03 · What You Need to Know
Where Generative AI Can Assist in Preparing a Journal Submission
Manuscript Completion and Submission Readiness Are Different Milestones
A scientifically complete manuscript may still fail to satisfy a journal's administrative and technical submission requirements.
For example, the manuscript may exceed the permitted word count, contain figures in unsupported formats, omit a required ethics statement, or use an abstract structure intended for another journal.
These problems do not necessarily indicate weaknesses in the research itself. Nevertheless, they may delay editorial screening or require authors to correct the submission before the manuscript proceeds further.
Generative AI can help identify such issues, particularly when researchers provide the relevant manuscript materials and current journal instructions.
Its appropriate role is to assist with preparation and checking rather than certify that the submission meets every requirement.
What Parts of Journal Submission Preparation Can AI Support?
Journal submission preparation involves several distinct activities. Some are primarily organizational, while others require verification of scientific facts, author agreements, or institutional records.
| Submission Task |
Potential AI Assistance |
What Authors Must Verify |
| Author guidelines |
Extract and organize stated submission requirements. |
Completeness and accuracy against current official instructions. |
| Manuscript formatting |
Identify apparent problems with headings, structure, and references. |
Actual compliance with the required format and file specifications. |
| Language editing |
Improve clarity, grammar, and readability. |
Preservation of scientific meaning, terminology, and intended claims. |
| Cover letter |
Prepare a draft using verified manuscript details. |
Accuracy of novelty, contribution, declarations, and journal-specific statements. |
| Highlights and keywords |
Suggest concise summaries and relevant search terms. |
Faithfulness to the research and compliance with journal limits. |
| Author declarations |
Organize or draft statements from supplied facts. |
Actual contributions, conflicts, approvals, and data conditions. |
| Submission checklist |
Identify required documents and track preparation status. |
Whether every required item exists and is complete. |
| Submission metadata |
Help prepare titles, abstracts, affiliations, and other entry fields. |
Exact agreement with approved manuscript and author records. |
The distinction between assistance and verification becomes especially important when a task involves declarations that carry ethical, legal, or institutional consequences.
Start With the Journal's Current Instructions for Authors
Journal requirements differ among publishers, disciplines, article types, and submission systems.
Some journals require structured abstracts, while others accept unstructured summaries. Reference formatting may be strictly prescribed or relatively flexible during initial submission.
Requirements for tables, figures, supplementary materials, reporting checklists, and author declarations may also differ.
Researchers should therefore obtain the current instructions from the journal's official website before asking AI to evaluate compliance.
AI can help convert these instructions into a practical checklist, but it should distinguish requirements explicitly stated in the source from recommendations based on general publishing practice.
The task of checking a manuscript against journal author guidelines requires attention to the actual documents rather than assumptions about what academic journals usually require.
AI Can Help Check Manuscript Structure and Formatting
Generative AI may identify apparent inconsistencies in manuscript organization, such as missing sections, irregular heading levels, inconsistent terminology, or references that do not follow a supplied format.
It can also help researchers prepare a list of elements requiring attention, including tables, figures, supplementary files, and acknowledgments.
However, AI may not accurately inspect every aspect of a formatted document. Page layout, embedded objects, figure resolution, line numbering, tracked changes, and file compatibility may require direct inspection using appropriate document software.
Researchers should also distinguish between submission formatting requirements and scientific reporting requirements.
A manuscript can follow the journal's preferred layout while still omitting essential methodological information.
AI-assisted formatting checks should therefore complement, not replace, substantive review of the research.
Language Editing Must Preserve the Scientific Meaning
AI can improve sentence clarity, reduce unnecessary repetition, and suggest more consistent terminology.
These functions may be particularly helpful when authors are preparing manuscripts in a language other than their first language.
However, language editing can introduce substantive changes even when the request appears limited to grammar.
For example, AI might change "was associated with" to "influenced," thereby implying causation where the study established only an association.
It might also strengthen tentative conclusions, remove important limitations, or replace a precise methodological term with a more familiar but inaccurate expression.
Authors should review AI-edited passages against the original text, data, and intended interpretation.
The goal is clearer scholarly communication without altering what the research actually demonstrates.
AI Can Draft a Cover Letter, but It Must Not Invent Contributions
Many journals request or permit a cover letter explaining the manuscript's contribution and suitability for the publication.
Generative AI can help organize this letter, improve its tone, and present the manuscript's significance concisely.
However, AI may exaggerate novelty, invent claims about previous research, or state that the manuscript satisfies requirements the authors have not checked.
For example, an AI-generated letter might describe a study as the first investigation of a particular phenomenon without evidence supporting that claim.
Researchers should ensure that AI-assisted submission cover letters accurately represent the manuscript and contain only verified declarations.
Claims concerning originality, exclusive submission, previous publication, and author approval should reflect the actual circumstances.
Highlights and Keywords Can Be Prepared With AI Assistance
Some journals request research highlights, key points, or keywords in addition to the abstract.
AI can suggest concise statements identifying the principal findings and contributions. It may also propose terminology relevant to the manuscript's subject area.
Nevertheless, researchers should ensure that highlights do not exaggerate findings or introduce conclusions unsupported by the study.
Keywords should accurately represent the research rather than simply include fashionable terms intended to increase discoverability.
When a journal specifies the number, length, or format of highlights and keywords, those requirements should govern the final submission.
Author Contribution Statements Require Actual Author Information
Many journals request information about the contributions made by individual authors. Some use the Contributor Roles Taxonomy (CRediT), which provides standardized descriptions of research contributions.
AI can help organize author-provided information into the required format.
However, it cannot independently determine who designed the study, collected data, performed analyses, obtained funding, or approved the manuscript.
Those facts must come from the research team.
Researchers should not allow AI to infer contributions from author order, academic rank, or assumptions about typical research roles.
Each author should verify the contribution information relevant to them, and the final statement should reflect the journal's applicable requirements.
Conflict-of-Interest Statements Cannot Be Invented
Journals may require authors to disclose financial or nonfinancial relationships that could be relevant to the submitted research.
Generative AI can explain disclosure terminology and help draft a statement using information supplied by the authors.
It should not assume that no conflicts exist merely because none were mentioned in the manuscript.
Similarly, AI should not generate a declaration that all authors have no competing interests unless the authors have confirmed that statement.
Disclosure requirements vary by journal and may include relationships, activities, or interests that extend beyond direct research funding.
The final statement should be based on actual author disclosures and the journal's current policy.
Data Availability Statements Must Reflect Real Access Conditions
Some journals require a data availability statement describing whether and how research data can be accessed.
AI can help draft the statement, but it cannot establish whether a dataset has actually been deposited, whether a repository link works, or whether sharing is permitted.
For example, an AI-generated statement might claim that data are publicly available in a repository when the authors have not uploaded them.
Other studies may involve confidential participant information, contractual restrictions, or ethical approvals that limit data sharing.
Researchers should verify the dataset's actual availability and applicable restrictions before approving the statement.
A data availability declaration is a factual representation of the research record, not a generic paragraph to be completed automatically.
Ethics Approval and Consent Information Require Particular Care
Research involving human participants, animals, or other regulated activities may require ethics approval, informed consent documentation, or additional declarations.
Generative AI can help identify where these statements belong in the manuscript and organize information supplied by the research team.
However, it must not invent ethics approval numbers, committee names, consent procedures, or exemptions.
Researchers should verify statements against actual approval documents and applicable institutional requirements.
Where an ethics determination is unclear, the responsible institutional body should provide clarification.
AI-generated explanations cannot substitute for an approval or exemption that was never obtained.
AI Can Help Organize Reporting Guideline Requirements
Some journals require authors to follow reporting guidelines appropriate to their research design.
Examples include CONSORT for randomized trials, STROBE for observational studies, and PRISMA for systematic reviews.
The EQUATOR Network provides a searchable collection of health research reporting guidelines.
AI can help identify potentially relevant guidelines and organize a manuscript assessment against supplied checklist items.
However, the applicability of a guideline depends on the actual study design and journal requirements.
Researchers should consult the original reporting guideline and verify that each claimed item is genuinely addressed in the manuscript.
Completing a checklist does not automatically establish that the underlying study was appropriately designed or conducted.
Can AI Reliably Identify Every Required Submission Document?
Not necessarily. Journal submission requirements may be distributed across multiple webpages, downloadable instructions, editorial policies, and the submission system itself.
Some requirements may depend on article type, research design, funding arrangements, or the journal's peer-review model.
For example, a journal using double-anonymized review may require a manuscript file that excludes identifying information, while another journal may require a different arrangement.
AI may overlook conditional requirements or apply instructions intended for another article category.
Researchers should therefore use AI-generated checklists as working documents rather than assume that every submission requirement has been identified.
Journal Policies on AI Assistance May Require Disclosure
Publishers and journals may impose requirements concerning the use of generative AI in manuscript preparation.
The International Committee of Medical Journal Editors (ICMJE) addresses the use of AI-assisted technologies in scholarly work and emphasizes that human authors remain responsible for the accuracy and integrity of submitted content.
Publisher policies may distinguish language assistance from substantive content generation and may specify where AI use should be disclosed.
Researchers should consult the current policy of the target journal rather than assume that disclosure rules are identical across publishers.
AI systems should not be listed as authors because they cannot assume the responsibilities associated with scholarly authorship.
Disclosure, where required, does not remove the obligation to verify the manuscript and supporting materials.
Protect Confidential Manuscripts and Research Data
Preparing a submission with AI may involve sharing unpublished research, datasets, reviewer-related information, or institutional documents.
Researchers should consider whether the selected AI service is appropriate for the information being processed.
Participant data, confidential research findings, proprietary information, and potentially patentable material may be subject to restrictions.
Before uploading documents, examine applicable institutional policies, research agreements, data protection requirements, and the provider's data-handling conditions.
Where possible, use appropriately sanitized material or an approved environment that meets the relevant confidentiality requirements.
Administrative convenience does not justify disclosing information that researchers are obligated to protect.
The Final Submission Must Be Checked Against the Actual Files
A common limitation of AI-assisted preparation is the difference between a proposed correction and a completed correction.
AI may suggest adding a declaration, renaming a figure, or revising a reference without actually modifying the final file.
Similarly, a generated checklist may mark an item complete based on a statement in the conversation rather than evidence in the manuscript.
Researchers should inspect the actual submission files, not merely the AI-generated summary of what supposedly has been prepared.
The final verification should cover the manuscript, supplementary files, declarations, submission metadata, and any required supporting documents.
This is the central principle behind independently checking AI-assisted journal submission materials before submission.
