Do I Need to Disclose AI Usage in My Paper?

The Short Answer

Yes, you almost certainly need to disclose AI usage in your paper. Every major scientific publisher now requires some form of AI disclosure, and the list of venues without such a policy is shrinking fast. If you are submitting to Nature, Science, IEEE, ACM, Elsevier, Springer, Wiley, Taylor & Francis, or PLOS, you are required to declare how you used AI tools. The same is true for top conferences in computer science, medicine, and the social sciences.

If you are unsure whether your specific journal requires it, check our AI disclosure policies by journal page for a quick reference.

Why Disclosure Is Now the Default

The reason is simple. AI tools like ChatGPT, Claude, Copilot, and others can generate text, code, analysis, and figures that end up in published research. Readers and reviewers need to know which parts of a paper were AI-assisted so they can evaluate the work properly. A literature review drafted by GPT-4 and then checked by the authors is a different product than one written from scratch by domain experts. Neither is inherently wrong, but the reader deserves to know the difference.

Publishers also have a practical concern. If AI-generated content contains errors, hallucinations, or fabricated references, someone must be accountable. Disclosure makes it clear that the human authors reviewed and take responsibility for all content, including anything an AI helped produce.

When Is Disclosure Required? A Quick Decision Guide

Here is a practical way to think about it. If you answer "yes" to any of the following, you should disclose.

1. Did you use an AI tool to generate, draft, or edit any text that appears in the paper? This includes ChatGPT, Claude, Gemini, Copilot, Grammarly's AI features, QuillBot, and similar tools. Even if you rewrote the output substantially, the initial AI involvement should be noted.

2. Did you use an AI tool for data analysis, coding, or statistical work? If an AI wrote or helped write your analysis scripts, generated visualizations, or suggested statistical approaches, that counts.

3. Did you use an AI tool to create or modify any figures, images, or tables? AI-generated or AI-modified visual content needs disclosure at every major venue.

4. Did you use an AI tool for literature search, summarization, or hypothesis generation? Tools like Semantic Scholar's AI features, Elicit, or ChatGPT used for brainstorming fall into this category.

5. Did you use AI-powered translation tools to prepare your manuscript? DeepL, Google Translate, and similar tools with AI backends are covered by most policies.

If you answered "no" to all of the above, you probably do not need a disclosure. Standard tools like reference managers, basic spell-checkers, and traditional grammar tools (without AI-powered suggestion features) are generally not covered by AI disclosure policies.

The Gray Areas

Some situations genuinely fall into a gray zone.

Grammarly and similar writing assistants. Older versions of Grammarly used rule-based grammar checking, which no one would call AI disclosure territory. Current versions use generative AI to suggest rewrites and paraphrases. If you used Grammarly's AI-powered features beyond basic spell-check, it is safer to disclose.

Search engines with AI features. Google and Bing now include AI-generated summaries in search results. If you read those summaries as part of your literature review, most publishers would not consider that AI usage requiring disclosure. But if you used an AI tool specifically to search, filter, or summarize academic literature, that is worth noting.

Code completion in IDEs. If you write code and your IDE offers AI-powered suggestions (GitHub Copilot, Cursor, or similar), this falls under AI usage. If you used code completion to write scripts that produced results in your paper, disclose it.

When in doubt, disclose. No reviewer has ever rejected a paper for being too transparent about its methods.

What Happens If You Do Not Disclose

The consequences of failing to disclose AI usage are real and getting more serious.

During review. Reviewers and editors increasingly use AI detection tools. If undisclosed AI usage is suspected during peer review, your paper may be returned for revision or rejected outright. At best, this delays publication. At worst, it creates a negative impression with editors you may want to submit to again.

After publication. If undisclosed AI usage is discovered after a paper is published, the journal may issue a correction or, in serious cases, retract the paper. Retractions are public and permanent. They follow your name in every database.

Reputational damage. Academic communities are small. Being known as someone who hid AI involvement is far worse than being known as someone who used AI transparently. The academic world is moving toward treating undisclosed AI usage the same way it treats other forms of research misconduct.

None of this means you should be afraid to use AI tools. The problem is never the usage itself. The problem is the hiding of it.

How to Make Disclosure Quick and Painless

The main reason researchers skip disclosure is that it feels like extra work on top of an already demanding publication process. But it does not need to be burdensome.

Keep a running log. When you start a project, create a simple note listing which AI tools you use and what for. Update it as you go. When it is time to write the disclosure, you already have everything you need.

Use a standardized format. Free-form disclosure statements work, but a structured format like AI Usage Cards makes the process faster and more complete. You fill in a guided form, and it generates a formatted disclosure that meets the requirements of every major publisher.

Generate your disclosure at ai-cards.org. The free generator walks you through each category of AI usage and produces a professional document you can attach to your paper as supplementary material or adapt into an in-text disclosure statement. It takes about ten minutes.

The Bottom Line

AI disclosure is no longer optional at any serious academic venue. The good news is that it is straightforward to do, and doing it well actually strengthens your paper. Transparency about your methods, including your use of AI tools, is a sign of rigor, not weakness.

If you are ready to create your disclosure now, generate a free AI Usage Card and have it done in minutes. For ChatGPT-specific guidance, see our page on how to disclose ChatGPT usage in academic papers.