FAQ

Do I Need to Disclose AI Usage in My Paper?

A practical disclosure test for researchers who used ChatGPT, Copilot, Grammarly AI, Claude, Gemini, or other AI tools while preparing a paper.

Yes, if AI shaped the paper

If an AI tool shaped your manuscript, analysis, code, figures, literature work, or research decisions, disclose it.

That rule will keep most authors out of trouble.

Do not treat disclosure as an admission of misconduct. Treat it like a methods note. You tell readers what you did, where the tool entered the workflow, and how you checked the output.

Major publishers now ask authors to report many forms of generative AI use. ACM tells authors to disclose generative AI when it generates new content such as text, images, tables, or code. Elsevier asks authors to include an AI declaration for manuscript preparation and to describe AI use in the research process in the Methods section. PLOS asks authors to report AI contributions to a study or article content in a dedicated Methods section, or in Acknowledgments for article types without Methods. Taylor & Francis asks authors to state the tool name, version, how they used it, and why. (acm.org)

So if you ask, "Do I need to mention ChatGPT, Claude, Gemini, Copilot, Cursor, Grammarly AI, Elicit, Perplexity, or an AI translation tool?", use this answer first: probably yes.

For venue rules, start with [[[AI disclosure](/ai-usage-cards-vs-datasheets/)](/how-to-disclose-ai-use-for-neurips-icml-and-acl-submissions/) policies by major journals](/ai-disclosure-policies-by-journal/). For the basic concept, read What Are AI Usage Cards?.

The quick test

Ask one question.

Did the AI tool contribute content, analysis, code, interpretation, screening, translation, images, or wording that affected the submitted work?

If yes, disclose it.

You do not need a disclosure for every piece of ordinary software. You do need one when a model generated or transformed scholarly content in a way that readers, reviewers, or editors would want to know.

A reference manager does not write your argument. A basic spellchecker does not invent a citation. A generative model can do both. That difference matters.

When disclosure usually belongs in the paper

Disclose AI use when you used it to draft, rewrite, paraphrase, translate, summarize, classify, screen, code, analyze, visualize, or generate material that appears in the paper or helped produce the paper.

That includes common cases:

  • drafting text that appears in the manuscript
  • rewriting paragraphs, claims, abstracts, titles, or responses to reviewers
  • summarizing papers for a literature review
  • helping screen studies for a review
  • writing or revising code that produced results
  • suggesting statistical tests, models, or data cleaning steps
  • generating figures, diagrams, tables, graphical abstracts, or captions
  • translating manuscript text
  • producing synthetic text, images, or data
  • using AI agents to run searches, compare sources, or prepare files

If you used ChatGPT for prose, see How to Disclose ChatGPT Usage in Academic Papers. If you used Copilot for code, see How to disclose Microsoft Copilot use in academic writing.

When you may not need a disclosure

Some uses sit below the disclosure threshold for many journals.

You usually do not need a special AI statement for ordinary spelling, punctuation, reference management, formatting, or citation style checks. ACM says authors do not need to disclose generative AI use when they use it like Grammarly or a basic word processor to improve spelling, grammar, punctuation, clarity, or similar text quality. Elsevier says basic grammar, spelling, and punctuation checks need no declaration. Sage says authors do not need to disclose AI-assisted tools that suggest corrections and improvements to content that authors wrote themselves, such as spelling and grammar checks. (acm.org)

But watch the switch.

Many tools now move from correction into generation. "Fix typo" differs from "rewrite this introduction for a Nature-style audience." "Check grammar" differs from "produce a new paragraph explaining this result."

Once the tool generates new content or changes the meaning, disclose it.

The gray area: language editing

Language editing causes the most confusion.

If you used a tool only to catch spelling, punctuation, or minor grammar errors, many policies will not ask for a statement. If you accepted rewritten sentences, paraphrased paragraphs, translated text, or tone changes that altered how you expressed claims, disclose it.

Do not write a grand confession. Write a plain sentence.

\section*{AI use disclosure}
The authors used Grammarly and ChatGPT to identify grammar issues and suggest sentence-level revisions in the Introduction and Discussion. The authors reviewed every suggestion and accepted only changes that preserved the intended meaning.

If your paper goes to a journal with a strict template, adapt the sentence. If you do not know the journal yet, keep the note in your project folder and generate a fuller [AI Usage Card](/) before submission.

Literature searches and review work

AI literature tools can affect a paper even when none of their text appears in the final manuscript.

If you used an AI tool to find papers, cluster themes, screen abstracts, summarize studies, compare methods, or draft a review matrix, disclose the role. This applies most strongly to systematic reviews, meta-analyses, scoping reviews, and evidence maps.

A reader needs to know if an AI system helped decide which records entered the review or how you interpreted prior work. That choice affects reproducibility.

For review papers, see AI Disclosure in Systematic Reviews and Meta-Analyses. A short note can work:

\section*{AI use disclosure}
The authors used Elicit and ChatGPT to support literature mapping during the scoping stage. The authors checked all candidate records against database search results and did not use AI-generated summaries as sources. Final inclusion decisions were made by the authors.

Notice the last sentence. It tells the editor where human judgment sat.

Code, statistics, and analysis

If AI helped write code that cleaned data, ran models, produced figures, or generated results, disclose it.

This belongs near the methods, not hidden in a generic acknowledgment. Elsevier tells authors to describe AI use in the research process in detail in the Methods section. PLOS asks authors to state which aspects of the study, article contents, data, or supporting files the AI tool affected or generated. (elsevier.com)

Use clear language:

\section*{AI use disclosure}
The authors used GitHub Copilot to suggest Python code for data cleaning and figure formatting. The authors inspected, tested, and revised the code before running the final analysis. The repository contains the final reviewed code used to produce the reported results.

If you used AI to choose a statistical model, say that. If you only used it to format a plot label, say that instead. Scale the disclosure to the use.

For discipline examples, see AI Disclosure for NLP Research Papers and AI Disclosure for Social Science Research.

Images and figures need extra care

Do not assume disclosure makes every image use acceptable.

Some publishers restrict AI-generated images and AI-assisted image manipulation. Nature says Springer Nature journals cannot permit AI-generated images and videos for publication while legal and integrity issues remain unresolved, with limited exceptions. Elsevier does not permit generative AI or AI-assisted tools to create or alter images in submitted manuscripts, except when the tool forms part of the research design or methods and authors describe that use in a reproducible way. Taylor & Francis permits some AI-assisted data visualizations and conceptual illustrations, but it does not allow generative AI to create or manipulate outputs of research or clinical testing. (nature.com)

This is the one place where "I disclosed it" may not solve the problem.

Before you submit an AI-generated graphical abstract, conceptual figure, microscopy image adjustment, diagnostic image, or research output visualization, check the target journal. If the figure records evidence, the rules will often run tighter than the rules for prose.

Where to put the disclosure

Put the disclosure where the reader expects to learn about that part of the work.

For manuscript preparation, many journals accept a statement near the end of the manuscript, in Acknowledgments, or in a dedicated AI disclosure section. Elsevier asks for a separate AI declaration statement for manuscript preparation and says AI use in the research process belongs in Methods. PLOS asks for a dedicated section of Methods, or Acknowledgments if the article type lacks Methods. Taylor & Francis asks authors to include a specific AI usage disclosure statement in article submissions. ICMJE says journals should require authors to disclose AI use at submission and that authors should describe the use in both the cover letter and the submitted work when applicable. (elsevier.com)

Use this placement guide:

AI useBest place to disclose
language editing or rewritingAI disclosure section or Acknowledgments
code that produced resultsMethods and, if needed, software availability
literature screening or extractionMethods
figure generation or image alterationFigure caption, Methods, and AI disclosure section
cover letter only requestCover letter plus manuscript statement if the journal asks

If your journal has its own template, follow that template first. For submission workflows, see AI Transparency Requirements for Journal Submissions.

What the statement should include

A useful disclosure answers four questions:

  • What tool did you use?
  • What did the tool do?
  • Where did it affect the workflow?
  • How did the authors check the output?

PLOS asks for the tool name, a description of how authors used the tool, how they evaluated the validity of the outputs, and which parts of the study, article, data, or supporting files the tool affected. Taylor & Francis asks for the full tool name with version number, how authors used it, and the reason for use. (journals.plos.org)

That gives you a practical template:

\section*{AI use disclosure}
The authors used [tool name and version] to [specific task] during [stage of the project].
The tool affected [sections, code, figures, data processing, literature screening, or other materials].
The authors reviewed [all outputs / selected outputs], verified [claims, citations, code, calculations, or classifications], and made the final scientific decisions.

A full AI Usage Card gives you more structure than one paragraph. You can attach it as a supplement, place it in an appendix, or copy its text into the journal's disclosure field. For examples, see AI Usage Cards Examples and Templates and the LaTeX Tutorial for AI Usage Cards.

What not to write

Vague disclosures create new problems.

Avoid this:

The authors used AI to improve the manuscript.

That sentence tells the editor almost nothing. It hides the tool, the task, the affected sections, and the checks.

Write this instead:

\section*{AI use disclosure}
The authors used ChatGPT (OpenAI, GPT-4.1) to suggest sentence-level edits in the Introduction and Discussion. The authors reviewed all suggestions, verified that the edits did not change the scientific meaning, and checked all citations independently. No AI tool was used for data analysis, figure generation, or interpretation of results.

If the model or version changed across the project, write what you know. If you cannot reconstruct every model version, do not invent details. Say the tool name, the period of use, the task, and the checks.

What happens if you do not disclose

A missing disclosure can turn a small workflow choice into an integrity problem.

ICMJE says nondisclosure of AI use may require corrective action and may count as misconduct in some circumstances. PLOS says noncompliance with its AI policy counts as misrepresentation of methods, contributions, or results, and the journal may reject the paper, retract it after publication, publish an editorial notice, or notify the authors' institutions. (icmje.org)

Editors dislike surprises. If they find unreported AI use after submission, they must ask a harder question: what else did the authors leave out?

You can avoid that problem with a short, accurate statement.

Keep a record before submission week

Most authors struggle with disclosure because they wait until the last day.

Keep a small log while you write:

Date or stageToolTaskOutput keptAuthor check
March 2026, draftingChatGPTsentence-level edits to Introduction8 edited sentencesauthors checked meaning and citations
April 2026, analysisCopilotPython plotting suggestionsrevised plotting functiontests run and code reviewed
April 2026, reviewElicitcandidate paper searchsearch notes onlyrecords checked in database

That table can become an AI Usage Card, a Methods note, or a disclosure paragraph.

If you write a thesis, see How to Disclose AI Usage in Your Thesis. If you review manuscripts, see AI Disclosure in Peer Review: What Reviewers and Editors Should Report.

The rule to remember

If AI shaped the paper, disclose it.

Name the tool. Say what it did. Say where it affected the work. Say how you checked it.

That is enough for most cases.

If you want to handle this now, generate a free AI Usage Card. You can use the card as a supplement, turn it into a short journal statement, or copy the relevant text into your Acknowledgments, Methods section, cover letter, or submission form.

Generate Your AI Usage Report

Create a standardized AI Usage Card for your research paper in minutes. Free and open source.

Create Your AI Usage Card