AI Use Policy - Gatha Cognition

Generative artificial intelligence (AI) tools — large language models such as ChatGPT, Claude, Gemini, and similar systems, as well as AI-assisted image, data-analysis, and coding tools — are now in routine use across academic writing, peer review, and editorial work. This policy sets out how such tools may and may not be used in work submitted to, reviewed for, or produced by Gatha Cognition’s journals and books, so that authors, reviewers, and editors share a common, transparent standard.

This policy applies to:

• Authors submitting manuscripts, revisions, or supplementary material to any Gatha Cognition journal or book.
• Peer reviewers evaluating manuscripts on our behalf.
• Editors and editorial staff (including editorial board members) handling submissions at any stage of the workflow.

It covers any tool that generates, transforms, or evaluates text, images, code, or data using machine learning — whether described as “AI,” “generative AI,” or a “large language model.” It does not apply to conventional, deterministic software such as spell-checkers, reference managers, or plagiarism-detection tools, which are addressed under our existing editorial policies.

1. Authorship and Accountability

• AI tools cannot be listed as an author or co-author on any manuscript, book, or chapter published by Gatha Cognition. Authorship carries accountability — the ability to approve a final version, take public responsibility for the work, and be held accountable for its accuracy and integrity — which an AI tool cannot hold or exercise.
• Every named human author remains fully responsible for the entire content of a submission, including any portion drafted, edited, translated, or analyzed with AI assistance. “The AI made an error” is not an acceptable explanation for inaccurate, fabricated, plagiarized, or biased content.
• Authors must independently verify every factual claim, citation, and reference produced or suggested by an AI tool before submission. AI language models are known to generate plausible-sounding but fabricated citations and facts (“hallucinations”); Gatha Cognition treats an unverified AI-fabricated citation the same as any other case of citation fabrication under our research-integrity policy.
• This section applies equally to editors and reviewers: an AI tool cannot be credited as a reviewer or listed as contributing to an editorial decision.
• Generative AI cannot be cited as a source or reference. A statement like “ChatGPT, personal communication” is not an acceptable citation — AI output is not a verifiable, retrievable source; cite the underlying verified literature instead.

2. Policy for Authors

Generally permitted, with disclosure:

• Improving language, grammar, and readability of the authors’ own already-written text.
• Summarizing background literature that authors have themselves read and verified, to help structure a section.
• Assistance with code, statistical scripts, or routine data-formatting tasks, where the author reviews and tests the output.
• Brainstorming structure, titles, or abstract phrasing.

Not permitted, under any circumstances:

• Using AI to generate the substantive scientific content, results, discussion, or conclusions of a study — the intellectual contribution must be the authors’ own.
• Submitting AI-generated text, data, or analysis without disclosure, or presenting AI-generated content as original human-authored work.
• Using AI to fabricate, alter, or embellish data, results, or experimental findings.
• Using AI to generate peer-review responses or rebuttal letters that misrepresent the authors’ actual reasoning.

Per ICMJE, authors must be able to assert there is no plagiarism anywhere in the paper, including in text and images produced by an AI tool, and must ensure full, correct citations for all attributed material — AI-generated content is not exempt from the usual plagiarism check.

Authors who use a generative AI tool anywhere in preparing a submission must disclose it per the Disclosure Requirements section below. Non-disclosure discovered after submission or publication will be treated as a research-integrity concern.

3. AI-Generated Images, Figures, and Data

• AI-generated or AI-altered images, figures, videos, or illustrations may not be used to represent real research data (e.g., a generated image standing in for an actual micrograph, gel, chart, or photograph). Doing so is treated as data fabrication.
• AI tools may be used to produce genuinely illustrative or conceptual graphics (for example, a schematic diagram or a graphical abstract) only with clear disclosure in the figure caption and Methods/Acknowledgments section, stating the tool used and that the image is illustrative, not primary data.
• AI-assisted data analysis (e.g., using an AI tool to help write or debug an analysis script) is permitted only when the author independently verifies the underlying calculations and results; the AI tool itself is never an acceptable source of a reported result.
• Authors remain responsible for ensuring any AI-generated visual or analytical content does not infringe third-party copyright or reproduce another author’s unpublished data.

4. Policy for Peer Reviewers

• Manuscripts under review are confidential, unpublished work. Reviewers must not upload a manuscript, its figures, data, or any substantive excerpt to ChatGPT, Claude, Gemini, or any other third-party AI tool. Doing so discloses confidential, unpublished material to an external party and is a breach of the reviewer confidentiality agreement, regardless of that tool’s own privacy settings.
• AI tools must not be used to generate the substantive content of a review — the scientific judgment, critique, and recommendation must be the reviewer’s own independent assessment.
• Reviewers may use AI-based grammar or language tools on their own already-written review text, provided no manuscript content is pasted into the tool, and provided this is disclosed to the handling editor if asked.
• Any reviewer found to have shared manuscript content with an AI tool will be removed from that review and reported to the editor-in-chief; repeated or serious breaches may result in removal from the reviewer pool.

5. Policy for Editors and Editorial Staff

• The same confidentiality restriction as above applies in full: unpublished manuscript content, author identities, reviewer reports, and any confidential correspondence must never be entered into a third-party AI tool, including for tasks like drafting decision letters, unless using an internal, contractually confidential enterprise AI deployment explicitly approved by Gatha Cognition for that purpose.
• AI must not make or substitute for an editorial decision. Accept, reject, revise, and desk-reject decisions must reflect the editor’s own scholarly judgment, informed by reviewer reports and editorial board input — not generated or decided by an AI tool.
• AI tools may be used for administrative and communication support that does not involve confidential manuscript content — for example, drafting generic template language, formatting checklists, or summarizing publicly available literature for background context.
• Editors are responsible for checking, at initial submission, that authors have completed any required AI-use disclosure before a manuscript proceeds to review.
• Human in the loop for any automated integrity checks. Gatha Cognition runs every submitted manuscript through similarity/plagiarism-detection screening (e.g., Turnitin) as a standard, disclosed part of the editorial process, with results shared with the handling editor and made available to reviewers and the author as needed. This routine screening is a contracted, confidentiality-bound service integrated into our own submission system — not the same category of risk as pasting a manuscript into a public chatbot — so it is not restricted by the confidentiality rule above, but the same human-in-the-loop principle still applies: a similarity score is a prompt for editorial judgment, never an automatic verdict on its own. The same applies to any future automated tool for figure manipulation/duplication checks, undeclared-AI detection, or automated reviewer suggestions.

5a. Publisher-Side Production Use of AI (Formatting, Copyediting, Graphical Abstracts, Highlights)

Gatha Cognition’s own editorial/production team may use AI tools to assist with template composition and layout, language and copyediting, and preparing supplementary elements such as graphical abstracts and highlights. This is treated differently from the peer-review confidentiality restriction above, because it happens after a manuscript is accepted (or well into production), not during confidential evaluation — but it is still governed by clear limits, following COPE case 26-03 (2026), “Should there be limits to the editorial use of assistive AI tools?”:

• Stay assistive, not generative. Permitted: formatting/layout into the journal or book template, mechanical language correction, spacing and citation-formatting fixes, consistency checks — the kind of change a human copyeditor or typesetter would make, without altering the author’s meaning, arguments, or voice. Not permitted without the author’s explicit involvement: AI rewriting that changes framing, emphasis, or authorial voice, even if no new ideas are introduced.
• Disclose the practice to authors upfront. Our author guidelines and submission process will state that Gatha Cognition uses AI-assisted tools in production (formatting, copyediting, graphical abstracts, highlights), so authors can make an informed choice to submit on that basis.
• Author review and approval before publication. Any AI-assisted work Gatha Cognition applies to author-submitted content is presented back to the corresponding author for review and approval — the same way proofs are today — before the final version is published. Nothing AI-assisted is published without the author having seen and agreed to it.
• Confidentiality still applies. Even at the production stage, manuscript content should only go through AI tools covered by Gatha Cognition’s own data-handling terms with the provider (not an ad hoc public consumer tool).
• Tools should be appropriately tested before routine use. Any AI tool adopted for routine production work should be evaluated on real (or representative) manuscripts first, so its output quality is known before it touches live author content.

6. Disclosure Requirements and Documentation

Authors who used any generative AI tool in preparing a submission must disclose it in two places:

1. Cover letter at submission — a brief statement of which tool(s) were used and for what purpose.
2. Methods (or Acknowledgments, for article types without a Methods section) of the manuscript itself — a statement naming the tool, its version/provider where known, and the specific purpose (e.g., “Language editing of the manuscript draft was assisted by [tool name, version]. All content was reviewed and verified by the authors, who take full responsibility for the final text.”).

Disclosure should specify what the tool was used for (language editing, literature summarization, coding assistance, image generation, data analysis, etc.), not just that “AI was used.” Routine, non-generative tools (spell-check, grammar-check plugins with no generative output, reference managers) do not need to be disclosed.

Reviewers and editors: any permitted AI use under the sections above (e.g., language-polishing a reviewer’s own comments) should be mentioned to the handling editor in the same submission where it occurred, so it is on record.

For AI used to generate substantive new text, analysis, or figures (not just light language editing), authors should also record the specific prompt(s) used — this level of detail is recommended, not mandatory, but makes any later question about the content much easier to resolve.

7. Non-Compliance and Consequences

How a concern is handled depends on when and what is found:

• Undisclosed AI use, discovered before publication: the editor will request the author add the required disclosure and re-verify the affected content; the manuscript is held pending this correction.
• AI-fabricated data, citations, or results, discovered at any stage: treated as a research-integrity/misconduct concern under our existing editorial-misconduct process — may result in rejection, and, if discovered post-publication, a correction or retraction per COPE guidelines, with the author’s institution notified where appropriate.
• Confidentiality breach by a reviewer or editor (manuscript content shared with a third-party AI tool): immediate removal from that review/assignment, a report to the editor-in-chief, and, for serious or repeated breaches, removal from the reviewer or editorial pool.
• Undisclosed AI use discovered after publication, where the content itself is accurate and not fabricated: typically handled with a published correction adding the required disclosure, rather than retraction — retraction is reserved for cases involving fabrication, plagiarism, or serious misconduct.

All cases are evaluated case-by-case by the editor-in-chief, consistent with COPE’s guidance on handling suspected AI-policy violations.

8. Alignment with Industry Standards

This policy follows the current consensus positions of the major academic-publishing integrity bodies:

• COPE — “Authorship and AI tools” (COPE position, last reviewed 13 February 2023; cite as COPE Council, https://doi.org/10.24318/cCVRZBms): “AI tools cannot be listed as an author of a paper… [they] cannot meet the requirements for authorship as they cannot take responsibility for the submitted work.” Authors using AI in writing, images/graphics, or data collection and analysis “must be transparent in disclosing in the Materials and Methods (or similar section)… how the AI tool was used and which tool was used,” and “are fully responsible for the content of their manuscript, even those parts produced by an AI tool.”
• ICMJE — “Artificial Intelligence (AI)-Assisted Technology” and “Use of Artificial Intelligence in Publishing”: chatbots “should not be listed as authors because they cannot be responsible for the accuracy, integrity, and originality of the work.” “Editors, reviewers, or publishers should not upload submitted manuscript into AI systems where confidentiality cannot be assured without authors’ explicit permission.”
• WAME (World Association of Medical Editors) — “Chatbots, Generative AI, and Scholarly Manuscripts” (revised 31 May 2023): chatbots cannot be authors; they “retain the prompts fed to them, including manuscript content, and supplying an author’s manuscript to a chatbot breaches confidentiality.”
• DOAJ (Directory of Open Access Journals) — Guide to Applying, “Artificial Intelligence (AI) and automated tools” (v2.9): expects journals to publish an AI-use policy covering authors and journals, including “human in the loop” oversight of automated integrity checks.

This policy will be reviewed periodically as AI tools, journal-industry norms, and COPE/ICMJE/WAME guidance continue to evolve. Questions about a specific use case not clearly covered here should be directed to the editorial office before submission.