How to Disclose ChatGPT Usage in Academic Papers

Why ChatGPT Disclosure Specifically

ChatGPT is by far the most widely used AI tool in academic research. It is also the tool that journal editors and reviewers are most likely to ask about. Whether you used it to brainstorm your introduction, polish a paragraph, or generate Python code for your analysis, your target journal almost certainly wants to know.

The good news is that disclosing ChatGPT usage is simple once you know what to include. This page gives you ready-to-use templates for three levels of usage, along with practical advice on what details matter most.

If you are unsure whether your situation even requires disclosure, check our guide on whether you need to disclose AI usage.

Always Record the Version

Before we get to the templates, one critical detail. Always note which version of ChatGPT you used. GPT-3.5, GPT-4, GPT-4o, and GPT-4.5 have significantly different capabilities and known limitations. A reviewer reading your disclosure needs to know the version to assess what the tool was capable of during your work.

If you used ChatGPT through the web interface, your account settings or conversation history will indicate which model was active. If you accessed it through the API, your code will specify the model string (for example, "gpt-4o-2024-08-06"). Record this information as you work so you do not have to reconstruct it later.

Also note the approximate dates of usage. Models change over time even within the same version label, so the timeframe helps readers and reviewers contextualize your work.

Level 1. Light Usage (Brainstorming and Outlining)

This covers situations where you used ChatGPT as a thinking partner. You asked it to suggest research angles, help you outline a section, generate keyword ideas for a literature search, or explain a concept you were unfamiliar with. The actual text in your paper is your own.

When to Use This Level

You did not copy or closely paraphrase any ChatGPT output into your paper. The tool influenced your thinking but not your text. Most journals still want this disclosed because it affected your research process.

Template

The author(s) used ChatGPT (GPT-4o, OpenAI, accessed [month/year]) during the early stages of this research for brainstorming research questions and outlining the structure of [specific sections]. No AI-generated text was included in the manuscript. All writing, analysis, and interpretation are the authors' own work.

Where to Place It

A brief mention in the Acknowledgments section is sufficient for most journals at this level.

Level 2. Moderate Usage (Writing Assistance, Editing, Paraphrasing)

This covers the most common scenario. You used ChatGPT to help draft, edit, or improve passages of text in your paper. Maybe you wrote a rough version of your related work section and asked ChatGPT to improve the flow. Maybe you used it to paraphrase dense technical content for a broader audience. Maybe English is not your first language and you used ChatGPT to polish your prose.

When to Use This Level

ChatGPT contributed to the language or structure of text that appears in the paper, but you reviewed, edited, and verified everything. The ideas and arguments are yours.

Template

The author(s) used ChatGPT ([model version, e.g., GPT-4], OpenAI, accessed [month range/year]) to assist with writing and editing tasks during the preparation of this manuscript. Specifically, ChatGPT was used to [describe tasks, e.g., "improve the clarity and readability of the Introduction and Discussion sections" or "suggest alternative phrasings for technical descriptions in Section 3"]. All AI-generated suggestions were critically reviewed, substantially revised where necessary, and verified for accuracy by the authors. The authors take full responsibility for the content of this publication.

Where to Place It

Most publishers want this in the Acknowledgments section. Some, like ACM, now request a dedicated AI disclosure section. Check your target journal's specific policy on our journal policy reference page.

Level 3. Heavy Usage (Text Generation, Code, or Analysis)

This applies when ChatGPT played a significant role in producing content that appears in your paper. Examples include generating first drafts of entire sections, writing substantial portions of your analysis code, producing content for tables, or creating text that you then edited but that retains the structure and much of the language of the original output.

When to Use This Level

The paper contains content that originated from ChatGPT, even if you revised it afterward. This is the level where a full AI Usage Card becomes particularly valuable because a simple paragraph cannot adequately capture the details.

Template

The author(s) used ChatGPT ([model version], OpenAI, accessed [month range/year]) as a significant writing and research aid in the preparation of this manuscript. The following describes the specific usage:

Text generation. ChatGPT was used to generate initial drafts of [specify sections, e.g., "the Related Work section (Section 2) and parts of the Methodology section (Section 3)"]. These drafts were substantially revised, restructured, and verified against primary sources by the authors.

Code generation. ChatGPT was used to generate [language, e.g., "Python"] scripts for [purpose, e.g., "data preprocessing and visualization in Section 4"]. All generated code was reviewed, tested, and modified by the authors.

[Other usage]. [Describe any additional usage, such as help with statistical analysis, literature summarization, or figure descriptions.]

All AI-assisted content was verified for accuracy and correctness. The authors take full responsibility for the integrity of this publication. A complete AI Usage Card detailing all AI tool usage is available as supplementary material.

Where to Place It

For this level of usage, place a summary statement in the Methods or Acknowledgments section and attach a full AI Usage Card as supplementary material. The AI Usage Card gives reviewers the structured detail they need without cluttering your main text.

Generating a Formatted Disclosure at ai-cards.org

Instead of writing your disclosure from scratch, you can generate one at ai-cards.org. The generator includes ChatGPT as a pre-selectable tool, and it walks you through each category of usage with guided prompts. You specify the model version, describe what you used it for, explain your verification process, and the tool produces a formatted AI Usage Card you can download as a PDF.

This is particularly useful if you used ChatGPT alongside other AI tools. The generator handles multiple tools in a single card, so you get one unified disclosure document instead of cobbling together separate statements.

Common Mistakes to Avoid

Being vague about the version. "ChatGPT was used" tells a reviewer almost nothing. Always specify GPT-3.5, GPT-4, GPT-4o, or GPT-4.5. These are different models with different capabilities and different known issues.

Underreporting usage. If you used ChatGPT for moderate or heavy assistance, do not describe it as "minor editing help." Reviewers and editors are becoming skilled at identifying AI-assisted text, and a mismatch between your disclosure and the apparent level of AI involvement will raise concerns.

Forgetting the verification step. Every publisher wants to know that you checked the AI output. If your disclosure says "ChatGPT generated the related work section" and nothing more, editors will worry about the reliability of that content. Always explain how you verified the output.

Not disclosing at all. This is the biggest mistake. As we explain in our guide on whether you need to disclose AI usage, failing to disclose can lead to rejection, retraction, and reputational harm. There is no downside to honest disclosure and significant risk in omitting it.

ChatGPT and Authorship

One more important point. Do not list ChatGPT as a co-author on your paper. Every major journal and conference has explicitly stated that AI tools cannot be authors because they cannot take responsibility for the work. Our page on whether AI can be a co-author covers this topic in detail.

The right approach is to credit ChatGPT's contribution through a disclosure statement or AI Usage Card, not through the author list.

Get Started

Creating a proper ChatGPT disclosure takes only a few minutes and protects your work from challenges during review or after publication. Generate your free AI Usage Card now with ChatGPT pre-selected as a tool, or use one of the templates above and adapt it to your situation.