03 · What You Need to Know
Why Coauthors Need Visibility Into Material AI Contributions
Coauthor Awareness Is Part of Informed Manuscript Approval
Academic authorship involves more than appearing on a manuscript's title page.
Authors are expected to make scholarly contributions and accept responsibilities according to the criteria governing their publication.
The International Committee of Medical Journal Editors (ICMJE), for example, expects authors to approve the final version and agree to be accountable for the work under its authorship criteria.
These recommendations apply to journals adopting them and should not be represented as universal requirements across every discipline.
Nevertheless, they illustrate a general problem in collaborative scholarship: authors cannot meaningfully evaluate research decisions about which they lack essential information.
When AI materially contributes to a method or interpretation, the relevant information may affect whether coauthors can responsibly approve the resulting work.
What Does Coauthor Awareness Actually Mean?
Coauthor awareness does not necessarily mean that every collaborator must inspect every AI conversation.
It means that the team has sufficient information about consequential AI assistance to understand its role and evaluate the resulting research.
For example, coauthors may need to know that AI generated the initial statistical code, that the code was subsequently tested, and that a qualified researcher reviewed the final implementation.
They may not need to read unrelated conversations in which the analyst asked AI to explain general programming terminology.
The distinction concerns the relevance of the information to collective scholarly responsibility.
Why Materiality Matters More Than Frequency
A researcher may use AI repeatedly to improve nonconfidential correspondence without materially influencing the study.
Another may consult AI once and adopt a recommendation that changes the primary statistical model.
The second interaction may have greater scientific consequences despite involving less AI use.
Materiality therefore concerns the significance of an AI contribution to research decisions, evidence, or reporting.
It should not be defined exclusively through generated word counts, prompt frequency, or the proportion of text that appears to have been written by AI.
A substantive AI contribution may remain influential even after its original wording has been rewritten.
Which AI Contributions Should Be Visible to All Coauthors?
Research teams should distinguish information needed by the entire author group from technical details that may initially be reviewed by designated specialists.
| Research activity |
Information coauthors should generally know |
Principal accountability concern |
| Research design |
Whether AI materially influenced the selected design |
Methodological justification |
| Measurement development |
Whether AI generated or substantially modified research instruments |
Validity of measurement |
| Literature synthesis |
Whether AI materially contributed claims or interpretations |
Accuracy of cited evidence |
| Data processing |
Whether consequential transformations involved AI-generated procedures |
Data integrity and reproducibility |
| Statistical analysis |
Whether AI influenced model selection or analytical implementation |
Validity of results |
| Qualitative interpretation |
Whether AI materially contributed codes, themes, or interpretations |
Methodological coherence |
| Substantive writing |
Whether AI influenced major arguments or conclusions |
Intellectual contribution and evidential support |
| Language editing |
Information required by team or publisher rules |
Accurate reporting and preservation of meaning |
The appropriate disclosure may be concise. The purpose is to make consequential AI involvement visible, not to create an exhaustive technological history.
Does Every Coauthor Need to Understand Every AI Tool?
Collaborative research often depends on specialized expertise.
A qualitative researcher may not understand every line of statistical code, just as a statistician may not be familiar with every interpretive decision in qualitative analysis.
AI does not eliminate this division of labor.
However, collaborators should know which consequential procedures were used and whether appropriately qualified researchers verified them.
They should also be able to identify who can answer questions about particular parts of the work.
Coauthor awareness therefore requires sufficient understanding of the research and its verification arrangements, not identical technical expertise among all authors.
Why AI-Generated Code Should Not Remain an Invisible Contribution
AI-generated analytical code can affect research findings without producing any text that appears in the manuscript.
For example, a script may incorrectly merge datasets, calculate derived variables, or remove observations.
The resulting tables may appear plausible even when the underlying procedure is incorrect.
Researchers should inspect and test consequential code rather than assume successful execution establishes correctness.
Wilson and colleagues (2017) describe good practices for scientific computing, including testing and documentation.
When AI materially contributes to analytical implementation, coauthors should know how that contribution was verified and who accepted responsibility for the final procedure.
What About AI-Assisted Literature Reviews?
AI may suggest references, summarize publications, or generate preliminary comparisons between studies.
These activities can influence the evidence presented in a manuscript.
However, AI-generated references may be fabricated, and genuine publications may be misrepresented.
Coauthors should therefore know when material literature claims originated from AI-assisted synthesis and whether the original sources were examined.
Checking that a DOI resolves does not establish that the article supports the claim attributed to it.
When such material forms part of the manuscript's central argument, relevant authors should be able to examine the evidential basis.
Should AI-Assisted Qualitative Interpretation Be Disclosed Internally?
Material AI involvement in qualitative interpretation generally warrants communication because it may influence how findings were developed.
For example, a researcher might use AI to suggest themes from interview transcripts or propose alternative explanations for participant experiences.
Whether that assistance is methodologically appropriate depends on the selected analytical approach.
Braun and Clarke (2022) emphasize the active interpretive role of researchers in reflexive thematic analysis.
Other qualitative approaches may support different forms of computational assistance.
Coauthors should understand how automated suggestions were evaluated and how the final interpretation was constructed.
Does Human Revision Eliminate the Need for Coauthor Awareness?
Not necessarily.
A researcher may substantially rewrite AI-generated text while retaining its underlying argument.
Similarly, an AI-generated methodological suggestion may influence the final analytical procedure even after the code has been modified.
The question is whether AI materially shaped the research rather than whether its original wording remains identifiable.
However, not every discarded suggestion requires disclosure.
Researchers should evaluate the actual influence of AI and the reporting requirements that apply.
Why Internal Disclosure and Journal Disclosure Are Different
Internal disclosure helps collaborators understand and verify the research before approving it.
Journal disclosure communicates relevant AI assistance according to the publication venue's requirements.
The two processes may involve different levels of detail.
For example, coauthors may need technical information about an AI-generated analytical script that is not appropriate for inclusion in a brief publication disclosure statement.
Conversely, a journal may require reporting of a particular form of language assistance that the team ordinarily handles through standing permission.
Researchers should distinguish undisclosed AI use within a collaboration from compliance with external publication policies.
What If the Team Has No AI-Use Agreement?
The absence of a formal agreement does not automatically mean that every AI interaction was unauthorized.
However, it may create uncertainty about which activities required consultation and how consequential outputs should be verified.
Researchers should establish what AI contributed, identify the requirements that actually applied, and evaluate whether the work remains scientifically defensible.
The team can then establish clearer expectations for future activities.
Written research team AI-use agreements may reduce ambiguity by distinguishing routine permissions from consequential activities requiring collective review.
How Much Detail Should an Internal AI Disclosure Include?
A useful internal disclosure should identify the task, nature of the AI contribution, significance for the research, and verification undertaken.
For example, a collaborator might report that AI generated preliminary code for calculating composite scores, which was tested against independently calculated examples before being used.
Another might explain that AI suggested alternative interpretations of a qualitative finding, but the final interpretation was developed through examination of the original transcripts.
The disclosure should provide enough information to support informed review.
It need not reproduce every unrelated interaction unless a relevant protocol or institutional requirement demands that level of documentation.
Who Should Coordinate Coauthor Awareness?
The corresponding author or principal investigator may coordinate the collection of information about AI contributions.
However, coordination should not be confused with exclusive responsibility for verification.
Each collaborator remains responsible for accurately reporting their own activities and fulfilling the authorship obligations applicable to them.
The team should establish who verifies AI-assisted contributions according to relevant expertise and research responsibilities.
Collective approval should be informed by those verification arrangements.
What If a Coauthor Learns About Material AI Use After Approving the Manuscript?
The team should determine what assistance occurred and whether it affects the validity or reporting of the research.
If AI-generated references or analytical procedures were not adequately checked, the relevant work should be verified.
Where inaccuracies are identified, appropriate corrections may be necessary.
If the manuscript has already been submitted or published, journal procedures may become relevant.
The late disclosure should also be evaluated against the requirements that applied at the time.
It should not automatically be characterized as misconduct without examining the evidence and governing rules.
Can a Coauthor Approve Work They Cannot Fully Verify Personally?
Collaborative research frequently requires authors to rely on appropriately qualified colleagues for specialized work.
That reliance should be informed rather than unquestioning.
A coauthor should understand the contribution's role, the relevant verification procedures, and who is responsible for answering technical questions.
Where serious uncertainties remain unresolved, approval may need to be postponed.
The objective is not to require every author to repeat every analysis but to preserve accountable scholarly collaboration.
Why Coauthor Awareness Should Not Become Surveillance
Research teams may be tempted to require complete AI interaction histories from every collaborator.
Such records can contain unrelated personal information, confidential research material, and preliminary ideas.
For many projects, a structured record of material contributions and verification may be more useful.
More extensive logging may be justified when AI interactions are themselves research data or when authorized procedures require them.
The appropriate level of documentation should reflect the scientific purpose and applicable information-governance requirements.
What Should Happen Before Final Manuscript Approval?
Before submission, the team should confirm whether material AI contributions have been communicated and adequately verified.
The corresponding author should establish which journal disclosure requirements apply.
Relevant collaborators should have an opportunity to examine consequential methods, evidence, and interpretations.
Any unresolved concerns should be addressed before the final manuscript is approved.
The aim is collective author accountability in AI-assisted research, supported by sufficient information rather than assumptions about individual tool use.
