Who Counts as an Author? ICMJE and CRediT

Authorship is both credit and a public promise of accountability. Under the International Committee of Medical Journal Editors recommendations, a person should satisfy four criteria together. The person must make a substantial contribution to the work, help draft it or revise it for important intellectual content, approve the version to be published, and agree to be accountable for questions about accuracy or integrity.

The Contributor Roles Taxonomy, or CRediT, answers a related but different question. Its 14 labels describe activities such as conceptualization, data curation, formal analysis, investigation, software, supervision, and writing. A CRediT statement can show who did what. It does not decide by itself who belongs in the byline, how much work was enough, or where each name should appear.

The four ICMJE criteria#

The January 2026 ICMJE recommendations state that authorship should rest on all four of these elements:

1. Substantial contributions to the conception or design of the work, or to acquiring, analyzing, or interpreting its data.
2. Drafting the work or revising it critically for important intellectual content.
3. Final approval of the version to be published.
4. Agreement to be accountable for all aspects of the work, including appropriate investigation and resolution of questions about accuracy or integrity.

The criteria are joined by “and.” Collecting data does not automatically qualify someone. Neither does obtaining money, holding a senior title, providing general supervision, or approving a final file without a substantial intellectual role.

The standard also carries a protection against strategic exclusion. People who meet the first criterion should have the opportunity to participate in review, drafting, and final approval; a team should not deny a qualifying contributor access to the manuscript and then argue that the person failed criteria two or three. Planning authorship early makes that obligation visible.

Accountability is not a claim that every author personally repeated every calculation. ICMJE expects each author to be accountable for the author's own contributions, to know which coauthors are responsible for other parts, and to have confidence in the integrity of those contributions. If a concern arises, authors share responsibility for helping resolve it.

What belongs outside the byline#

Many essential contributions do not meet all four criteria. ICMJE lists activities such as acquiring funding, general supervision, administrative support, writing assistance, technical editing, language editing, and proofreading as insufficient on their own.

“Not an author” does not mean “not valued.” An acknowledgments or contributor section can name the work precisely: a data collector credited for investigation, a project coordinator for administration, a medical writer disclosed for writing assistance. The recognition should match the journal's policy, and acknowledged people should give permission because readers may infer that they endorse the work.

Ghost authorship hides a contributor whose role should have been disclosed or who met the authorship standard. Gift or honorary authorship adds a name that did not qualify. Both break the link between work, credit, and responsibility. One conceals influence; the other creates the appearance of oversight that may not have occurred.

CRediT provides a role map#

CRediT is an ANSI/NISO standard with 14 roles:

• conceptualization;
• data curation;
• formal analysis;
• funding acquisition;
• investigation;
• methodology;
• project administration;
• resources;
• software;
• supervision;
• validation;
• visualization;
• writing the original draft;
• writing through review and editing.

The taxonomy allows several roles per person and several people per role. The optional degree labels lead, equal, and supporting can add detail where a publisher implements them; cRediT is most useful when teams apply the role definitions consistently rather than treating them as flattering keywords.

A role statement does not replace the methods section. “Formal analysis: A.B.” identifies responsibility but does not explain the model, code, missing-data plan, or sensitivity analyses. It also does not prove that the analysis was correct. Contributorship metadata improves traceability; the scientific report still has to support the claim.

Authorship and CRediT are not interchangeable#

Some CRediT roles can qualify for the first ICMJE criterion, depending on substance. Methodology, formal analysis, investigation, or conceptualization may do so. Other roles, such as funding acquisition or general supervision, may not. A person could hold several CRediT roles and still not satisfy manuscript revision, final approval, and accountability.

The reverse also needs care. Every author should have a truthful contribution statement, but no one role is mandatory for all authors. A senior author might contribute to conception, methodology, critical revision, and supervision without writing the first draft. A statistician might contribute to methods, formal analysis, validation, visualization, and revision. Both can qualify if all four ICMJE conditions are met.

CRediT does not determine name order. First, last, and corresponding positions carry different meanings across fields, and alphabetical order is common in some disciplines. A paper should follow the journal's instructions and make local conventions explicit when order could be misunderstood. A detailed role statement reduces, but does not eliminate, ambiguity.

Corresponding authors and guarantors#

The corresponding author manages communication with the journal during submission, peer review, and publication. That role includes administrative duties such as ensuring required forms, disclosures, and materials are complete; it does not make that person solely responsible for the scientific content, and it does not lower the responsibilities of other authors.

Some journals also ask for a guarantor who accepts overall responsibility for the integrity of the work. The exact role is journal-specific. A guarantor label should not be used to turn other authors into passive names. Every listed author still must meet the applicable authorship policy.

For large collaborations, a group name may appear in the byline. The article and indexing records should identify which group members qualify as authors and which are nonauthor collaborators. The group should define who approves the final manuscript, who can answer questions, and how contributor roles are recorded. Vague collective credit can make accountability harder rather than easier.

Local contributors and research fairness#

The current ICMJE recommendations encourage collaboration and coauthorship with colleagues in locations where research is conducted, and they state that an unexplained absence of local investigators as authors should prompt questions and may lead to rejection.

This does not create automatic authorship based on geography. It recognizes a recurring integrity problem: researchers who understand the setting, participants, data generation, and local implications may be omitted while distant collaborators control the manuscript. The fair response is early, substantive collaboration with genuine opportunities to meet all criteria, not honorary addition at submission.

A prospective authorship plan should cover local leadership, interpretation, language, data governance, dissemination, capacity, and access to results. Revisit the plan when contributions change. The final record should describe actual work rather than the team's initial expectations.

AI tools cannot accept authorship#

ICMJE says authors should not list or cite AI or AI-assisted technologies as authors. A tool cannot approve a final manuscript, sign a disclosure, answer an integrity inquiry, protect participant confidentiality, or accept responsibility for an error. Human authors remain responsible for text, images, analysis, citations, and any plagiarism introduced through tool use.

Disclosure requirements depend on the journal and how the tool was used. ICMJE asks authors to disclose AI use at submission and to describe use in the appropriate section. Assistance with writing belongs in an acknowledgment. Use for data collection, analysis, or figure generation belongs in methods. Simple spelling and grammar correction may be treated differently by a journal, so its instructions are the ones that bind you.

Confidentiality is another boundary. Unpublished manuscripts, peer-review material, identifiable participant information, and proprietary data should not be entered into a system unless confidentiality and authorized use are assured. Disclosure after the fact cannot repair an unauthorized transfer.

Plan authorship before conflict appears#

At project start, create a living contribution table. Record the expected roles, the authorship standard you intend to apply, the decision rules for order, who will draft and revise, who will approve, and how you will handle a disagreement. Update it at protocol completion, after data collection, during analysis, and before submission.

Use specific evidence. “Helped with analysis” is vague. “Specified the missing-data model, wrote the analysis code, checked output against a second implementation, interpreted the sensitivity analyses, and revised the methods and results” gives you a much clearer decision.

Authorship disputes belong first with the collaborators and, if needed, their institutions. ICMJE notes that journal editors are not the appropriate arbiters because they lack the evidence to investigate contributions. Institutions can review records, agreements, drafts, code history, and correspondence.

Changes after submission require exceptional care. Adding or removing a name, or changing order, generally requires an explanation and agreement from affected authors under journal policy, and a team should not use submission pressure to force a last-minute gift name or erase a contributor.

A reader's audit trail#

When reading a paper, look beyond the name count. Find the authorship policy, CRediT statement, acknowledgments, funding, conflicts, data availability, and code availability. Check whether conceptual, analytical, validation, and writing roles are assigned. For a group byline, see whether individual members and roles are indexed.

No contribution statement proves quality. It helps you know whom to ask and whether the reported division of labor is plausible, and that transparency is valuable because modern research is often too complex for a byline alone to tell you anything.

Related guides explain why the registered protocol matters and how trial audit trails protect data integrity. The site's research overview connects these practices to accountable evidence use.

References#

1. ICMJE: Defining the Role of Authors and Contributors
2. ICMJE: Use of Artificial Intelligence in Publishing
3. NISO CRediT Contributor Roles Taxonomy
4. Definitions of the 14 CRediT roles
5. COPE discussion document on authorship and AI tools