Policy Note: Current as of 18 June 2026
This guide reflects publisher policies and publication ethics guidance reviewed in June 2026. AI policies may change, and individual journals may set stricter requirements than their parent publisher. Always verify the latest Instructions for Authors, submission portal questions, figure and image guidelines, and peer review policies for your target journal before submission.

AI Publishing Policies Have Become Submission-Governance Rules

This 2026 update reflects how AI policies in academic publishing have moved from broad principles into practical submission rules. In 2025, the core question was whether publishers allowed tools like ChatGPT. In 2026, the focus: how research teams document, verify, and disclose AI assistance before journal submission.

This update is prompted by a documented transparency gap in scientific publishing. Springer Nature Group reported that AI tools are already part of many research and publishing workflows, while one third of surveyed researchers had never disclosed their AI use when submitting or publishing. Citation accuracy is another pressure point. Retraction Watch reported on an audit finding that roughly one in 277 PubMed-indexed papers published in the first seven weeks of 2026 referenced a nonexistent paper. In this context, publishers are placing more emphasis on documentation, verification, and integrity checks before publication.

For labs and research groups, managing academic writing in the age of AI is now a matter of routine data governance. This guide explains where AI disclosures may need to appear, which basic editing uses may be exempt, how figure rules have become more specific, and how research teams can audit AI use before submission.

Track AI Contributions in Coauthor

Subscribe to thesify to access Coauthor and test AI contribution tracking in a shared manuscript workspace. Coauthor helps research teams keep drafting, revision, and contribution decisions visible before journal submission.

Quick Pre-Submission AI Compliance Checklist

Use this journal AI policy checklist after the manuscript is assembled and before the corresponding author begins the submission process. The aim is to identify disclosure, authorship, figure, confidentiality, and documentation issues before they become submission problems.

For broader guidance on responsible AI-supported writing practices, see thesify’s guide to ethical use cases of AI in academic writing.

Before submission, confirm the following:

• No AI tool is listed as an author: Check the title page, author contribution statement, acknowledgements, metadata, and any submission-form fields. AI tools should not be credited as authors or cited as authors.

• All substantive AI use has been identified: Review whether AI was used for drafting, revising, translation, literature synthesis, coding support, data analysis support, figure preparation, data visualization, reference checking, cover-letter drafting, or reviewer-response drafting.

• Grammar-only or copy-editing use has been separated from substantive use: Many publishers distinguish basic grammar, spelling, punctuation, and narrowly defined copy editing from AI use that shapes content, structure, interpretation, translation, analysis, or visuals. Do not assume an exemption applies until you have checked the target journal’s policy.

• The publisher-level AI policy has been reviewed: Confirm the baseline rules for authorship, disclosure, figures, data, peer review, confidentiality, and tool terms.

• The journal-specific instructions have been reviewed: Check the target journal’s Instructions for Authors, publishing ethics page, image guidelines, figure preparation rules, graphical abstract instructions, cover-art policy, and submission portal questions. Journal-level rules may be stricter than the publisher’s general policy.

• The disclosure location is clear: Identify whether AI use belongs in the Methods section, Acknowledgements, a dedicated AI declaration, a figure caption, the cover letter, the submission form, or more than one location.

• The disclosure includes the necessary details: At minimum, record the tool name, version or model, purpose, affected manuscript section, output retained, and verification step. Add manufacturer, prompt history, date of use, or validation details where the journal requires them.

• Figures and visual materials have been audited separately from text: Review all figures, tables, charts, data visualizations, visual abstracts, graphical abstracts, cover art, and primary research images. Confirm whether AI was used to create, edit, label, redraw, clean, translate, or format any visual material.

• Primary research images and data-derived outputs have been checked against the strictest applicable rule: Treat microscopy, histology, radiology, western blots, gels, patient images, clinical images, and other evidence-bearing visuals as high-risk materials. If AI was used in any part of the workflow, confirm that the method is permitted, reproducible, disclosed, and supported by source files.

• AI tool terms have been reviewed before submission: Check whether the tool’s terms address confidentiality, data retention, model training, ownership, copyright, output reuse, privacy, and publication rights. Avoid entering unpublished manuscript material, third-party copyrighted material, patient information, or identifiable data into tools with unclear terms.

• Peer review confidentiality has not been breached: If you acted as a reviewer or editor, confirm that no unpublished manuscript, figure, table, supplementary file, reviewer report, or editorial correspondence was uploaded into a public AI system. If AI was used in a permitted private or journal-approved environment, record and disclose that use according to the journal’s policy.

• Coauthors have signed off on AI use: Ask each coauthor to confirm whether they used AI, what material they entered, what output was retained, which manuscript sections were affected, and how the output was checked.

• The final AI disclosure statement and supporting records have been saved: Keep a copy of the AI use log, final disclosure wording, source files for figures, prompt or output records where relevant, and any coauthor confirmations. These records may be useful if the editor requests clarification during review.

Verifying Policy Alignment Behind the Checklist

These checklist items reflect recurring requirements across major publisher policies and publication-ethics guidance:

• Authorship and Accountability: Major publishers and publication-ethics bodies reject AI authorship and keep responsibility with authors, including ICMJE, Springer Nature, Elsevier, and Wiley.

• Disclosure Formats and Locations: Disclosure placement has become more specific in 2026. ICMJE recommends disclosure at submission and description in both the cover letter and the submitted work, where applicable. Elsevier uses a dedicated AI declaration, while Wiley requires disclosure at submission and in the manuscript where relevant.

• The Copy-Editing Exemption: Clear boundaries for non-reportable assistive AI (grammar, readability, spelling) are defined explicitly by Springer Nature's copy editing guidelines and Elsevier's author policies.

• Visual Data Controls: Elsevier, Wiley, and Springer Nature distinguish more clearly between primary research images, data-derived visuals, and conceptual or illustrative materials.

• Peer Review and Terms Verification: Requirements to protect confidentiality and explicitly audit the legal data usage terms of external software are dictated by Wiley, Elsevier, and Springer Nature.

What Changed in Academic Publishing AI Policies in 2026?

The 2026 update is not a reversal of the basic publishing position on AI. Major publishers and editorial bodies still converge on three core principles: AI tools cannot be authors, human authors remain responsible for the submitted work, and substantive AI use must be disclosed. 

Academic publishing AI guidelines 2026 now focus directly on how AI use is handled operationally at submission: where it is disclosed, which uses are exempt, how visual material is reviewed, what reviewers may do with AI tools, and what records authors must maintain.

Disclosure has become more structured. The International Committee of Medical Journal Editors (ICMJE) now gives biomedical journals a clear, high-authority reference point by stating that authors should disclose AI use at submission and describe it explicitly in both the cover letter and the submitted work, where applicable. Similarly, the World Association of Medical Editors (WAME) provides guidance that is highly operational because it links disclosure directly to function: drafting, data analysis, code generation, figure preparation, review reports, and editorial correspondence each carry distinct reporting expectations.

Publisher policies have also become much more specific about non-text outputs. Figures, tables, charts, data visualizations, graphical abstracts, cover art, and primary research images are no longer treated as a single category. For authors, this means AI-assisted language editing and AI-assisted visual production cannot be reviewed under the same broad rule. A figure based on observed data, a conceptual diagram, and a graphical abstract may each require a separate policy check depending on the target publisher and journal.

Reviewer and editor use of AI has also become a separate compliance issue. Author-facing rules may allow limited AI support for drafting, editing, or translation, but peer review rules are usually stricter because unpublished manuscripts are confidential. Reviewers may be required to follow journal policy, request permission, disclose AI use, or avoid AI tools where confidentiality cannot be assured.

Another major change is the expansion of research integrity screening. Publishers are using screening tools to identify potential integrity risks before publication, including reference problems, image concerns, authorship anomalies, conflicts of interest, and paper-mill signals.

This infrastructure does not remove editorial judgment, but it raises the practical value of keeping accurate records. A research team that can show what AI was used for, where it affected the manuscript, and how outputs were verified is far better prepared for submission queries.

The Transparency Gap Is Now a Submission Risk

AI adoption and formal AI disclosure are not moving at the same pace, creating a distinct "transparency gap" in scholarly publishing. Data published by the Springer Nature Group indicates that AI tools are already part of many research and publishing workflows, especially for translation, summarization, manuscript writing, editing, data analysis, and modeling. The same data reports that one third of surveyed researchers had never formally disclosed their AI use when submitting or publishing their work.

For authors, this gap is an immediate submission risk. If AI use is common but the disclosure is reconstructed only at the end of the project, important details are easily lost. Research teams frequently struggle to remember which tool was used, which version, which prompt, which output, which exact manuscript section was modified, and which verification step was taken. This missing documentation creates avoidable uncertainty and policy risk at submission, especially for multi-author papers.

ICMJE and WAME Make Disclosure More Operational

The ICMJE Recommendations provide a concrete disclosure standard for biomedical publishing. They state that journals should require authors to disclose at submission whether AI-assisted technologies were used and that authors should describe this use in both the cover letter and the submitted work, where applicable. ICMJE also states that AI-assisted tools should not be listed as authors, that human authors remain responsible for submitted material, and that nondisclosure may require corrective action or be construed as misconduct in some circumstances.

The WAME Chatbot Guidelines add another operational layer. They ask authors, reviewers, and editors to specify chatbot use where it affects scholarly work, manuscript evaluation, review generation, or editorial correspondence. WAME also warns that chatbot prompts and content may be retained by tool providers, creating confidentiality risks if unpublished manuscript text, reviewer reports, or editorial correspondence are entered into public tools. 

Together, ICMJE and WAME move the discussion away from general permission and toward documentation: who used AI, for what purpose, in which part of the work, and with what specific disclosure.

Publisher-Side Screening Makes Verification More Important

Publisher-side screening is now part of the submission environment. Elsevier expanded its Check Integrity tool across nearly 2,000 journals to identify potential ethics issues before publication, including unauthorized authorship changes and editorial conflicts of interest.

Concurrently, Springer Nature has integrated research integrity tools used in submission screening, including an irrelevant reference checker, Geppetto for identifying AI-generated fake content, and SnappShot for identifying problematic images.

For authors, the practical lesson is not to try and write for detection tools, but rather to remove avoidable integrity risks before submission.

• AI-generated references must be checked against original sources to protect against fabricated citations.

• Figure files should be auditable, with source files and processing steps preserved.

• Author contributions must be transparent and accurate.

• Disclosure statements should match the actual drafting and revision process.

The more structured the submission workflow becomes, the less reliable it is to treat AI disclosure as a final administrative sentence.

What Major Publishers Agree On About AI Use

While publisher AI policies for authors differ in disclosure formats and journal-level rules, the baseline is increasingly consistent across major publishers and publication-ethics bodies. Before looking at specific exceptions, research teams should align their submission workflow with these core consensus points:

The Shared Policy Baseline

AI Tools Cannot Be Listed as Authors

The authorship rule is the clearest point of industry-wide agreement. AI tools cannot approve a final manuscript, respond to editorial queries, manage copyright transfers, or accept responsibility if integrity or ethical concerns arise after publication. For these reasons, publishers and editorial bodies reject AI authorship entirely, even when tools contribute heavily to writing, code support, translation, or data processing.

This rule extends directly to citation practice. Generative tools and chatbots cannot be treated as primary or scholarly sources for factual claims. If an AI tool suggests a reference, statistic, methodology, or interpretation, the research team is required to locate, verify, and cite the original source directly.

Authors Remain Accountable for AI-Assisted Content

Using artificial intelligence does not shift accountability away from the research group. Authors remain responsible for the integrity of the submitted manuscript, including false claims, plagiarized text, altered figures, or invalid references introduced through AI-assisted work.

This dynamic becomes particularly complex in collaborative, multi-author environments where different contributors may use AI tools independently. To manage this risk, research teams should use a clear protocol for AI contribution tracking in co-authored papers before submission.

Verification Workflow by Task Type

Substantive AI Use Usually Requires Disclosure

Most policies draw a line between superficial language polish and substantive AI interaction. While basic editing for style, grammar, or punctuation is often exempt, any tool usage that modifies the content, meaning, layout, or interpretation of a study requires formal disclosure.

Classifying Substantive AI Uses

Before submission, separate narrow line editing from substantive AI contributions. While basic grammar polishing may fall safely under an exemption, any substantive contribution must be carefully logged, verified, and matched to your journal's exact disclosure format.

Where AI Disclosure Requirements Differ by Publisher

A primary compliance issue for research teams is that AI disclosure in academic publishing is not fully standardized. The exact same AI-assisted task may need to be reported in entirely different ways depending on the publisher, journal, article type, and submission interface. Authors can run into problems even when they disclose AI use if the disclosure appears in the wrong section or leaves out details the journal requires.

Disclosure in the Manuscript

When AI use affects the submitted work, most publishers require disclosure somewhere within the manuscript body. However, the exact location is determined by the specific function of the tool.

The main distinction across these manuscript locations is functional. AI used to polish author-written prose is treated differently from AI used to generate analytical output, prepare visual material, write code, or shape reported results. Do not assume that a single disclosure sentence placed in the acknowledgements will cover all tool interactions.

Disclosure in the Cover Letter or Submission Portal

AI disclosure frequently must extend outside the manuscript file itself. This is especially true in medical and biomedical publishing under ICMJE Recommendations, which state that journals should require authors to disclose AI use at submission and describe it in both the cover letter and the submitted work when applicable.

Submission platforms also use direct questions to record AI use. These portal inputs form an explicit part of the editorial record. Before starting the upload process, verify that your records are ready for the following fields:

If your manuscript text says one thing and your submission form says another, the inconsistency will delay your review. Align your manuscript disclosure, cover letter, portal responses, and coauthor records before submission.

What to Include in an AI Disclosure Statement

A strong AI disclosure statement should allow an editor, reviewer, or reader to understand what the tool contributed and how the authors verified the final output.

A weak disclosure statement says only that AI was "used during manuscript preparation." A stronger disclosure identifies the tool, the task, the specific manuscript area affected, and the precise verification step taken. The goal is to make substantive AI use completely traceable for editorial review.

Comparative Publisher Policy Matrix for 2026

The tables below summarize the most relevant policy differences for authors preparing a journal submission in 2026, based on publisher and publication-ethics guidance reviewed in June 2026. This publisher AI policy comparison is a starting point, not a substitute for checking the target journal’s current author instructions and submission portal.

To keep the matrix readable and scannable, the comparison is divided into two sections: manuscript disclosure rules and integrity, confidentiality, and legal-risk rules.

Authorship, Disclosure, and Editing Exemptions

Images, Peer Review, Legal Risk, and Journal-Level Cautions

This matrix should be used as an initial orientation tool for your research group. Before submission, authors should still verify four items directly on the target journal’s website: the specific author instructions, the AI policy addendums, the image or figure policy, and the submission portal questions.

AI-Generated Images, Figures, Tables, and Visual Abstracts

In 2026, AI-generated image and figure rules are among the most detailed parts of publisher policy. Textual AI use is often handled through disclosure rules, while visual material is reviewed more cautiously. The stricter treatment reflects several policy concerns that recur across publisher guidance: data integrity, evidence validity, copyright, licensing, and reproducibility.

To safely navigate editorial checks, research teams should evaluate every visual element against this localized risk hierarchy before submitting:

The Visual Output Risk Hierarchy

Primary Research Images and Original Data

Primary research images represent the highest risk bracket in academic compliance. This category includes microscopy images, DNA gels, western blots, histology slides, radiology images, clinical photographs, patient images, and any other visual material that serves as direct empirical evidence for a scientific claim.

For evidence-bearing visuals, visual clarity does not override data integrity. The submitted image must remain a faithful representation of observed or measured data. Generative models function by calculating probabilistic pixel outputs, meaning they can inadvertently introduce artificial artifacts, delete valid biological structures, or smooth away critical data anomalies.

Pre-Submission Figure Audit

Before routing files to the submission portal, principal investigators should evaluate every primary image using this technical protocol:

Charts, Tables, Data Visualizations, and Code

Data charts, numerical tables, statistical plots, mathematical formulas, and software scripts occupy a different policy category from prose or primary research images. While they do not consist of raw pixel data, they fundamentally shape the reporting and interpretation of your results. For this reason, 2026 publisher guidelines treat AI-generated tables or code as core components of the research-output workflow rather than routine manuscript preparation.

Submission Risk Classification by Workflow Task

To reduce submission risk, research teams should distinguish lower-risk technical assistance from higher-risk generation tasks:

• Manuscript Formatting:

  • Lower-Risk/Permitted: Utilizing software to adjust chart labels, grid layouts, axis readability, or cell spacing.

  • Higher-Risk/Substantive: Letting an algorithm dynamically change or smooth out a trendline to alter what the empirical data appears to show.

• Data Visualization:

  • Lower-Risk/Permitted: Rendering a clean scatterplot or heatmap directly from verified, raw source data using a transparent script.

  • Higher-Risk/Substantive: Generating an illustrative plot or visual summary without a verifiable, underlying numeric dataset or reproducible methodology.

• Software Coding & Scripts:

  • Lower-Risk/Permitted: Prompting a model to debug syntax errors, optimize loops, or add comments to author-written data processing scripts.

  • Higher-Risk/Substantive: Copying and pasting complex, unverified AI-generated analysis code directly into your study's operational pipeline without independent testing.

• Data Tables & Matrices:

  • Lower-Risk/Permitted: Reformatting a static, verified block of numbers into the specific column layout required by a journal style guide.

  • Higher-Risk/Substantive: Instructing a model to autonomously calculate averages, generate data classifications, or compile statistical summaries that are not manually checked.

When using AI tools for code, calculations, or visualization, authors should keep a verification trail that includes source data, relevant prompts or task descriptions, code iterations, and validation outputs. If the tool directly affects data processing or results presentation, the workflow may need to be reported in the Methods section.

Unchecked AI outputs can introduce fabricated categories, incorrect calculations, or misleading summaries. Verification should include both technical testing and substantive review.

Explanatory Diagrams, Graphical Abstracts, and Cover Art

Explanatory diagrams, theoretical schematics, and model overviews are regulated under separate parameters because they illustrate a concept, mechanism, or hypothetical workflow rather than present raw, observed data. However, they must still be completely accurate, properly licensed, and explicitly disclosed.

Graphical abstracts and cover art require special caution. While they sit outside the evidentiary text of the paper, publishers enforce strict rules here because these assets are highly visible, heavily indexed, used for public promotion, and present severe copyright liabilities.

Visual Asset Policy Checklist

Research groups must never assume that permission to use tools for text editing translates to visual generation. A target journal may permit AI-assisted language editing while restricting or rejecting undisclosed generative visual material.

Master Pre-Submission Figure Audit Workflow

Run this figure audit before submission, separate from the standard text review, to reduce avoidable visual-policy risks:

• Isolate and Categorize Every Asset: Catalog every visual item into a distinct structural bucket: primary research data, data chart, tabular matrix, conceptual diagram, graphical abstract, or cover art.

• Verify Empirical Status: Explicitly mark whether each asset represents observed data, measured data, algorithmically derived data, or a pure theoretical concept.

• Map Algorithmic Touchpoints: Document whether an AI tool was used to create, edit, label, sharpen, clean, translate, redraw, or format any portion of the graphic.

• Check Pipeline Reproducibility: Confirm that all raw source data, baseline capture images, underlying script code, and exact prompt parameters can be immediately retrieved for editorial audit.

• Audit Local Journal Parameters: Review the specific journal’s image integrity, figure preparation, graphical abstract, and cover-art policy pages to verify no local overrides exist.

• Finalize Disclosure Placement: Determine if the visual disclosure statement belongs inside the Methods section, the data acknowledgements, the master AI declaration block, or directly within the figure caption text.

• Archive Source and Compliance Logs: Save all original, uncompressed image files, raw instrument outputs, data scripts, and multi-author validation logs into a secure project folder.

• Secure Direct Author Confirmation: Require the specific researcher or co-author responsible for figure creation to formally sign off that the asset is completely accurate and policy-compliant.

By treating every visual element as part of the submission record, you make the figure workflow easier to verify if editors request clarification.

Peer Review Boundaries and Confidentiality Rules

Peer review AI confidentiality rules are often significantly stricter than author-facing guidelines. A journal may permit authors to use AI for language support while prohibiting reviewers and editors from entering the manuscript into a public AI system. The distinction is straightforward: peer review inherently involves unvetted, confidential research content.

For authors, understanding these boundaries is critical for two reasons. First, many researchers concurrently serve as journal referees or handling editors. Second, reviewer reports and editorial correspondence may become part of the author’s revision workflow, where authors may be tempted to use AI to organize responses in ways that create confidentiality risk.

Why Public AI Tools Create Confidentiality Risks

Unpublished manuscripts contain highly sensitive intellectual property—including novel findings, proprietary methodologies, patent-sensitive concepts, identifiable clinical data, raw code, and supplementary datasets.

According to the World Association of Medical Editors (WAME), uploading a manuscript to a chatbot can breach confidentiality because user prompts and uploaded content may be retained by the provider.

The STM Association similarly warns that entering copyrighted or unpublished material into public generative AI platforms can create privacy, copyright, and confidentiality risks.

A safe baseline is to avoid pasting confidential review material into public AI systems unless the journal explicitly permits the tool and its confidentiality safeguards are clear.

What Reviewers and Editors May Be Allowed to Do

Reviewer and editor AI rules differ significantly by publisher. Some policies completely prohibit any AI use for manuscript assessment. Others allow extremely narrow uses—such as polishing the grammar of a review report—provided confidentiality is guaranteed and the use is formally disclosed to the handling editor.

The central rule is that scientific judgment remains with the reviewer or editor. For a broader explanation of how referees fit into the publication lifecycle, see thesify’s guide to Mastering the Peer Review Process.

Reviewer-Response Drafting for Authors

Authors face a related compliance issue when preparing their "Response to Reviewers" documents. AI can be highly useful for organizing a response table, improving professional tone, or checking clarity. The risk arises when confidential reviewer comments, editorial correspondence, or unpublished manuscript text are entered into consumer tools with unclear privacy terms.

Before using AI in a reviewer-response workflow, authors should separate low-risk formatting from higher-risk content processing:

If AI contributes materially to a revised manuscript, a new statistical analysis, an added figure, or a rewritten interpretation, authors must log the use and update their AI disclosure statement.

Reviewer Confidentiality: Do's and Don'ts

Publisher Screening Tools and Research Integrity Checks

Research integrity screening tools in academic publishing are now part of the submission environment. These systems are not limited to detecting AI-generated text. They may also evaluate references, image files, authorship changes, conflicts of interest, and broader paper-mill patterns.

For authors, the implication is highly practical: a manuscript must be systematically checked internally before upload, not just before publication. AI disclosure, citation verification, figure integrity, and author contribution records must be fully consistent before the paper enters editorial triage.

What Submission Screening Tools Look For

Suggested Submission-Screening Flow:

Manuscript Upload → Technical Checks → Integrity Screening → Editorial Triage → Peer Review

Reference Checks and Hallucinated Citations

AI-assisted writing creates a highly specific reference problem: citations that look perfectly plausible but do not exist, or citations that exist but do not support the specific claim being made. AI tools can produce inaccurate titles, incorrect author lists, mismatched DOIs, or citations that are contextually irrelevant.

A 2026 analysis reported by Retraction Watch highlighted a sharp increase in fabricated references in the biomedical literature. An audit of nearly 2.5 million papers reported that about one in 277 PubMed-indexed papers published in the first seven weeks of 2026 referenced a nonexistent paper. The practical lesson for authors is: do not treat an AI-generated reference as a valid citation until it has been manually verified.

If AI was used for literature synthesis, citation discovery, or reference formatting, keep a record of the task and the verification step. A reference list is part of the evidentiary structure of the paper, not a formatting afterthought.

Image Integrity Screening

Image integrity screening is also becoming much more visible in publisher workflows. These checks look for duplicated panels, manipulated images, unexpected visual similarities, inappropriate reuse, or inconsistencies between the published figure and the underlying data.

This is why figure documentation must be handled before submission. Authors should preserve original image files, processed versions, analysis scripts, figure assembly files, and caption notes.

The aim is to make the visual record auditable. If an editor asks how a figure was produced, the author team should be able to answer immediately without reconstructing the workflow from memory.

Authorship, Conflict, and Paper-Mill Signals

Publishers are deploying screening systems to assess risks well beyond text and images. For example, Elsevier's 2026 expansion of its Check Integrity tool spans nearly 2,000 journals to screen submissions for potential publishing ethics issues. This includes checking for unauthorized authorship changes and editorial conflicts of interest, with flagged concerns routed directly to specialist integrity analysts.

Similarly, Springer Nature describes research integrity workflows that include an irrelevant reference checker, Geppetto, and SnappShot for identifying reference, text, and image concerns.

For authors, these screening checks reinforce the value of basic submission governance:

These checks are not a reason to write defensively or over-disclose every routine spell-checker. They are a reason to keep submission records accurate and consistent.

What This Means for Authors

Publisher screening tools do not replace author responsibility; they increase the cost of weak documentation, vague disclosure, and unverified outputs. Before clicking submit, corresponding authors should complete five foundational checks:

A clean submission record will not guarantee peer-review acceptance, but it can reduce avoidable integrity questions before the paper reaches the handling editor.

Journal-Specific AI Policy Deviations Authors Should Check

A complete generative AI policy journal submission check should not stop at a publisher’s homepage. While publisher-level frameworks establish the structural baseline, individual journals retain the autonomy to enforce much narrower rules regarding text generation, visual abstracts, primary data figures, and automated portal queries. This is particularly common in medicine, life sciences, image-heavy disciplines, and high-impact venues with strict reporting standards.

Consequently, AI compliance has become a core variable in journal selection. Research teams must assess a venue's specific artificial intelligence criteria alongside traditional factors like scope, indexing, and publication timelines. For step-by-step guidance on aligning your project with the right venue, see thesify's guide to conference submission and journal recommendations.

When Journal Rules Are Stricter Than Publisher Rules

Journal-level deviations exist because editors must protect field-specific reporting standards, clinical data pathways, and visual records. Assuming a general publisher-level permission applies identically to every journal in that portfolio creates submission risk.

Where Local Journal Rules Regularly Diverge from Publisher Standards

The practical rule for authors is: treat your parent publisher's corporate guidelines as the policy floor, never as the definitive submission rule.

Examples to Verify Before Submission

The table below is a verification guide, not a definitive policy summary. Use it to identify the journal-family rules that should be checked directly before submission.

Highly Scrutinized Journal Families and Localized Deviations

Where to Look for Local AI Rules

To ensure your research group is fully compliant with AI policy parameters, audit your target journal's interface across these specific sub-pages and administrative areas before starting the file transfer process:

• Instructions for Authors / Submission Guidelines: Review for broad baseline parameters, permissible text-generation limits, and required declaration phrasing.

• Editorial Policies / Publishing Ethics: Scan the sections on authorship criteria, definition of research misconduct, and plagiarism indicators.

• Research Integrity / Image Integrity Guidelines: Look specifically for structural constraints on digitizing gels, western blots, microscopy captures, and radiology scans.

• Figure Preparation / Graphical Abstract Instructions: Verify the explicit rules regarding the generation or design of visual models, graphical summaries, or journal cover art.

• Peer Review Policy: Check what parameters regulate AI tools on the referee side, ensuring you do not cross boundaries if writing a review or revision response.

• Portal Submission Walkthrough: Preview the online submission screens if possible; look for hard-coded checkboxes, mandatory data blocks, or text upload forms explicitly asking for AI documentation.

Before clicking final submission, compare your manuscript text, formal cover letter, figure captions, internal laboratory logs, and online portal answers. These materials should describe AI use, data handling, and verification consistently.

Global Standards and Emerging AI Disclosure Frameworks

AI disclosure standards in academic publishing are no longer shaped solely by individual publishing houses. Central publication-ethics bodies and international medical associations are active in setting structural expectations for authors, reviewers, and editors. These frameworks help explain why many publishers converge on the same baseline rules: accountability, transparent disclosure, confidentiality, and verification.

Core Global Frameworks at a Glance

ICMJE AI Guidance for Authors, Reviewers, and Editors

The ICMJE Recommendations provide an influential standard for authors, reviewers, and editors in medical publishing. This guidance moves AI disclosure away from optional acknowledgement language and into the submission workflow.

ICMJE Stakeholder Responsibility Matrix

WAME Guidance on AI Disclosure by Function

The WAME Chatbot Guidelines are useful because they separate tool use into functional categories. Rather than treating all AI use as one broad category, WAME links disclosure to the role the tool played in the work.

WAME Function-Based Reporting Logic

• Simple Word-Processing Support: Basic spelling, style, and formatting tools are treated separately from substantive research and do not carry heavy reporting mandates.

• Drafting New Prose Content: If an LLM is used to draft text segments, authors must explicitly report this use within the Acknowledgements, specifying the exact prompts used where relevant.

• Converting Prompts to Visuals: If AI converts text prompts into data tables or illustrations, authors must specify the prompt strings and precisely describe the system's role.

• Analytical Work & Interpretation: If AI is used to evaluate data or perform calculations, this intervention must be disclosed directly in the body of the paper, specifically within the Abstract and Methods section.

• Code Generation: Any automated code generation that impacts data processing, filtering, or downstream experimental results must be declared clearly within the text.

• Generating Tables, Figures, or Reported Results: If AI generates tables, figures, or reported results, authors should record the tool, version, prompt, date or time where relevant, and verification steps.

• Editorial & Review Correspondence: Both editors and peer reviewers must explicitly disclose any chatbot use to authors and to each other within their official communications.

Documenting the tool’s function first makes it easier to match AI use to the journal’s required disclosure location.

STM Guidance and Activity Classifications

The STM Generative AI Paper Framework serves as an industry-wide advisory blueprint that shapes how major publishers separate low-risk formatting from impermissible research activities.

STM Systemic Boundary Classifications

• Basic Text Refinement: Routine text correction, language editing, spelling checks, and bibliography formatting are classified as basic author support and are generally permitted.

• Substantive Functional Applications: Any usage that goes beyond basic editing requires explicit disclosure so that the journal's editorial team can verify legitimacy before review.

• Data Manipulation and Fabrication: STM guidance treats the use of generative AI to create, alter, or manipulate original research data, images, or results as outside acceptable author-support use.

• Public AI Infrastructure Risk: Uploading unreleased research text into public tools introduces severe privacy, copyright, data storage, and training reuse risks.

• Reviewer Data Lockdowns: Peer reviewers are strongly restricted from processing manuscripts through public utilities due to strict confidentiality parameters.

• Bespoke Publisher Tools: Accessing secure, publisher-controlled AI tools is appropriate only when verified data privacy, IP protection, and data security safeguards are active.

STM’s taxonomy reinforces a practical point for authors: basic editing, programming scripts, visual mapping, data analysis, and referee reports should not be collapsed into a single disclosure category. Each interaction must be recorded separately, verified manually, and checked against local rules before submission.

How Research Teams Should Document AI Use Before Submission

Knowing exactly how to disclose AI use in a manuscript is easier when the research team records tool use throughout drafting and revision. In modern multi-author papers, artificial intelligence may be distributed across several disconnected phases: one author might use it for translation, another for debugging R code, and a third for drafting a conceptual figure. If these uses are reconstructed from memory at submission, policy-relevant details can be missed.

A shared AI use log provides the corresponding author with an auditable, practical record to formulate accurate disclosures, manage coauthor review, and answer editorial queries. It also establishes a clear governance structure for AI contribution tracking in co-authored papers, ensuring individual accountability across complex projects.

Create a Shared AI Use Log

Make your log specific enough to capture what the tool did, what material was entered, what output was kept, and how the authors verified the result.

For laboratories that already utilize structured workflows, this AI use log should sit alongside the main manuscript tracker, figure checklist, and author contribution statement. It should be updated during drafting rather than reconstructed at the end.

Map AI Use to Disclosure Requirements

The shared log is most useful  when it is mapped to the target journal’s mandatory disclosure formats. Because the WAME Chatbot Guidelines dictate that AI should be reported based on its specific function, different tasks will belong in entirely different sections of your submission package.

A single manuscript may trigger multiple disclosure locations. For instance, AI-assisted translation might belong in the Acknowledgements, while AI-assisted code generation belongs in the Methods section. A comprehensive log helps the team avoid collapsing distinct uses into one vague disclosure sentence.

Build an AI Sign-Off Step Into Coauthor Review

Before finalizing the submission, the corresponding author should ask each coauthor to confirm any AI use. This supports responsibility and helps identify any AI use that has not yet been reviewed or disclosed.

This sign-off can be handled as a short checklist step alongside author order, conflicts of interest, funding statements, and data availability.

Save Prompt and Output Records When Needed

Prompt and output retention is most critical when generative AI contributes to structural analysis, coding, tables, visual figures, or reported empirical results. Some publishers, including Frontiers and IEEE, place stronger emphasis on prompt, output, or caption-level records in specific AI-assisted workflows, especially where AI contributes to figures, analysis, code, or reported results. Authors should preserve prompt histories, raw outputs, source files, and verification notes so the workflow can be explained if the journal requests clarification.

Even if your target journal does not formally demand prompt logs at submission, preserving these records allows authors to immediately and accurately answer editorial integrity queries during peer review.

For a broader understanding of how this preparation fits into your overall workflow, see thesify’s guide on how to write a scientific paper. The core principle remains the same: a strong submission workflow makes the methodology of the paper visible long before a journal editor ever has to ask for clarification.

The aim of these logs is to preserve enough information to explain substantive AI use, verify the final work, and align the manuscript, cover letter, figures, and online submission forms.

Final Submission Workflow for Policy-Safe AI Use

A final review of AI tools and academic publishing ethics in the 24 to 48 hours before submission can reduce avoidable disclosure, citation, figure, and confidentiality questions. At this stage, the goal is  to confirm that your tool records, disclosure text, figures, references, coauthor approvals, and submission portal inputs are consistent.

Research teams balancing active grants alongside publication workflows must also align journal disclosures with broader research-governance expectations. For related context on funding compliance, see thesify’s guide to Funding Agency AI Policies: What Researchers Need to Know, which highlights how institutional and publisher governance often overlap.

24 to 48 Hour Pre-Submission AI Tracker

Verifying Cross-Document Alignment

A reliable final check requires cross-referencing five critical documents: the manuscript file, the formal cover letter, your internal AI use log, your online submission form answers, and your co-author confirmations. These assets must describe an identical workflow. If your internal log shows an author used an AI platform for translation or data visualization, your final disclosure statement must accurately reflect where that work is reported, or explicitly outline why it qualifies for a line-editing exemption under the target journal’s policy.

Before you click submit, ensure that your AI disclosure statement is technically precise without over-reporting routine spelling checks. A compliant final disclosure should clearly detail the tool name, specific build version, developer, precise purpose, affected section, and the validation steps taken. For figure, code, or statistical visualization pipelines, the record must also preserve raw source files and prompt strings to ensure reproducibility if requested by an editor during triage. This final workflow can reduce avoidable policy delays before the manuscript moves into editorial assessment.

FAQs About AI Policies in Academic Publishing 2026

Can AI Be Listed as an Author on a Journal Article?

No. Major global publishers and core publication-ethics standards strictly reject AI authorship because AI tools cannot approve a final manuscript, accept responsibility for errors, respond to editorial integrity queries, or handle legal agreements. If artificial intelligence contributed to any stage of your project, the authors remain accountable for checking, verifying, and disclosing that use according to the journal’s policy.

Do Grammar and Spelling Tools Need to Be Disclosed?

Often no, but this depends entirely on your target journal's policy. Many international standards treat routine grammar, spelling, punctuation, and narrow copy-editing support as basic assistive editing that does not require disclosure. This exemption generally applies when the tool polishes author-written text without introducing new ideas, altering argument structures, or shifting data interpretations.

However, if AI drafts new prose text, rewrites chapters, translates substantive sections of the manuscript, summarizes literature, or builds visual elements, it crosses into substantive use and must be reported.

Where Should AI Use Be Disclosed in a Manuscript?

The designated location varies significantly based on publisher rules and the tool's functional application.

Overview of Common Disclosure Placements

• Methods Section: Typically required when AI is integrated into data analysis, computational modeling, software coding, or data visualizations.

• Acknowledgements or Dedicated Declarations: Common placements when tools support text translation, preliminary drafting, or structural manuscript formatting.

• Cover Letters or Submission Portals: Some journals require AI disclosure in the cover letter, the submission portal, or both before editorial review begins.

Are AI-Generated Figures Allowed in Academic Publishing?

AI-generated figures are highly restricted or outright prohibited, particularly when they represent primary research data. Many major publishers restrict or prohibit generative AI use for evidence-bearing visuals such as microscopy panels, western blots, gels, histology slides, radiology images, and clinical patient images.

While conceptual diagrams, explanatory flowcharts, data visualizations, graphical abstracts, and cover art may carry different, more flexible rules, they remain heavily governed by journal-specific guidelines and licensing terms.

Can Peer Reviewers Use AI Tools?

Reviewers generally must not upload unpublished manuscripts, figures, tables, supplementary data files, or editorial correspondence into public AI platforms. Peer-review submissions are confidential, and uploading them to public chatbots may breach confidentiality, intellectual property, or data privacy expectations.

Some publishers may permit reviewers to use secure, private, or internal environments to polish the language of drafted feedback reports. The substantive evaluation of novelty, methodology, and scientific merit should remain reviewer- or editor-led.

How Should Coauthors Track AI Use?

Co-authors should collaborate using a centralized, shared AI use log from the start of the drafting process. At a minimum, this team log should document:

• The exact tool name and software build version.

• The date of the query and the specific co-author who ran it.

• The exact task performed and the manuscript section or figure file affected.

• The raw inputs entered alongside the specific outputs retained.

• The manual verification steps taken to validate the data accuracy.

Maintaining this clear internal infrastructure makes it simple for the corresponding author to compile accurate disclosure text, protect data confidentiality, and respond clearly to editorial or integrity questions during submission review.

AI Policy Compliance Is a Research Governance Task

Achieving AI compliance in academic publishing is a research governance task that depends on documented oversight, accurate disclosure, careful handling of figures and data, and verification against the target journal’s rules.

A sound submission workflow keeps AI tool use traceable, verifies outputs that affect the manuscript, protects confidential material from public platforms, and ensures the final disclosure matches the team’s actual research and writing process.

While global publisher guidelines share stable foundational principles, the exact disclosure locations, image categories, peer-review restrictions, and portal-level expectations remain unstandardized. To reduce avoidable delays, research teams should treat the target journal’s instructions as the final authority before submission.

Related Articles

• Tracking AI Provenance in Co-Authored Research Papers: Learn how research teams use thesify Coauthor to map text provenance, attribute machine suggestions, and secure full transparency before submission. Recognize the risks of unmapped structural changes across distributed project workflows and understand how centralized workspaces preserve text integrity.

• How to Co-Author an Academic Paper: Roles & Order: Learn how to co-author academic papers by aligning on authorship order, creating clear agreements, and letting the first author anchor the target journal. Read about how different fields handle author order and why a one‑size‑fits‑all approach often fails. Get guidance on aligning on the norms of the first author’s discipline when selecting a target journal.

• How to Submit a Paper to a Journal: Step-by-Step: Get a step-by-step guide on how to submit manuscript to journal from shortlisting journals based on aims, scope, and article type to tracking editorial screening and preparing for peer review outcomes. We cover checking “Instructions for Authors”, submission requirements, preparing files and entering metadata into submission systems, cover letters and declarations, and more.