Author resources · Research integrity
A multidisciplinary guide to the scholarship behind human accountability, transparent AI contributions, and responsible academic writing.
Which studies should researchers start with?
For ethical models of AI-assisted authorship, useful starting points are Hosseini et al. (2024; first online 2023) on accountability, van Woudenberg et al. (2024) on the meaning of authorship, Suchikova and Tsybuliak (2025) on AI contribution reporting, and Resnik and Hosseini (2025) on research ethics. Smith et al. (2026) extend the discussion to university governance.
The practical starting point: check the target publisher’s policy, describe the tool’s actual role, verify the resulting work, and retain human responsibility.
For researchers and early career scholars, the challenge is no longer simply whether an AI tool can help with a manuscript. It is how that assistance affects intellectual contribution, evidence, transparency, and responsibility. Language polishing, generating an argument, and analyzing interview data raise different reporting questions.
This guide organizes the literature around four themes: accountability and disclosure; authorship and epistemic agency; contribution transparency; and linguistic equity and institutional governance. These are organizing themes for this article, rather than four universally adopted or formally validated models. The sources include philosophical arguments, empirical research, commentaries, and editorial policies.
Core studies and ethical perspectives compared
The table separates what each publication contributes from the practical lesson a researcher can draw from it. Final issue years are used consistently; several papers first appeared online in an earlier year.
| Publication | Perspective | Scope / type | Practical relevance |
|---|---|---|---|
| Hosseini et al. (2024) | Accountability and transparency | Scholarly publishing; editorial | Disclose consequential AI assistance and check the content it produces. |
| Cotton et al. (2024) | Academic integrity | Higher education; discussion | Consider educational benefits alongside honesty, assessment, and institutional policy. |
| Donato et al. (2023) | Editorial transparency | Medical publishing; editorial | Understand early medical-journal responses to LLM-assisted writing. |
| van Woudenberg et al. (2024) | Conservative view of authorship | Philosophy; conceptual argument | Distinguish generating sentences from asserting claims and accepting responsibility. |
| Resnik & Hosseini (2025) | Responsible research practice | Research ethics; normative analysis | Address error, bias, disclosure, synthetic data, and human oversight. |
| Suchikova & Tsybuliak (2025) | AI contribution taxonomy | Research publishing; commentary | Report specific AI tasks rather than a vague statement that AI was used. |
| Lin (2025) | Role-specific human credit | Team science; proposed approach | Make human contributions and credit decisions more explicit. |
| Hamm et al. (2024) | Journal governance | Radiology; editors’ policy statement | Apply the relevant journal’s authorship, disclosure, and confidentiality rules. |
| Werdiningsih et al. (2024) | Language support and authenticity | EFL writing; small qualitative study | Consider how students balance writing assistance with their own voice. |
| Smith et al. (2026) | Institutional responsibility | University research governance | Support researchers with clear principles, training, infrastructure, and processes. |
1. Human accountability and transparent disclosure
Hosseini, Rasmussen, and Resnik’s editorial, first published online in 2023, argues that researchers remain accountable for text and ideas developed with AI. Its proposed policy emphasizes checking factual and citation accuracy, reasoning, bias, relevance, and originality. The paper offers an influential practical starting point, but its draft journal policy should not be treated as a universal rule (Hosseini et al., 2024).
Resnik and Hosseini (2025) examine the wider ethics of AI in research. Their recommendations include identifying error and bias, explaining AI use accessibly, consulting affected stakeholders, and clearly identifying synthetic data. This broadens the question from who wrote the manuscript to how the research itself was conducted.
Biomedical editorial responses make the policy implications concrete. Donato et al. (2023) discuss transparency in medical publishing, while the ESR editors’ statement specifies human authorship, AI disclosure, and confidentiality expectations for its journals (Hamm et al., 2024).
Current policy anchor: ICMJE guidance excludes AI tools from authorship and asks authors to explain their use in the cover letter and the appropriate manuscript section. COPE’s position likewise places responsibility on authors and calls for disclosure of the tool and its use. Follow the policy of the journal or publisher receiving your work.
2. Authorship, intention, and epistemic responsibility
A philosophical approach asks what it means to author a text. In their conservative view, van Woudenberg, Ranalli, and Bracker (2024) argue that ChatGPT does not qualify as an author because it lacks the relevant beliefs, intentions, capacity to perform speech acts, and responsibility for what it produces.
This is a philosophical position, not an experimental finding about every possible AI system. Its practical value is the distinction between producing fluent sentences and standing behind an argument. For a scholar, accepting a passage involves understanding its claims, examining its evidence, and being prepared to defend or correct it.
From an educational perspective, Cotton et al. (2024) discuss ChatGPT’s opportunities and academic-integrity risks, alongside the need for institutional policies and support. Their discussion provides context for responsible learning and writing; it is not a standalone authorship standard for every discipline.
A useful self-check: can you explain why the argument is sound, identify the evidence supporting it, and recognize what would make it wrong? If not, the passage needs further scholarly work before submission.
3. Contribution transparency: beyond “AI was used”
A binary disclosure conceals important differences. Using AI to revise a paragraph is different from asking it to classify interview excerpts or produce analytical code. Suchikova and Tsybuliak (2025) propose a standardized AI contribution taxonomy to make such roles more visible. Their commentary is a proposal for better reporting, not evidence that one taxonomy has been adopted throughout scholarly publishing.
Related work on human credit also matters. Lin (2025) proposes project-specific contribution thresholds and more detailed role reporting for team science. Aggarwal (2025) offers a quantitative approach based on contributors’ time and its assessed value. Both contribute to debates about attribution; neither establishes a universal rule for assigning AI authorship.
The CRediT taxonomy describes 14 contributor roles, such as conceptualization, software, validation, and writing. Its official guidance explicitly distinguishes contribution description from determining who qualifies as an author. Keep human contribution statements and AI-use disclosures clear and complementary.
A practical contribution record
For each substantive use, record the tool and model or version where available, the date, the research task, the material provided, the output retained, and the person responsible for checking it. Explain the checking procedure and any relevant limitations. Retain prompts and outputs when useful and permitted, while protecting participant data and confidential material.
How to disclose AI assistance by task
The following three levels are a practical planning aid for this guide. They do not override publisher instructions or establish that a particular AI use is permitted.
Level 1 · Basic mechanics
Spelling, references, and routine corrections
Ordinary spellchecking and reference management can be treated differently from generative rewriting. Check the actual feature being used and the applicable policy before concluding that disclosure is unnecessary.
Level 2 · Writing assistance
Language refinement, translation, and restructuring
Describe the tool’s role in the declaration, acknowledgments, or other section requested by the publisher. Check that revisions preserve the intended meaning, argument, and attribution.
Level 3 · Research assistance
Screening, coding, analysis, or synthetic data
Report substantive methodological use in Methods or the equivalent section, with enough detail to understand the procedure, validation, and limitations. Identify synthetic data explicitly and follow applicable research approvals and reporting standards.
The ESR statement, for example, distinguishes writing/editing disclosures from reporting AI used to generate research results (Hamm et al., 2024). This illustrates why placement should be checked against the destination’s instructions.
For Lumina Literati submissions: our Plagiarism and AI Policy requires authors to identify the AI tools used and describe their contribution to manuscript preparation. Consult that policy together with the author instructions.
4. Linguistic equity and institutional governance
Language assistance and intellectual authorship should be considered carefully, especially for scholars writing in an additional language. In a qualitative study involving three EFL students, Werdiningsih et al. (2024) explore experiences of ChatGPT-assisted academic writing, including language support and concerns about authenticity. The small sample offers useful experiences to consider; it cannot establish how multilingual researchers generally use AI.
For writers, a practical approach is to keep a record of the ideas and arguments they developed, compare revised text with the original, and reject changes that alter meaning or introduce unsupported claims. Assistance with English should not be mistaken for evidence that the underlying scholarship belongs to the tool.
Institutions also have responsibilities. Drawing on consultation at two Australian universities, Smith et al. (2026) propose a framework grounded in honesty, transparency, and accountability, supported by training, communication, infrastructure, and process change.
Kamali et al. (2024) examine educators’ AI ethics perspectives through activity theory and identify a need for professional development and collaborative policy work. This supports attention to the conditions in which people make ethical decisions, rather than assuming that a policy document alone resolves uncertainty.
For early career researchers, useful support includes clear examples of acceptable reporting, access to approved tools, advice on handling sensitive material, and an opportunity to discuss uncertainties with supervisors or editors before submission.
Practical disclosure templates
Adapt these examples to the work actually performed and the destination’s policy. Replace the bracketed fields; include only statements you can substantiate.
Template A: Language and readability assistance
Suggested location: the AI declaration, acknowledgments, or other section specified by the publisher.
During preparation of this manuscript, the authors used [tool name, model/version where available, and date or date range] to [describe the language or translation task] in [identify sections]. The authors reviewed and revised the output for meaning, accuracy, and appropriate attribution and take responsibility for the final manuscript. [If accurate: The tool was not used to generate research findings, interpret data, or identify references.]
Template B: Methodological or coding assistance
Suggested location: Methods or its equivalent, with any additional declaration required by the publisher.
We used [tool name, provider, model/version where available, and date or date range] to assist with [specific research task]. The tool received [describe inputs without disclosing confidential information] and produced [describe outputs]. [Responsible researcher(s)] evaluated the outputs using [validation procedure, extent of checking, and criteria]. We addressed identified errors by [describe corrections] and retained responsibility for the analysis and interpretation. Relevant prompts, settings, and validation records are [state their actual availability or explain access restrictions].
Do not claim that every output was independently validated, that an appendix exists, or that data were never shared unless those statements accurately describe your workflow.
Before you submit
- Check the destination’s rules. Review its AI, authorship, data, image, and confidentiality policies.
- Keep a task record. Note the tools, dates, inputs, outputs, and substantive changes involved.
- Verify evidence. Open cited sources and check that they support the claims made; test code and inspect analytical results appropriately.
- Describe human contributions. Agree on authorship and contributor roles with coauthors.
- Disclose accurately. Match the declaration to the actual work, including images or analysis when relevant.
- Review the final manuscript. Ensure the named authors understand and accept responsibility for the submitted content.
Frequently asked questions
Which early paper should I read first?
Hosseini et al.’s Using AI to write scholarly publications is a useful starting point for disclosure and accountability. It appeared online in January 2023 and in a 2024 journal issue. It is an early contribution, not a single source that established all subsequent ethical rules.
Why can’t ChatGPT be listed as a coauthor under COPE and ICMJE guidance?
These guidelines connect authorship with responsibility for the work. AI tools cannot fulfill that role. Describe their use transparently while keeping authorship and accountability with the people who meet the applicable criteria. See COPE and ICMJE.
Does using AI for grammar always require disclosure?
There is no single answer across all publishers and tools. Basic mechanical corrections may be treated differently from generative rewriting. Check the relevant policy and the function actually used. For a Lumina Literati submission, consult the AI policy and disclose the tools and their contribution as required.
How should early career researchers document AI use?
Retain drafts and an appropriate task log, record the tool and model where available, document substantive prompts and validation, and make an accurate disclosure. Store records in ways that respect confidentiality and institutional requirements.
Does CRediT make an AI tool eligible for authorship?
No. CRediT describes contributor roles; it does not determine authorship eligibility. Use it to clarify people’s contributions and report AI assistance according to the publisher’s instructions.
Publishing with integrity
Prepare your work for Lumina Literati Publishing
Planning a research monograph, edited collection, or book proposal? Review our submission requirements and editorial policies, and describe any AI assistance clearly when you submit.
References
References use final issue years. Earlier online dates explain why some sources are also cited elsewhere with a preceding year. This is a selected narrative guide, not a systematic review.
- Aggarwal, R. (2025). Addressing authorship dilemmas in scholarly publications: A solution-oriented study. Studies in Higher Education, 50(9), 2011–2029. https://doi.org/10.1080/03075079.2024.2407947
- Cotton, D. R. E., Cotton, P. A., & Shipway, J. R. (2024). Chatting and cheating: Ensuring academic integrity in the era of ChatGPT. Innovations in Education and Teaching International, 61(2), 228–239. https://doi.org/10.1080/14703297.2023.2190148
- Donato, H., Escada, P., & Villanueva, T. (2023). The transparency of science with ChatGPT and the emerging artificial intelligence language models: Where should medical journals stand? Acta Médica Portuguesa, 36(3), 147–148. https://doi.org/10.20344/amp.19694
- Hamm, B., Marti-Bonmati, L., & Sardanelli, F. (2024). ESR Journals editors’ joint statement on guidelines for the use of large language models by authors, reviewers, and editors. European Radiology Experimental, 8, Article 7. https://doi.org/10.1186/s41747-023-00420-2
- Hosseini, M., Rasmussen, L. M., & Resnik, D. B. (2024). Using AI to write scholarly publications. Accountability in Research, 31(7), 715–723. https://doi.org/10.1080/08989621.2023.2168535
- Kamali, J., Alpat, M. F., & Bozkurt, A. (2024). AI ethics as a complex and multifaceted challenge: Decoding educators’ AI ethics alignment through the lens of activity theory. International Journal of Educational Technology in Higher Education, 21, Article 62. https://doi.org/10.1186/s41239-024-00496-9
- Lin, Z. (2025). Modernizing authorship criteria and transparency practices to facilitate open and equitable team science. Accountability in Research, 32(8), 1564–1587. https://doi.org/10.1080/08989621.2024.2405041
- Resnik, D. B., & Hosseini, M. (2025). The ethics of using artificial intelligence in scientific research: New guidance needed for a new tool. AI and Ethics, 5(2), 1499–1521. https://doi.org/10.1007/s43681-024-00493-8
- Smith, S. M., Tate, M., Freeman, K., Walsh, A., Ballsun-Stanton, B., & Lane, M. (2026). A university framework for the responsible use of generative AI in research. Journal of Higher Education Policy and Management, 48(1), 17–36. https://doi.org/10.1080/1360080X.2025.2509187
- Suchikova, Y., & Tsybuliak, N. (2025). ChatGPT isn’t an author, but a contribution taxonomy is needed. Accountability in Research, 32(8), 1590–1595. https://doi.org/10.1080/08989621.2024.2405039
- van Woudenberg, R., Ranalli, C., & Bracker, D. (2024). Authorship and ChatGPT: A conservative view. Philosophy & Technology, 37, Article 34. https://doi.org/10.1007/s13347-024-00715-1
- Werdiningsih, I., Marzuki, & Rusdin, D. (2024). Balancing AI and authenticity: EFL students’ experiences with ChatGPT in academic writing. Cogent Arts & Humanities, 11(1), Article 2392388. https://doi.org/10.1080/23311983.2024.2392388
Policy and contributor resources
- Committee on Publication Ethics. Authorship and AI tools.
- International Committee of Medical Journal Editors. Use of AI by authors.
- CRediT / NISO. Contributor role descriptors.
Policy resources checked on 28 October 2026. Publisher requirements can change; consult the current version when preparing a submission.