03 · What You Need to Know
How AI Assistance Relates to CRediT and Author Contribution Statements
What Is an Author Contribution Statement?
An author contribution statement identifies the roles performed by contributors to a scholarly publication.
Depending on the journal, it may describe responsibility for study design, data collection, analysis, manuscript preparation, supervision, or other activities.
Such statements help readers understand how the work was distributed among the researchers and may reduce ambiguity concerning individual contributions.
They do not automatically establish the quality of a contribution or determine responsibility for every research decision.
Importantly, an author contribution statement is not necessarily a complete record of every tool, service, or form of assistance involved in producing the manuscript.
What Is the CRediT Taxonomy?
The Contributor Roles Taxonomy (CRediT), maintained through the National Information Standards Organization, provides a standardized vocabulary for describing contributions to scholarly work.
It includes 14 roles:
- Conceptualization
- Data curation
- Formal analysis
- Funding acquisition
- Investigation
- Methodology
- Project administration
- Resources
- Software
- Supervision
- Validation
- Visualization
- Writing – original draft
- Writing – review & editing
These categories provide a common vocabulary for documenting different kinds of scholarly work.
CRediT does not, however, replace a journal's authorship criteria. It is a contributorship classification system rather than an independent test of who qualifies for the author byline.
Why Is Contributorship Different From Authorship?
Contributorship describes what was done. Authorship determines who qualifies for formal scholarly recognition and responsibility.
A person may contribute to a project without satisfying every requirement for authorship. For example, technical assistance or language editing may be acknowledged without resulting in an author byline.
Similarly, a software tool may assist with analysis or visualization without becoming a responsible academic author.
CRediT contribution role
A standardized description of an activity performed in producing scholarly work.
Academic authorship eligibility
A determination that a contributor satisfies the intellectual, approval, and accountability requirements established by the applicable publication policy.
This distinction explains why assigning an activity label to an AI system does not resolve whether AI can satisfy academic authorship criteria.
Can AI Perform Tasks Resembling CRediT Roles?
Yes. Generative AI systems can assist with several activities that resemble CRediT contribution categories.
For example, an AI tool may generate preliminary manuscript text, recommend analytical code, suggest data visualizations, or propose methodological alternatives.
However, the resemblance between a task and a CRediT category does not establish that the system should receive a formal contributor record.
| CRediT Role |
Possible AI Assistance |
Reporting Consideration |
| Writing – original draft |
Generating preliminary manuscript paragraphs. |
Disclose substantive drafting according to the journal's AI policy. |
| Writing – review & editing |
Suggesting revisions to structure or language. |
Distinguish routine editing from substantive generative assistance. |
| Formal analysis |
Generating analytical code or preliminary interpretations. |
Describe consequential methodological assistance and verification. |
| Software |
Assisting with the development of research code. |
Document relevant software development and testing. |
| Visualization |
Generating chart code or suggesting visual representations. |
Report material AI involvement where required. |
| Methodology |
Suggesting study procedures or analytical approaches. |
Ensure that human researchers evaluate and justify the adopted methods. |
These examples illustrate functional similarities. They do not imply that CRediT formally recognizes AI systems as authors or that every listed activity requires an identical disclosure.
Can ChatGPT Receive a Writing – Original Draft Credit?
ChatGPT can generate manuscript text, including material that resembles an original draft.
However, under major publisher policies prohibiting AI authorship, the system should not be entered as a named author merely because it generated the initial wording.
Instead, researchers should disclose the assistance through the mechanism required by the journal.
For example, a journal may request a dedicated AI-use statement describing how ChatGPT contributed to manuscript preparation.
The human authors must still meet the journal's authorship criteria and accept responsibility for the final work.
It would also be misleading to assign a human researcher a drafting contribution that substantially overstates the work they actually performed.
Accurate contribution reporting requires describing human activities truthfully while documenting AI assistance separately where appropriate.
Can AI Be Credited for Formal Analysis?
AI may assist with statistical analysis, data processing, or interpretation. Such assistance can be methodologically consequential.
For example, a researcher may use a language model to generate code for a regression analysis or to suggest analytical procedures.
However, the researcher must verify the code, evaluate the assumptions, and confirm that the reported findings are supported by the data.
Where the journal requires methodological transparency, the AI tool's role should be described in the methods section or another appropriate location.
A formal analysis contribution assigned to an eligible human author should reflect that person's actual analytical work, including substantive verification and interpretation.
Simply executing AI-generated code without understanding its operation may not substantiate the claimed level of analytical contribution.
Can AI Be Credited for Conceptualization or Methodology?
AI systems can suggest research questions, theoretical relationships, methodological alternatives, and analytical frameworks.
These outputs may influence a researcher's thinking, but the system's suggestions do not independently establish an accountable intellectual contribution under the prevailing authorship policies.
The researcher must evaluate whether the proposed approach is appropriate, original, ethically defensible, and consistent with the research objectives.
If an AI suggestion materially shapes the study, its use may warrant disclosure according to the applicable policy.
However, the human researcher should not claim conceptualization or methodology contributions that they cannot meaningfully explain or defend.
The distinction is between receiving a suggestion and exercising scholarly judgment in adopting, adapting, or rejecting it.
What Does NISO Say About CRediT and Authorship?
NISO's CRediT documentation distinguishes contribution description from authorship eligibility.
CRediT provides a structured way to describe roles but does not itself determine who qualifies as an author.
This distinction matters when interpreting AI involvement because the existence of a corresponding contribution category does not automatically establish eligibility for the author list.
Researchers should therefore follow the journal's implementation of CRediT and its separate authorship and AI-disclosure policies.
Where a submission system allows only named authors to receive CRediT roles, researchers should not attempt to add an AI tool as an author merely to document its assistance.
What Do ICMJE and COPE Say About AI Contributions?
The International Committee of Medical Journal Editors states that AI-assisted technologies should not be listed as authors because they cannot take responsibility for the accuracy, integrity, and originality of submitted work.
COPE similarly excludes AI systems from authorship and emphasizes the responsibility of human authors for AI-assisted material.
Both positions recognize that AI may be used during manuscript preparation or research, subject to appropriate transparency and applicable restrictions.
Neither establishes that AI-generated writing or analysis should be credited through a conventional author contribution entry as though the system were an eligible human author.
The distinction between acknowledging AI assistance and granting academic authorship therefore remains important even when CRediT is used.
Should Human Authors Claim CRediT Roles for AI-Assisted Tasks?
Human authors should report their actual contributions accurately.
Using AI assistance does not automatically prevent a researcher from qualifying for a contribution role. A researcher may meaningfully perform formal analysis while using AI-generated code as part of the workflow.
Likewise, a researcher may substantially draft or critically revise a manuscript with AI-assisted language suggestions.
However, operating an AI tool should not automatically be equated with making the full intellectual contribution associated with the resulting output.
The relevant question is what the human researcher actually contributed, including the decisions, verification, interpretation, and revisions undertaken.
Researchers should avoid both understating legitimate human work and overstating contributions that were largely performed by AI without meaningful human involvement.
Can an Author Contribution Statement Mention AI Without Assigning It a CRediT Role?
Possibly, where the journal expressly permits contextual information about AI assistance within or alongside its contribution statement.
However, this should not be assumed to be a universal practice.
Many journals use structured contribution statements intended to identify the roles of named authors. Adding AI systems directly to those statements may conflict with the publisher's reporting format.
A separate AI-use disclosure is often clearer because it allows the researcher to describe the tool and its role without presenting it as an eligible author.
If the journal provides a designated field for AI assistance, that mechanism should be used according to its instructions.
Researchers should not invent a new CRediT role or assign an AI system an author identifier merely to record its involvement.
What If AI Performed Most of the Tasks in a Contribution Category?
The extent of AI involvement may raise questions about whether a human contributor's claimed role accurately represents their participation.
For example, a researcher who asks AI to generate an entire analysis and incorporates the results without meaningful verification should not automatically claim to have performed a substantial formal analysis.
Similarly, a person who submits an AI-generated manuscript without critically reviewing its content may not have demonstrated the scholarly participation required for authorship.
These circumstances do not make AI eligible for the corresponding role as an author.
Instead, they require examination of the human contribution and the journal's authorship criteria.
The broader issue of authorship when AI produces most of a manuscript cannot be resolved simply by transferring contribution labels from the human researcher to the AI system.
Does Disclosing AI Assistance Change Human Accountability?
No. Reporting AI involvement does not transfer responsibility to the tool.
The human authors remain accountable for the accuracy and integrity of the submitted work under major publishing policies.
If AI generates incorrect analytical code, fabricated citations, or unsupported interpretations, the authors must address those problems.
Contribution statements can help identify which human authors were responsible for particular research activities, but they do not eliminate the broader accountability requirements established by the journal.
This is why extensive AI assistance does not reduce human authorship responsibility.
How Should AI Assistance Be Disclosed Alongside CRediT?
The most appropriate approach is to keep the human contribution statement accurate and report AI assistance through the journal's required mechanism.
For example, a journal may require a CRediT statement identifying the researchers responsible for formal analysis and manuscript preparation, together with a separate declaration describing AI-generated code or language assistance.
Where AI affected the research methodology, the methods section may also need to describe the procedure and verification undertaken.
The statements should be consistent. A researcher should not claim exclusive manual drafting while separately reporting that AI generated substantial portions of the text without explaining the nature of their own contribution.
Nor should the AI disclosure be used to obscure responsibility for the final manuscript.
