Can an AI System Meet Standard Academic Authorship Criteria?

If AI Can Perform Scholarly Tasks, Why Doesn't It Qualify as an Author?

An AI system helps formulate a research question, analyzes a dataset, interprets preliminary findings, and drafts sections of a manuscript. These activities resemble contributions that might qualify a human researcher for authorship.

Does that mean the AI system has satisfied academic authorship criteria?

The answer requires examining what authorship frameworks actually demand. Producing research content is only one part of the equation. Many journals also require authors to review the manuscript, approve its final version, and accept accountability for the published work.

Rather than treating AI authorship as a simple question of how much work the system performed, researchers need to distinguish between completing scholarly tasks and satisfying the requirements for recognition as an academic author.

AI Can Perform Some Authorship-Related Tasks but Cannot Satisfy All the Criteria

The crucial distinction is between a system's ability to produce a scholarly output and its eligibility to become a responsible member of the author group.

How AI Performs Against the Four ICMJE Authorship Criteria

What Are the Four ICMJE Authorship Criteria?

The International Committee of Medical Journal Editors provides one of the most widely recognized authorship frameworks in scholarly publishing.

Its recommendations identify four criteria that must all be met by an individual listed as an author.

1. Substantial scholarly contribution: Contributing substantially to the conception or design of the work, or to the acquisition, analysis, or interpretation of data.
2. Manuscript drafting or critical review: Drafting the work or reviewing it critically for important intellectual content.
3. Final approval: Approving the version that will be published.
4. Accountability: Agreeing to be accountable for all aspects of the work and ensuring that questions about accuracy or integrity are appropriately investigated and resolved.

These criteria are cumulative. Meeting one or two does not establish eligibility for authorship.

The ICMJE also emphasizes that authors should understand the responsibilities of their coauthors and have confidence in the integrity of their contributions.

Although developed for medical journals, the framework has influenced authorship practices more broadly. It should not, however, be presented as a universal requirement imposed on every academic discipline.

Criterion 1: Can AI Make a Substantial Scholarly Contribution?

The first ICMJE criterion concerns substantial contributions to research conception, design, data acquisition, analysis, or interpretation.

AI systems can assist with activities resembling several of these contributions.

For example, a language model may suggest research questions, compare methodological approaches, generate analytical code, identify patterns in supplied data, or propose interpretations.

Some of these outputs may materially influence the researcher's work.

However, producing an output that resembles a qualifying contribution does not automatically establish that the system has made a contribution in the authorship sense required by a journal.

The distinction involves intellectual responsibility, the actual research context, and the capacity to stand behind the contribution.

A researcher who uses an AI-generated analytical suggestion must still determine whether it is methodologically appropriate and supported by the evidence.

Under prevailing publisher policies, the tool's contribution may need to be disclosed, but it does not make the system an eligible author.

Criterion 2: Can AI Draft or Critically Review a Research Manuscript?

AI systems can plainly generate manuscript text. They can also review supplied writing, identify possible inconsistencies, suggest revisions, and produce critiques of arguments.

These activities resemble the tasks described in the second ICMJE criterion.

Nevertheless, the ability to draft or review text is not sufficient for authorship.

The ICMJE's second criterion is one component of a cumulative framework. A contributor must also meet the other requirements.

Furthermore, AI-generated criticism may be inaccurate, incomplete, or unsupported. A model can identify genuine weaknesses, but it can also invent problems or produce plausible yet misleading explanations.

Human researchers must therefore evaluate the suggestions before incorporating them into the manuscript.

It is more accurate to say that AI can perform drafting and review activities than to conclude that it thereby satisfies the full authorship framework.

Criterion 3: Can AI Give Final Approval to a Manuscript?

The third criterion requires an author to approve the version that will be published.

An AI system can examine a final manuscript and generate a statement such as "I approve this version for publication."

However, producing those words is not equivalent to final approval by an eligible academic author.

Final approval is an act through which the author accepts the submitted version as work for which they are professionally accountable.

A language model does not independently enter that relationship with the journal.

Its generated statement cannot substitute for the approval required from a human author.

This distinction matters because final approval is not simply another editing task. It is a formal commitment associated with the submitted work.

Criterion 4: Can AI Accept Accountability for the Entire Research Paper?

The fourth criterion requires authors to agree to be accountable for all aspects of the work, including ensuring that questions concerning accuracy or integrity are appropriately investigated and resolved.

AI systems cannot assume this responsibility under current scholarly publishing frameworks.

They cannot independently fulfill the professional and institutional obligations associated with authorship, participate as accountable researchers in ordinary misconduct procedures, or undertake the relevant publication agreements in their own capacity.

An AI system may generate an explanation for an error or suggest a correction, but that does not make it responsible for the submitted research.

For example, if a manuscript contains fabricated citations generated by AI, the human authors remain responsible for addressing the problem.

This limitation is central to why AI systems cannot accept academic authorship responsibility.

How Does AI Compare Across All Four Criteria?

The table distinguishes technical capability from authorship eligibility.

It does not imply that AI contributions are necessarily insignificant. Rather, the cumulative criteria mean that performing some relevant tasks cannot compensate for an inability to meet the remaining requirements.

Why Must All Four Criteria Be Satisfied?

The ICMJE framework is designed to connect scholarly credit with responsibility.

It seeks to prevent individuals from receiving authorship solely because they provided funding, general supervision, administrative assistance, or another contribution that does not satisfy the full criteria.

At the same time, the ICMJE states that individuals who meet the first criterion should have an opportunity to participate in manuscript drafting or critical review and final approval so that qualifying contributors are not unfairly excluded.

This is an important qualification. The criteria should not be used to deny legitimate human contributors authorship by preventing them from fulfilling the remaining requirements.

AI systems present a different issue because their exclusion under prevailing policies concerns their inability to assume authorship accountability, not merely a lack of opportunity to review the manuscript.

Does COPE Apply the Same Four Criteria?

COPE does not simply reproduce the ICMJE's four-criterion framework as a universal authorship test for every discipline.

Instead, its position on AI authorship emphasizes the responsibilities attached to publication.

COPE states that AI tools cannot be listed as authors because they cannot take responsibility for submitted work.

It also notes that AI systems are not legal entities capable of declaring conflicts of interest or managing copyright and licensing agreements.

The conclusion is consistent with the ICMJE position, even though the frameworks are expressed differently.

Researchers should therefore distinguish agreement on the exclusion of AI authorship from complete uniformity in authorship criteria across organizations.

Do All Academic Disciplines Use the ICMJE Criteria?

No. Authorship conventions vary across fields.

Some disciplines emphasize substantial intellectual contribution and responsibility without formally adopting all four ICMJE criteria.

Others use detailed institutional or publisher requirements concerning author approval, contributorship, and accountability.

Large collaborative projects may also apply group-authorship arrangements or specialized contribution reporting.

These differences matter when determining whether a human contributor qualifies for authorship.

However, major publishing policies across disciplines generally distinguish AI assistance from eligible human authorship.

Researchers should verify the requirements of their intended journal rather than assuming that one framework automatically governs every publication.

Can the CRediT Taxonomy Make AI Eligible for Authorship?

The Contributor Roles Taxonomy (CRediT) describes specific contributions to scholarly work, such as conceptualization, data curation, formal analysis, methodology, software, and writing.

It is designed to make contributions more transparent.

However, CRediT does not itself determine who qualifies for authorship. Authorship eligibility remains governed by the applicable journal or institutional criteria.

An AI system may assist with tasks resembling CRediT roles, such as generating code or drafting text.

That does not make it an eligible author or justify assigning it a position in the author byline.

Questions about how AI assistance relates to author contribution statements should therefore be distinguished from the eligibility requirements for authorship.

What If AI Performs More Work Than a Human Coauthor?

The quantity of work performed does not automatically determine authorship eligibility.

For example, an AI system might generate extensive manuscript text while a human coauthor contributes a narrower but qualifying intellectual role.

Under the ICMJE framework, the human contributor may qualify if all four criteria are met. The AI system remains ineligible because it cannot satisfy the accountability requirements.

However, this does not mean that any human who merely submits AI-generated material automatically qualifies as an author.

The human contributor must still meet the journal's applicable criteria.

When AI produces most of the manuscript, determining legitimate human authorship may require careful examination of the actual intellectual contributions and responsibilities undertaken.

Can Human Researchers Accept Accountability on Behalf of AI?

Human authors must accept responsibility for AI-assisted content included in their manuscripts.

However, this does not make the AI system an eligible coauthor.

Under the ICMJE framework, each named author must satisfy all four criteria. One author cannot fulfill another author's eligibility requirements by proxy.

Human responsibility for AI-generated material therefore supports the use of appropriate disclosure rather than AI authorship.

The distinction also explains why AI assistance does not reduce the responsibility of human authors.

How Should AI Contributions Be Reported Instead?

Where AI assistance is permitted, researchers should follow the disclosure requirements of the intended journal.

The ICMJE recommends describing AI writing assistance in the acknowledgment section and reporting AI use for data collection, analysis, or figure generation in the methods section, as applicable.

COPE similarly emphasizes transparency about the AI tool used and its role in preparing or conducting the research.

The reporting mechanism may vary across publishers, and some journals provide designated AI-use disclosure statements.

The important distinction is that reporting a tool's contribution does not confer authorship status.

Researchers should consider appropriate acknowledgment of AI assistance as a transparency obligation separate from the authorship criteria.

Applying the Four Authorship Criteria to an AI-Assisted Research Project

The example shows why authorship cannot be determined simply by counting tasks performed or paragraphs produced.

The AI system may have contributed materially to the research workflow, but it does not satisfy the complete authorship framework.

At the same time, the researcher's own eligibility should not be assumed merely because the AI is excluded. The researcher must independently meet the relevant criteria and remain capable of defending the work.

Misconceptions About AI and Authorship Eligibility

How Should You Apply Authorship Criteria When AI Contributes to Research?

Researchers should evaluate authorship eligibility and AI assistance as related but separate questions.

The first concerns which human contributors qualify for authorship. The second concerns how the use of AI should be documented and disclosed.

The practical consequence is that human authorship cannot be established merely by excluding AI from the byline.

Researchers must still demonstrate qualifying contributions and accept responsibility for the work.

The authorship decision should therefore be based on the applicable criteria and the actual work performed by the human contributors.

What to Verify Before Assigning Authorship in AI-Assisted Research

Frequently Asked Questions About AI and Academic Authorship Criteria

Performing Authorship-Related Tasks Is Not the Same as Qualifying as an Author

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.