MDPI (Multidisciplinary Digital Publishing Institute) does not allow a generative-AI tool such as ChatGPT to be listed as an author on any manuscript, and requires authors to disclose substantive use of such tools in a manuscript’s Acknowledgments and Materials and Methods sections. MDPI first issued this position in April 2023, built on the general guidance the publisher cites from the Committee on Publication Ethics (COPE) and the International Association of Scientific, Technical and Medical Publishers (STM). This page walks through what MDPI’s policy actually requires, what is exempt from disclosure, and how it compares to the AI-authorship rules at other major scholarly publishers. For the wider landscape across publishers, see CASRAI’s Journal and Publisher Policies on Generative AI in Manuscripts guide; for background on MDPI itself, see the MDPI dictionary entry and CASRAI’s guide on whether MDPI journals are peer reviewed.
The core rule: generative AI cannot be an author
MDPI’s generative AI and large language model (LLM) policy states plainly that tools like ChatGPT “do not meet the criteria for authorship and cannot be listed as authors on manuscripts.” This is the same accountability-based reasoning used across scholarly publishing, and it tracks directly onto the ICMJE authorship criteria that most journals, including MDPI’s, already apply: an author must be able to take public responsibility for a work’s content and agree to be accountable for its accuracy. A software tool cannot assume legal responsibility, cannot hold a conflict of interest, cannot sign a copyright transfer, and cannot be contacted to answer questions about the work post-publication — so it cannot satisfy the authorship test, no matter how much of the drafting it performed. See CASRAI’s entry on AI co-authorship rejection (ICMJE 2023) for the parallel reasoning at ICMJE, which MDPI’s policy is consistent with.
What must be disclosed, and where
Where an author uses a generative-AI tool to generate content for a manuscript — drafting text, generating or reworking substantive passages, or otherwise contributing material beyond simple correction — MDPI requires two things:
- A disclosure statement in the Acknowledgments section, naming the tool used.
- A description in the Materials and Methods section (or the equivalent section for article types without a formal Methods section) explaining specifically how the tool was used — which parts of the work it contributed to and in what way.
This two-part structure mirrors the general disclosure model COPE set out in its February 2023 position statement, “Authorship and AI tools”: name the tool, and explain its role, rather than a bare one-line acknowledgment that AI was “used.” A vague statement such as “ChatGPT was used to improve this manuscript” without further detail does not meet MDPI’s stated standard of “a detailed explanation of how the tool was used.”
What does not require disclosure
MDPI draws a line between generative content creation and routine language polishing. Using an AI-assisted tool purely for basic language editing — correcting grammar, spelling, punctuation, or formatting, without generating new substantive content — does not trigger a disclosure requirement. This mirrors the distinction most other major publishers draw (see SAGE’s policy and Taylor & Francis’s policy): a spell-checker or grammar tool is treated as no different from a human copy-editor, while a tool that drafts sentences, paragraphs, or arguments is treated as a substantive contribution that must be named and explained. Authors who are uncertain which side of that line a given use falls on should default to disclosure — MDPI’s ethics guidance frames the requirement in terms of transparency, and over-disclosure carries no penalty, while under-disclosure is treated as a breach of the publisher’s research and publication ethics standards.
Authors remain fully accountable for AI-assisted content
MDPI places full responsibility for a manuscript’s content on its human authors, regardless of how much of it was AI-assisted. That includes the standard research-integrity risks generative-AI use can introduce or amplify: fabricated or inaccurate citations, plagiarized or insufficiently paraphrased passages lifted from a model’s training data, factual errors presented with unwarranted confidence, and violations of third parties’ intellectual property in AI-generated text or images. An author cannot point to an AI tool as the source of an error to avoid responsibility for it — the accountability sits with the named human authors, exactly as it would for an error introduced by a human co-author or research assistant. This is consistent with acknowledgement-vs-authorship treatment of AI tools across the industry: crediting a tool’s contribution does not transfer any of the accountability that comes with authorship, because the tool was never eligible to hold that accountability in the first place.
Where this policy comes from
MDPI states that its generative-AI guidance is based on the general guidance provided by STM and COPE. Two source documents are worth knowing for context:
- COPE’s “Authorship and AI tools” position statement (published February 13, 2023) — the widely-adopted reference point across scholarly publishing: AI tools cannot be listed as authors because they cannot take responsibility for the work, cannot assert conflicts of interest, and cannot manage copyright or license agreements. It requires disclosure of AI use in writing, in producing images or graphical elements, and in collecting or analyzing data.
- STM’s position on generative AI, developed alongside COPE’s, which most large commercial and society publishers — including MDPI — have aligned their own author-facing policies to.
MDPI’s own announcement of its updated authorship policy dates to April 2023, shortly after COPE’s statement, and the publisher has not published a materially different version of the core rule since. Compare this to CASRAI’s coverage of ICMJE’s generative AI policy and IEEE’s generative AI policy, both of which reach the same “not an author, must be disclosed” conclusion via their own separate statements issued around the same period.
How MDPI compares to other major publishers
On the two rules that matter most in practice — no AI authorship, and disclosure of substantive AI-generated content — MDPI’s policy is functionally aligned with the rest of the field:
- SAGE runs separate journal and book/learning-resource policies, both built on the same “not an author, disclose substantive use” logic, and additionally lists specific prohibited uses (e.g., AI-conducted qualitative interviews, AI-generated images presented as real research images). See SAGE’s generative-AI and authorship policy.
- Taylor & Francis applies the same no-authorship and disclosure rules across its journal portfolio. See Taylor & Francis’s generative AI policy.
- AMA Manual of Style / JAMA Network takes the same authorship position and sets out its own style-specific disclosure conventions for biomedical manuscripts. See AMA/JAMA’s position on AI and authorship.
- ICMJE, whose criteria MDPI and most biomedical and life-science journals apply to determine authorship in the first place, reaches the identical conclusion on AI: it cannot meet the authorship criteria and its use must be disclosed.
Where MDPI’s public author-facing guidance is comparatively thinner than some peers is in prohibited-use specifics and reviewer/editor-facing rules: MDPI’s ethics pages, as published, are framed primarily around author obligations (authorship eligibility, disclosure, accountability) rather than an itemized list of prohibited AI uses or a detailed AI-in-peer-review policy of the kind SAGE publishes separately. Authors and editors handling manuscript confidentiality during peer review should not assume a specific MDPI rule exists beyond the general research-and-publication-ethics obligations that already govern reviewer conduct, and should check the current MDPI ethics page for the journal in question if that specific question is load-bearing for a submission.
Practical checklist for authors submitting to an MDPI journal
- Do not list any AI tool, including ChatGPT or any other LLM, as an author or co-author.
- If a generative-AI tool contributed substantive content (drafted text, generated a figure, assisted with data analysis or interpretation), disclose the tool by name in the Acknowledgments section.
- In the Materials and Methods section (or equivalent), describe specifically how the tool was used — not just that it was used.
- Basic language editing (grammar, spelling, formatting) does not require disclosure, but when in doubt, disclose anyway.
- Independently verify every citation, fact, and figure an AI tool touched — responsibility for errors sits with the human authors, not the tool.
- Check the specific journal’s current instructions for authors, since MDPI operates hundreds of journals and disclosure mechanics can be refined at the individual-journal level even when the core policy is set centrally.
Frequently asked questions
Can ChatGPT be listed as a co-author on an MDPI journal article?
No. MDPI’s policy explicitly states that generative-AI tools do not meet the criteria for authorship and cannot be listed as authors, regardless of how much of the manuscript they contributed to.
Do I have to disclose AI use if I only used it to fix grammar?
No. MDPI’s policy exempts basic language-editing tasks — correcting grammar, punctuation, spelling, and formatting — from the disclosure requirement, treating it the same as conventional copy-editing.
Where in an MDPI manuscript does the AI disclosure go?
MDPI asks for two things: a disclosure statement in the Acknowledgments section naming the tool, and a more detailed explanation of how it was used in the Materials and Methods section (or the equivalent section for the article type).
Is MDPI’s AI policy different from ICMJE’s or COPE’s?
Not on the core rule. MDPI states its policy is based on general guidance from STM and COPE, and reaches the same conclusion as ICMJE, IEEE, and every other major scholarly publisher CASRAI has reviewed: AI cannot be an author, and substantive use must be disclosed.
What happens if I don’t disclose AI use in an MDPI submission?
MDPI treats undisclosed generative-AI use as a research and publication ethics issue, evaluated under the same framework as other undisclosed-methods or transparency failures. Authors remain fully accountable for all manuscript content, disclosed or not.
