SAGE Publishing runs two separate generative-AI policies rather than one: a journal-article policy (governing what authors, editors, and peer reviewers may do with a manuscript under submission or review) and a distinct set of guidelines for book and learning-resource authors. Both share the same core logic — AI tools cannot hold authorship or accountability, substantive AI-generated content must be disclosed, and purely assistive language/grammar tools do not — but the required process and disclosure point differ by publication type. This page covers both, sourced directly from SAGE’s own current policy pages. For how SAGE’s approach compares to other major scholarly publishers, see CASRAI’s wider Journal and Publisher Policies on Generative AI in Manuscripts guide.
SAGE’s journal-article policy: what’s prohibited
SAGE’s journal Artificial Intelligence policy lists a set of uses that disclosure cannot make acceptable — these are prohibited outright, not “permitted with a declaration”:
- Generating incorrect text or content and presenting it as the author’s own analysis or writing.
- Generating data, or an entire submission, from a series of prompts — i.e., using a generative AI tool to produce the substance of a study rather than assisting with an existing one.
- Conducting interviews with generative AI tools in place of human research participants in qualitative research, or having an AI tool perform the analysis of experiences and themes that is properly the researcher’s own interpretive work.
- Plagiarism or inappropriate attribution to prior sources produced via an AI tool — treated as a standard research-integrity violation, not a distinct AI-specific offense.
- AI-generated images presented as unique or novel research images — i.e., passing off a generated image as if it depicted real data or results.
- Fabricated references or falsified claims introduced or missed because an AI tool’s output wasn’t checked.
- Use of generative AI tools for editorial work or to conduct peer review — this restriction applies to editors and reviewers handling a manuscript, not just to authors submitting one (see the peer-review section below).
Consistent with every other major scholarly publisher’s policy, SAGE does not permit a generative AI tool to be listed as an author or co-author of a journal article. The reasoning tracks the same accountability logic used across the industry: authorship requires the ability to take responsibility for a work’s accuracy and integrity, which a tool cannot do. See AI as author and AI co-authorship rejection (ICMJE 2023) for the underlying Dictionary-level treatment of this principle.
What must be disclosed, and where
SAGE draws a two-tier line between AI use that needs no disclosure and AI use that does.
No disclosure required: assistive AI tools
Per SAGE’s own wording, “AI tools that make suggestions to improve or enhance your own work, such as tools to improve language, grammar, or structure, are considered assistive AI tools and do not require disclosure.” This covers ordinary use of tools like spell-checkers, grammar checkers, and built-in suggestion features in standard writing software — the same exemption every major publisher’s policy carves out for basic language-polishing tools.
Disclosure required: substantive generative AI use
Where generative AI contributes actual content rather than just polishing the author’s own draft, SAGE requires disclosure. Uses it names explicitly as requiring disclosure include:
- Literature review assistance
- Translation services
- Software or code generation
- Data visualization
- Representative illustrations or infographics (as distinct from images presented as actual research results, which are prohibited outright — see above)
- Code error-checking
- Reference compilation
SAGE’s stated disclosure location for journal articles is the Methods or Acknowledgements section — the same two locations most other major publishers use, though exact placement conventions still vary journal to journal. See the generative-AI disclosure statement Dictionary entry for what a compliant statement typically needs to name (tool, version, and what it was used for).
Whatever the location, the underlying obligation is unchanged: authors “must make every effort to ensure that the output is factually correct, and the references provided reflect the claims made” — disclosure does not shift responsibility for accuracy away from the author. SAGE’s own guidance is explicit that current language models “are not fully objective or factual,” which is the stated rationale for requiring authors to verify AI output rather than simply disclosing it and moving on.
Book and learning-resource authors: a separate, more detailed process
SAGE’s guidelines for book and learning-resource authors are distinct from the journals policy above, with their own vocabulary and a more prescriptive editorial process.
Two categories: AI-assisted vs. AI-generated
- AI-assisted (no disclosure required): tools that make suggestions, corrections, and improvements to the author’s own writing — SAGE names Microsoft Outlook and Word’s built-in suggestion features as examples.
- AI-generated (disclosure required): content actually produced by a tool such as ChatGPT or an image generator like DALL-E. For AI-generated content, authors must specify, in their disclosure to their editor, what tool was used, how it was used, and the rationale for use.
What’s prohibited for book authors
- Artificially creating or modifying core research data with generative AI.
- Sharing sensitive personal or proprietary information with a generative AI tool.
- Entering confidential peer-review materials into a generative AI system.
The required process
Book and learning-resource authors are expected to disclose AI-generated content to their editor at submission, and to:
- Verify the accuracy and validity of all AI outputs before relying on them.
- Cross-check any AI-suggested citations for correctness.
- Include proper citation for AI-generated material used in the work.
- Review content the tool produced for bias and plagiarism.
- Acknowledge the limitations of the large language model used.
- Retain full authorship responsibility for the finished work, regardless of what a tool contributed.
SAGE frames disclosure at this stage as informing an editorial decision, not just satisfying a formality: the point of naming the tool and describing how and why it was used is to let the editor assess the submission with that information in hand, rather than discovering undisclosed AI involvement later.
AI and peer review at SAGE
SAGE’s restrictions extend to how editors and peer reviewers may use generative AI when handling a manuscript, which is a distinct question from what an author may disclose:
- Reviewers may not use tools like ChatGPT to generate a peer review. SAGE states that reviewers who do so will not be invited to review again.
- Editors must not use generative AI tools to generate decision letters or summaries of unpublished research.
- Undisclosed generative AI use, once identified, triggers investigation under COPE’s guidelines rather than being handled as a purely internal editorial matter.
This mirrors the confidentiality logic behind reviewer-side AI restrictions industry-wide: an unpublished manuscript entered into a third-party AI tool can expose unpublished data, methods, or findings to that tool’s provider. See the AI in peer review Dictionary entry and CASRAI’s publisher-by-publisher comparison of reviewer-side AI policies for how this compares across other publishers.
Detection and non-disclosure
SAGE’s guidance to its own editors flags specific failure patterns to watch for, including submissions containing images “reported as primary data but were generated using AI” and instances where a language model was allowed to “falsify references or insert incorrect ones.” Consistent with COPE’s guidelines, undisclosed AI use identified after submission is treated as a genuine integrity concern rather than a minor oversight, and can trigger investigation, corrective action, or rejection at any stage of review or publication — more serious outcomes are more likely where the undisclosed use involves plagiarism, fabricated data or references, or misleading attribution of the work’s authorship.
How SAGE compares to the wider publisher landscape
SAGE’s policy follows the same four-part shape found across most major scholarly publishers: no AI authorship, disclosure of substantive use, an exemption for basic language tools, and retained human accountability for accuracy. Two things about SAGE’s approach are worth flagging as more specific than the industry-average version of that shape:
- SAGE names a genuinely separate, more prescriptive process for book and learning-resource authors, on top of its journal-article policy — most publisher AI policies referenced elsewhere are written primarily with journal articles in mind and don’t address monographs or textbooks as their own category.
- SAGE’s prohibited-use list is unusually specific about qualitative research methodology — explicitly naming AI-conducted interviews in place of human participants, and AI-performed thematic analysis, as prohibited rather than merely undisclosed. This reflects SAGE’s particular strength in social-science and qualitative-methods publishing, where those specific misuse patterns are a live concern in a way they aren’t for a primarily biomedical or physical-science publisher.
For the fuller cross-publisher picture — including ICMJE, Elsevier, PLOS, Springer Nature, Wiley, Taylor & Francis, and Science/AAAS — see CASRAI’s Journal and Publisher Policies on Generative AI in Manuscripts guide. As with any publisher policy in this fast-moving area, treat the specifics above as current as of when they were checked, and confirm against SAGE’s own current policy pages before submission — publisher AI policies have already been revised multiple times since 2023.
Frequently asked
Does SAGE allow a generative AI tool to be listed as an author?
No. Consistent with every other major scholarly publisher, SAGE does not permit an AI tool to be listed as an author or co-author, because authorship requires the ability to take accountability for the work, which a tool cannot do.
Do I need to disclose using Grammarly, spell-check, or Word’s built-in suggestions for a SAGE submission?
No. SAGE classifies tools that “make suggestions to improve or enhance your own work, such as tools to improve language, grammar, or structure” as assistive AI, which does not require disclosure. Disclosure is required once a tool generates substantive content rather than polishing the author’s own writing.
Where do I disclose generative AI use in a SAGE journal manuscript?
In the Methods or Acknowledgements section, per SAGE’s journal AI policy. Book and learning-resource authors instead disclose directly to their editor as part of the submission process, following a more detailed process specific to that publication type.
Can SAGE peer reviewers use ChatGPT to help write their review?
No. SAGE states that reviewers who use tools like ChatGPT to generate a review will not be invited to review again, and editors are separately barred from using generative AI to draft decision letters or summarize unpublished research.
What happens if I don’t disclose AI use and SAGE finds out later?
It’s treated as a genuine integrity concern, handled per COPE’s guidelines, and can lead to investigation, corrective action, or rejection at any stage of review or publication — with more serious consequences where the undisclosed use also involves plagiarism, fabricated data, or falsified references.
Is SAGE’s AI policy the same for books as it is for journal articles?
No. SAGE runs a separate set of generative-AI guidelines for book and learning-resource authors, requiring disclosure to the editorial contact at submission (naming the tool, how it was used, and the rationale), plus explicit steps to verify AI output, cross-check citations, and disclose limitations — a more prescriptive process than the journal-article policy’s Methods/Acknowledgements disclosure.
