If AI Writes Most of a Research Manuscript, Who Is the Author?

If AI Produces Most of the Words, Who Deserves Authorship Credit?

A researcher conducts a study, collects the data, and interprets the findings. Instead of writing the manuscript from the beginning, the researcher asks a generative AI system to produce the introduction, literature review, methodology, results narrative, and discussion. After reviewing and editing the output, the researcher submits the paper under their own name.

Who is the author of that manuscript? Is it the researcher who conducted the study, the AI system that produced most of the text, or both?

The question becomes more difficult when the researcher contributes little beyond prompting the AI and approving the generated output. In such cases, authorship cannot be resolved simply by identifying who typed the words or whose name appears on the submission.

Academic authorship involves intellectual contribution and responsibility for the published work. The extent of AI-generated writing may raise important questions about both.

Human Researchers May Be Authors, but AI Cannot Qualify as a Coauthor

The essential distinction is between producing the manuscript's wording and making the scholarly contributions that justify authorship. These activities can overlap, but they are not equivalent.

How Authorship Is Determined When AI Generates Most of a Manuscript

Is the Person Who Writes Most of the Words Necessarily the Author?

No. Academic authorship is not determined solely by the number of words a contributor produces.

A researcher may qualify for authorship through substantial contributions to study design, data analysis, or interpretation, even if another eligible human contributor performs much of the initial manuscript drafting.

Conversely, producing large amounts of text does not automatically establish authorship eligibility. Professional language editors, for example, may contribute substantially to manuscript wording without necessarily meeting the journal's complete authorship criteria.

Generative AI makes this distinction more visible because it can produce extensive, polished prose without independently assuming responsibility for the scholarship.

The relevant question is not who produced the largest proportion of sentences. It is whether the proposed human authors satisfy the requirements governing scholarly contribution, manuscript approval, and accountability.

Why Can't the AI System Be Named as the Author?

Major scholarly publishing authorities do not recognize AI systems as eligible authors.

The International Committee of Medical Journal Editors (ICMJE) 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.

The Committee on Publication Ethics (COPE) similarly excludes AI tools from authorship because they cannot accept responsibility for submitted work or fulfill relevant obligations concerning conflicts of interest and publication agreements.

Springer Nature also states that large language models do not satisfy its authorship criteria because they cannot take accountability for the work.

These policies do not deny that AI can generate useful academic content. They distinguish producing content from becoming an accountable scholarly author.

Consequently, listing an AI system as an academic author is not an acceptable solution under these publishing frameworks, even when it generated most of the manuscript.

What Authorship Criteria Must Human Researchers Satisfy?

The ICMJE framework provides a useful basis for examining authorship eligibility, although its four criteria are not universally imposed across every discipline.

Under this framework, each author must satisfy all four requirements:

1. Make substantial contributions to the conception or design of the work, or to the acquisition, analysis, or interpretation of data.
2. Draft the manuscript or critically review it for important intellectual content.
3. Approve the final version for publication.
4. Agree to be accountable for all aspects of the work, including ensuring that questions about accuracy or integrity are appropriately investigated and resolved.

A researcher who designed the study, collected the data, interpreted the findings, and critically revised the AI-generated manuscript may satisfy these criteria, provided the remaining requirements are met.

However, someone who merely provides prompts, accepts generated content without meaningful review, and submits the manuscript may not have demonstrated the contributions or accountability required for authorship.

Whether a particular individual qualifies depends on the journal's actual criteria and the work performed.

The detailed question of whether AI systems can meet academic authorship criteria is distinct from assessing the eligibility of the humans who used them.

Does Conducting the Research Matter More Than Drafting the Manuscript?

Conducting the underlying research can provide a substantial basis for authorship, but it does not automatically eliminate the remaining requirements.

Consider a researcher who independently designs a study, recruits participants, analyzes the data, and interprets the findings. If AI subsequently helps draft the manuscript, the researcher's intellectual contributions do not disappear simply because the wording was generated with assistance.

Nevertheless, under the ICMJE framework, the researcher must also participate in drafting or critical intellectual review, approve the final version, and accept accountability.

Writing assistance cannot substitute for these responsibilities.

The reverse situation also matters. A person who uses AI to generate a manuscript based on research conducted entirely by others cannot automatically claim authorship simply because they assembled or submitted the text.

Authorship must reflect genuine qualifying contributions, not merely control over the final document.

What Counts as Meaningful Human Review of AI-Generated Writing?

Meaningful review requires more than reading the generated text and deciding that it sounds academically appropriate.

Depending on the content, the researcher may need to verify the existence and relevance of cited sources, compare literature summaries with original publications, check reported results against analysis records, and determine whether interpretations follow from the evidence.

For example, an AI-generated discussion may claim that a study establishes a causal relationship even though the research design supports only an association. A responsible researcher must identify and correct that overstatement.

Similarly, a generated methods section may describe procedures that were never actually performed.

The researcher's responsibility is to ensure that the manuscript accurately represents the research conducted, not simply to make the prose readable.

However, meaningful review alone does not automatically establish authorship if the individual lacks other qualifying contributions.

Does Editing AI-Generated Text Make Someone Its Academic Author?

Editing may constitute a meaningful intellectual contribution when it involves substantive analysis, interpretation, or critical development of the work.

Yet editing is not a single activity. Correcting punctuation differs from reconstructing an argument, verifying evidence, or revising the interpretation of findings.

The distinction between editing and qualifying authorship is especially important when AI produces most of the manuscript's wording.

What If AI Generated the Introduction and Literature Review but Not the Research?

The human researchers may still qualify as authors if they meet the applicable criteria.

However, the introduction and literature review are not merely decorative sections. They establish the research problem, interpret existing scholarship, and justify the study's contribution.

If AI generates these sections, researchers must ensure that the literature is accurately represented, that cited sources are genuine, and that the argument reflects their own defensible scholarly judgment.

They should not assume that a fluent literature synthesis demonstrates an accurate understanding of the cited studies.

Any disclosure requirements for substantive AI drafting must also be followed.

The relevant issue is not simply which sections AI wrote, but whether the resulting manuscript accurately communicates research for which the human authors can accept responsibility.

What If AI Also Designed the Study or Analyzed the Data?

AI assistance may extend beyond manuscript writing into research design, coding, statistical analysis, qualitative interpretation, or other methodological activities.

Such involvement can raise additional concerns because errors may affect the research findings themselves.

For example, AI-generated statistical code may contain incorrect variable definitions or inappropriate analytical procedures. An AI-generated interpretation may overlook assumptions or exaggerate the meaning of a result.

Researchers must verify that the methods and analyses are appropriate and accurately reported.

Extensive methodological AI assistance may also require disclosure in the methods section or another designated location, depending on the journal.

The AI system still does not become an author. Human researchers remain responsible for ensuring that the study's methods and findings are defensible.

Can a Researcher Claim Sole Authorship of a Mostly AI-Written Paper?

Potentially, but only if the researcher meets the applicable authorship criteria and the journal permits the relevant AI assistance.

Sole authorship does not necessarily mean that the author manually typed every sentence without technological assistance.

Researchers routinely use software, analytical tools, and editorial support while remaining the authors of their work.

However, generative AI may contribute substantive content rather than merely assisting with mechanical tasks. This makes transparency and verification particularly important.

A researcher cannot justify sole authorship merely by pointing out that AI is ineligible for the author byline.

The researcher must still demonstrate qualifying intellectual contribution, appropriate review, final approval, and accountability under the relevant framework.

Questions about claiming authorship of AI-generated material that was not meaningfully reviewed are particularly consequential when the human contribution is minimal.

Does Extensive AI Writing Automatically Constitute Academic Misconduct?

No. The answer depends on the applicable institutional or publisher policy and the nature of the assistance.

Some journals permit generative AI assistance subject to disclosure and human verification. Others impose restrictions on particular uses or require additional documentation.

Extensive AI-generated drafting may violate a policy in one setting while being permitted under specified conditions in another.

Even where the assistance is permitted, researchers remain responsible for the accuracy and integrity of the submitted manuscript.

It is therefore necessary to distinguish three questions: whether AI was used, whether the human contributors qualify for authorship, and whether the assistance complied with the relevant policy.

These questions are related but cannot be answered by a single detection score or by counting AI-generated words.

Who Owns the Copyright in a Mostly AI-Generated Manuscript?

Copyright ownership is a separate legal question from academic authorship.

Copyright rules differ across jurisdictions, and the legal treatment of AI-generated material may depend on the extent and nature of human creative contribution.

For example, the US Copyright Office has explained that copyright protection generally requires human authorship and that purely AI-generated material is not protected merely because a person supplied prompts.

Human-authored elements of an AI-assisted work may qualify for protection where the applicable legal requirements are satisfied.

These principles should not be generalized automatically to every country.

Importantly, copyright ownership does not determine whether someone meets a journal's academic authorship criteria. A person might possess certain rights in a manuscript without satisfying the requirements for scholarly authorship, or qualify as an academic author while copyright is governed by institutional or publishing agreements.

How Should Extensive AI Writing Be Disclosed?

Researchers should consult the journal's current instructions and accurately describe the assistance used.

Depending on the policy, relevant information may include the AI tool, the purpose of its use, the sections or activities affected, and the human verification undertaken.

The International Committee of Medical Journal Editors recommends reporting AI writing assistance in acknowledgments and describing AI use for data collection, analysis, or figure generation in the methods section, as applicable.

Other publishers may require a designated AI-use statement.

Disclosure should describe what actually happened. It should not exaggerate the tool's contribution or minimize substantive assistance merely because the researcher later edited the generated material.

Determining where AI assistance belongs in a research manuscript requires attention to the journal's reporting requirements.

What Happens if No Human Contributor Meets the Authorship Criteria?

The exclusion of AI from authorship does not mean that an otherwise ineligible human automatically becomes the author.

If a manuscript is generated largely by AI and no human contributor has made the contributions or accepted the responsibilities required by the journal, the authorship problem remains unresolved.

The appropriate response is not to invent an author or assign authorship to the AI system.

Instead, the work may need substantial human scholarly development before it is suitable for submission.

Where the circumstances are uncertain, researchers should seek guidance from the intended journal or an appropriate research-integrity office.

Authorship should describe legitimate scholarly participation, not merely provide a name required by a submission form.

Two Researchers Use AI to Write Most of Their Manuscripts, but Their Contributions Differ

The example demonstrates why the percentage of AI-generated wording is not an adequate authorship test.

Researcher A may satisfy the relevant authorship criteria despite extensive AI-assisted drafting, provided the journal permits that assistance and the remaining requirements are met.

Researcher B cannot establish legitimate authorship merely by submitting the generated manuscript.

Neither scenario makes the AI system eligible for the author byline.

Misconceptions About Authorship of Mostly AI-Written Manuscripts

How Should You Determine Authorship When AI Generated Much of the Manuscript?

Begin by identifying the scholarly work performed by the human contributors independently of the amount of text generated by AI.

Then examine whether those contributions satisfy the journal's authorship requirements and whether the AI assistance complies with its policies.

Researchers should also distinguish authorship eligibility from responsibility for AI-assisted material. A human author who qualifies under the journal's criteria remains responsible for the manuscript, regardless of how much wording was generated by AI.

That responsibility includes verifying sources, checking methods, evaluating interpretations, and correcting errors.

The appropriate standard is not whether the manuscript sounds human-written. It is whether the human authors can substantiate the research and legitimately claim responsibility for the submitted work.

What to Verify Before Claiming Authorship of a Mostly AI-Written Manuscript

Frequently Asked Questions About Authorship of AI-Written Research Papers

Authorship Depends on Human Scholarly Contribution, Not the Number of AI-Generated Words

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.