Should AI-Assisted Contributions Affect Academic Authorship Decisions?

Does Using AI Change Who Deserves Authorship Credit?

Two researchers collaborate on a manuscript. One develops the research question, designs the methodology, and interprets the findings. The other uses generative AI to draft substantial portions of the introduction, produce analytical code, and revise the discussion.

Both have contributed to the project, but the nature of their contributions differs.

Should the second researcher receive authorship credit because they produced a substantial amount of manuscript material with AI? Should their contribution receive less recognition because some of the work was automated? And should the AI system itself be acknowledged as a contributor?

These questions concern AI-assisted contributions and academic authorship. As generative AI becomes involved in more research activities, authorship decisions must distinguish the intellectual work performed by human researchers from the outputs produced by the technologies they use.

The difficulty is not simply determining how much work AI completed. It is establishing which human contributions satisfy the applicable authorship criteria and who can accept responsibility for the resulting research.

Should AI Assistance Change Academic Authorship Decisions?

How AI Assistance Changes the Assessment of Human Research Contributions

Academic Authorship Recognizes Responsibility, Not Merely Output Production

Academic authorship serves several purposes. It identifies individuals who made qualifying contributions to scholarly work, provides appropriate recognition, and establishes responsibility for the published research.

Authorship is therefore different from simply producing a certain number of manuscript pages, lines of code, tables, or figures.

A researcher who develops a study's conceptual framework may make a substantial intellectual contribution even if another collaborator prepares most of the manuscript's initial text.

Conversely, producing a large volume of text does not necessarily establish authorship eligibility if the person made no qualifying intellectual contribution under the applicable criteria.

Generative AI makes this distinction especially important because the amount of material produced can become increasingly disconnected from the amount and nature of human intellectual work involved.

The appropriate question is not how much content a person generated with AI, but what the person contributed to the research and whether that contribution satisfies the relevant authorship requirements.

What Do Established Authorship Criteria Require?

Authorship standards vary across disciplines, journals, and scholarly organizations.

The International Committee of Medical Journal Editors (ICMJE) recommends that authorship be based on four criteria:

1. Substantial contributions to the conception or design of the work, or the acquisition, analysis, or interpretation of data.
2. Drafting the work or critically reviewing it 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.

Under the ICMJE framework, individuals designated as authors should satisfy all four criteria. Contributors who do not meet all four should be acknowledged appropriately rather than automatically listed as authors.

These criteria are influential, particularly in biomedical publishing, but they are not universal rules for every academic discipline.

Researchers should consult the authorship requirements of the target journal, relevant professional society, and participating institutions.

AI assistance does not remove the need to apply these criteria.

Why AI Systems Should Not Be Listed as Academic Authors

Generative AI can produce text, code, images, and analytical suggestions. However, generating material is not equivalent to assuming scholarly authorship responsibilities.

The ICMJE explicitly states that AI-assisted technologies should not be listed as authors because they cannot be responsible for the accuracy, integrity, and originality of the work in the manner required of human authors.

The Committee on Publication Ethics (COPE) similarly states that AI tools cannot meet the requirements for authorship because they cannot take responsibility for submitted work.

These positions distinguish a tool's contribution to producing research material from the responsibilities associated with authorship.

For example, an AI assistant may suggest a statistical model, but it cannot independently assume the obligations of a researcher who approves and reports that analysis.

Likewise, an AI-generated discussion paragraph does not make the system accountable for whether the interpretation accurately reflects the study findings.

Human researchers must evaluate and accept responsibility for material incorporated into their scholarly work.

Does Using AI Reduce the Value of a Human Contribution?

Not automatically.

Researchers have long used software and other technologies to support intellectual work. Statistical programs calculate estimates, reference managers organize citations, and visualization tools produce figures.

Using these tools does not ordinarily eliminate the human contribution involved in designing, interpreting, or evaluating the research.

Generative AI can similarly assist researchers in carrying out substantive activities.

For example, a researcher may develop an original analytical strategy, use AI to generate an initial implementation, test the code, evaluate assumptions, and interpret the findings.

The researcher's intellectual contribution may remain substantial even though AI generated some of the implementation.

However, the opposite situation is also possible.

A person who merely requests an analysis, copies the generated results, and cannot explain or evaluate the method may have contributed less substantive intellectual work than the volume of output suggests.

Authorship assessment should examine what the person actually did rather than automatically reward or penalize AI use.

Why Prompting Alone Does Not Automatically Establish Authorship

Writing prompts can involve intellectual effort. A researcher may formulate detailed instructions, define analytical constraints, evaluate competing outputs, and iteratively refine an approach.

Some of these activities may constitute meaningful research contributions.

However, entering a prompt does not automatically satisfy authorship criteria.

For example, asking AI to write an introduction about a topic does not necessarily establish that the person developed the research question, evaluated the literature, or contributed substantive scholarly interpretation.

Similarly, asking AI to analyze a dataset does not establish that the person designed the analysis or verified its validity.

The relevant contribution lies in the human intellectual activity associated with the work, not simply in operating the interface.

This distinction is especially important when assessing how AI can make human intellectual contributions harder to identify.

How Should AI-Assisted Activities Be Evaluated?

Different AI-assisted activities may involve different forms of human contribution.

The table illustrates how human contributions can be assessed. It does not establish universal thresholds for authorship.

A person's overall eligibility must be evaluated against the applicable criteria, including responsibilities beyond the initial contribution.

Can the CRediT Taxonomy Help Identify Contributions?

The Contributor Roles Taxonomy, commonly known as CRediT, provides a standardized vocabulary for describing contributions to scholarly work.

Its roles include Conceptualization, Methodology, Software, Formal analysis, Investigation, Data curation, Writing – original draft, Writing – review & editing, Supervision, and other recognized activities.

CRediT can help research teams describe what individual contributors actually did.

For example, one researcher may contribute to conceptualization and methodology, while another performs formal analysis and software development.

However, CRediT is not itself an authorship eligibility standard.

Assigning someone a contributor role does not automatically establish that the person satisfies the target journal's authorship criteria.

Research teams can use CRediT to improve transparency while applying authorship criteria separately.

AI systems should not be assigned human contributor roles as though they were accountable scholarly contributors.

Can AI Assistance Make Authorship Contributions Difficult to Distinguish?

Yes. When AI assists with several stages of research, the visible outputs may not reveal who performed the underlying intellectual work.

For example, a manuscript section may have been generated by AI following detailed methodological instructions from one researcher and critically revised by another.

Determining each person's contribution requires examining their actual activities rather than inferring responsibility from the final text.

Research teams may therefore benefit from documenting substantive contributions during the project rather than reconstructing them only at submission.

Such records can help distinguish conceptual decisions, technical implementation, verification, and intellectual revision.

They may also reduce disputes when AI-generated material is extensive.

What If One Researcher Uses AI and Another Does Not?

Different patterns of AI use do not automatically justify different authorship treatment.

One researcher may complete a substantive contribution with AI assistance, while another performs comparable work without AI.

The appropriate assessment concerns the intellectual contribution and applicable authorship responsibilities.

However, collaborators should know when AI materially affects shared research outputs, especially when its use creates verification, confidentiality, or disclosure obligations.

The question of whether a coauthor can use AI without informing the other authors is therefore relevant to collaboration governance, even when the AI use does not change authorship eligibility.

Should AI-Generated Text Count Toward Writing Contributions?

Word count is not a reliable measure of scholarly contribution.

A researcher may write relatively little text while making a substantial contribution to conceptualization, methodology, or interpretation.

Another may produce many pages through AI without critically evaluating their content.

Under authorship frameworks that consider drafting or critical intellectual revision, the relevant issue is the person's substantive involvement in developing or reviewing the manuscript.

For example, a researcher who critically restructures an AI-assisted discussion, corrects unsupported interpretations, and develops the final scholarly argument may make a meaningful intellectual contribution.

By contrast, merely transferring generated text into a manuscript without substantive evaluation should not automatically be treated as equivalent to intellectual authorship.

The complete authorship criteria must still be satisfied.

Can AI-Assisted Coding Constitute a Qualifying Research Contribution?

Potentially.

Software development and computational analysis can be substantial research contributions, depending on the project and applicable authorship criteria.

A researcher who designs an algorithm, establishes analytical requirements, validates generated code, and interprets computational results may contribute significant intellectual work.

However, generating code with AI does not by itself demonstrate that the researcher made those decisions.

Research teams should distinguish between operating a tool and taking intellectual responsibility for the computational method.

The researcher should also ensure that the generated implementation is appropriately tested and documented.

Authorship credit should not be used as a substitute for verifying computational accuracy.

Why Accountability Cannot Be Delegated to AI

Academic authors must be able to answer questions about their contributions and participate in resolving concerns about the integrity of the work.

An AI system cannot independently fulfill these scholarly responsibilities.

For example, if an AI-generated citation is fabricated, human authors must investigate and correct the problem.

If AI-assisted statistical code produces incorrect results, the relevant researchers must examine the implementation and determine the implications for the study.

Similarly, AI-generated interpretations must be checked against the underlying evidence.

These obligations connect to responsibility for verifying AI-assisted work produced by a coauthor.

Individual roles may differ, but authorship entails responsibilities that cannot be transferred to an automated system.

How Should Research Teams Handle Authorship Disagreements Involving AI?

Authorship disputes may arise when collaborators disagree about whether AI-assisted work constitutes a substantial intellectual contribution.

For example, one researcher may argue that generating and editing most of the manuscript establishes authorship eligibility. Another may question whether the person contributed to the research design, analysis, or substantive interpretation.

The team should examine the actual contributions and apply the agreed authorship criteria.

Relevant documentation may include research plans, analytical decisions, manuscript revisions, contribution records, and evidence of substantive intellectual review.

Disagreements should not be resolved solely by comparing word counts, prompt counts, or the number of AI-generated outputs.

Where necessary, researchers should follow institutional authorship-dispute procedures or seek appropriate guidance from responsible research offices.

COPE provides guidance concerning authorship and contributorship issues, although the resolution of a particular dispute depends on the applicable institutional and publishing arrangements.

Should AI Use Be Disclosed Separately From Human Contributions?

Yes, when required by the relevant policy, and the two forms of information should not be confused.

An authorship or contributor statement identifies the roles and responsibilities of human contributors.

An AI-use disclosure describes how AI-assisted technologies contributed to the research or manuscript, according to the applicable requirements.

For example, a contribution statement may identify a researcher as responsible for formal analysis, while an AI disclosure explains that the researcher used an AI assistant to generate preliminary analytical code that was subsequently verified.

These statements answer different questions.

Researchers should consult the current policies of their target journal, publisher, institution, and funder because disclosure requirements vary.

The need for transparency is also relevant to ensuring that coauthors understand which parts of the project involved AI.

Can AI Assistance Change Authorship Eligibility During a Project?

Potentially, but not merely because a researcher begins or stops using AI.

Authorship eligibility may change when a person's actual intellectual contribution changes.

For example, a collaborator initially responsible only for routine editing may later contribute substantially to interpreting findings and critically revising the manuscript.

That additional work may affect authorship eligibility under the applicable criteria.

Conversely, a person originally expected to conduct substantive analysis may ultimately perform only routine administrative tasks while other researchers complete the intellectual work.

The final authorship decision should reflect the actual contributions and responsibilities rather than the original assignment alone.

Research teams should revisit authorship expectations when project responsibilities change and before submission.

Determining Authorship When AI Produces Much of the Manuscript Text

Under a framework such as ICMJE, the three researchers could qualify for authorship if each satisfies all applicable criteria. The scenario alone does not establish that they have done so.

Researcher C should not automatically receive greater authorship credit merely because the AI-assisted drafting process produced most of the manuscript's words.

Equally, C's contribution should not automatically be discounted because AI assisted with the initial text.

The appropriate decision depends on the substantive intellectual work performed and the responsibilities accepted by each contributor.

Misconceptions About AI-Assisted Authorship

How to Make Fair Authorship Decisions When AI Is Used

Research teams should evaluate authorship through documented human contributions rather than the quantity of AI-generated outputs.

The relevant criteria should be agreed upon early and revisited as the project develops.

Agree on Authorship Criteria Early

Collaborators should identify the authorship framework they intend to apply before disputes arise.

For projects targeting a particular journal, the journal's requirements should be consulted.

Where authorship criteria differ among collaborating disciplines, the team should discuss how those differences will be managed.

These expectations may be incorporated into research collaboration agreements that address AI use or related project governance documents.

Document Human Contributions Throughout the Project

Researchers should record substantive activities such as conceptualization, methodology, data analysis, interpretation, software development, and critical manuscript revision.

CRediT can provide a useful vocabulary for describing these contributions.

However, the team should apply authorship eligibility criteria separately.

Documentation should reflect actual work rather than planned responsibilities that were never completed.

Distinguish AI Assistance From Human Intellectual Responsibility

When AI contributes to an output, identify what the human researcher decided, verified, or interpreted.

For example, an analyst may use AI-generated code but remain responsible for the analytical specification, testing, and interpretation.

A manuscript author may use AI to improve language while retaining responsibility for the accuracy of claims and the coherence of the scholarly argument.

These distinctions help prevent both exaggerated and unfairly diminished recognition of AI-assisted human contributions.

Review Authorship Before Submission

Research teams should confirm that proposed authors satisfy the applicable criteria and have reviewed the final manuscript.

They should also ensure that contributor statements accurately represent the work performed.

Where AI assistance requires disclosure, the statement should be prepared separately and accurately.

Researchers should avoid treating the disclosure as evidence that generated material has been verified.

Resolve Disagreements Through Established Procedures

When collaborators cannot agree on authorship, the issue should be examined through the relevant institutional or publishing procedures.

Researchers should not assume that AI use automatically determines eligibility or contribution ranking.

Decisions should be based on documented human intellectual work, the applicable criteria, and the responsibilities associated with authorship.

Before Assigning Authorship in AI-Assisted Research

Common Questions About AI-Assisted Academic Authorship

Recognize Human Intellectual Work, Not the Volume of AI Output

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.