authorship or to arbitrate authorship conflicts. If agreement cannot be reached about who qualifies for authorship, the institution(s) where the work was performed, not the journal editor, should be asked to investigate.” ICMJE itself has no adjudication mechanism of its own either — it sets the criteria; applying them, and resolving disagreement about who met them, is left to the authors first and the employing or hosting institution second. For what that looks like in practice, see CASRAI’s guide on resolving authorship order disputes.
Other bodies describe authorship in broadly compatible but not identical terms. The APA Publication Manual and ethics code frame authorship around substantial scientific contribution and caution against both over- and under-inclusion, without ICMJE’s specific four-part, all-must-be-met structure. NIH does not publish a separate authorship-criteria standard of its own — its guidance to investigators points to ICMJE’s recommendations directly. In practice, ICMJE’s four criteria function as the de facto reference standard even for non-medical journals that haven’t formally adopted them.
Authorship order conventions
ICMJE deliberately stays out of byline order — determining it is left to the co-authors, not editors. In the silence that leaves, discipline-specific conventions have developed:
- In biomedicine and most life sciences, the first author is conventionally the person who did the largest share of the hands-on work, and the last author position is conventionally reserved for the senior author — often the lab head or principal investigator who supervised the work and secured funding, and who frequently (though not always) also serves as corresponding author.
- Where two or more people contributed comparably, journals increasingly accept co-first authorship (and, less commonly, co-corresponding authorship), typically flagged with a footnote stating the individuals contributed equally.
- Some fields — mathematics, economics, and parts of physics among them — favor strict alphabetical order regardless of contribution size, on the reasoning that ranked order invites exactly the kind of dispute ICMJE’s criteria are meant to sidestep entirely.
None of these conventions is an ICMJE requirement, which is exactly why disagreements over order — as distinct from disagreements over who qualifies as an author at all — are common and are explicitly left to the author group to work out. CASRAI’s guide on resolving authorship order disputes covers how to do that, and the guide on corresponding author responsibilities covers the separate, ICMJE-defined duties of managing communication with the journal — a role independent of where someone sits in the byline.
Related CASRAI guidance
- Corresponding author responsibilities — the separate, ICMJE-defined role of managing communication with the journal, distinct from meeting the four authorship criteria.
- Resolving authorship order disputes — what to do when the author group agrees on who qualifies but not on byline order (ICMJE deliberately leaves order to the authors, not editors).
- Can AI be listed as an author? — why criterion 4 (accountability) is the specific reason ICMJE, COPE, and major publishers reject AI tools as authors.
- Disclosing AI assistance in a CRediT statement — how to record a human author’s use of AI tools within their own contributor role, once authorship itself is settled.
- CRediT for non-traditional contributors — how the same author/acknowledgment binary applies to community, patient, and public-involvement contributors.
Frequently asked questions
What are the ICMJE’s four criteria for authorship?
Substantial contribution to conception/design or to acquisition, analysis, or interpretation of data; drafting the work or reviewing it critically for important intellectual content; final approval of the published version; and agreement to be accountable for all aspects of the work. All four must be met — they are not alternatives.
Is there one “primary” criterion for authorship?
No single criterion is designated primary — ICMJE requires all four jointly. If forced to identify the one that does the most work distinguishing authorship from a lesser contribution, it’s usually criterion 4 (accountability), since it’s the hardest for a large contribution to satisfy without also being substantively involved throughout the work.
What happens if someone contributes but doesn’t meet all four criteria?
ICMJE’s own answer: they should not be listed as an author, but should be acknowledged. Common examples the recommendations name explicitly include funding acquisition, general supervision, and writing or technical-editing assistance.
Can a journal editor decide who qualifies as an author?
No — ICMJE states this is the author group’s collective responsibility, not the editor’s. Editors can and do ask for disclosure (a CRediT statement, a signed authorship form) to make that determination verifiable, but the determination itself is left to the authors.
Do NIH or APA use different authorship criteria than ICMJE?
NIH doesn’t publish its own separate criteria — its guidance to investigators points to ICMJE’s recommendations directly. APA’s Publication Manual and ethics code describe authorship in terms of substantial scientific contribution and caution against over- or under-inclusion, but without ICMJE’s specific four-part, all-must-be-met test. In practice the frameworks are compatible, not competing.
Are ICMJE’s authorship criteria only for medical journals?
They were written for biomedical journals, but they function as the de facto reference definition of authorship well beyond medicine — most journals across disciplines that publish an authorship policy at all either adopt ICMJE’s criteria directly or restate them closely.
Who decides authorship order?
ICMJE does not — it explicitly leaves byline order to the co-authors themselves. Conventions vary by discipline (see above), but order is a decision for the author group, not the journal.
Who resolves an authorship dispute if the author group can’t agree?
Not ICMJE, and not the journal editor. ICMJE’s recommendations state plainly that “it is not the role of journal editors…to arbitrate authorship conflicts,” and that unresolved disagreements should go to “the institution(s) where the work was performed” to investigate.
How does group or consortium authorship work under ICMJE’s criteria?
The same four criteria apply to each named author individually, even in a large group. A paper may publish under a group name, but the corresponding author must still be able to identify which named individuals meet all four criteria (authors) and which contributed without meeting all four (collaborators or non-author contributors, listed separately).
