How to disclose Google Gemini use in academic writing

Gemini use belongs in your research record

Gemini can help with a paper in many ways. It can polish sentences, summarize PDFs, suggest code, draft search strings, translate passages, or inspect images. Some uses change the manuscript. Some touch the research process before a sentence reaches the page.

That difference matters.

A reader does not need to know that you asked Gemini for five title ideas and ignored them. A reader may need to know that you used Gemini to rewrite the discussion section, generate Python code, classify interview excerpts, translate participant quotes, or summarize studies during screening.

The safest rule is simple: disclose any Gemini use that could affect the content, evidence, analysis, interpretation, or presentation of the work. If you feel unsure, create a short record now. You can decide later what the journal needs.

If you want a structured way to do this, generate an [[AI Usage Card](/ai-disclosure-for-social-science-research/)](/) while the details remain fresh. You can use the card as an appendix, adapt it into an acknowledgments note, or keep it in your project folder for submission questions.

Name the Gemini environment you used

"Gemini" can mean several things in a research workflow. Do not stop at the brand name.

Record whether you used the consumer Gemini app, Gemini inside Google Workspace, Google AI Studio, the Gemini API, Gemini Code Assist, or a Gemini model through another platform. These services can differ in account type, data handling, available model versions, file upload behavior, and admin controls.

Google's Gemini Apps Privacy Hub says human reviewers may review some data that Google collects, and it warns users not to enter confidential information that they would not want a reviewer to see or Google to use. Google Workspace gives different commitments for eligible Workspace use: Google says Workspace content is not human reviewed or used for generative AI model training outside the domain without permission. (support.google.com)

That difference can matter for ethics approval, data protection, and journal questions. A disclosure does not need to quote a privacy policy. It should name the setting you used so a reader can understand the context.

A thin disclosure says:

"Gemini was used during writing."

A useful disclosure says:

"Google Gemini, accessed through a university Google Workspace account, was used to draft alternative phrasings for the introduction and to summarize reviewer comments. The authors checked all output and made the final edits."

That second version gives readers something to work with.

Separate writing help from research help

Journal editors care about what the tool did.

If Gemini helped with grammar, you can say that. If Gemini helped with analysis, say that too, but do not hide it under "language editing." A vague phrase can make a routine use look suspicious.

For language work, record the sections Gemini touched. Did you ask it to shorten the abstract? Improve transitions? Translate a paragraph from German to English? Suggest simpler wording for a methods section?

For research work, record the task and the human check. Did you ask Gemini to summarize papers, generate R code, classify responses, extract variables, or compare themes? That deserves more detail because it may affect results.

This distinction also appears in other tool guides, such as how to disclose ChatGPT usage in academic papers and [how to disclose Claude use in academic writing](/how-to-disclose-claude-use-in-academic-writing/). The tool name changes. The disclosure logic does not.

Record five facts before you submit

A good Gemini disclosure answers five questions:

Which Gemini service or model did you use?
When did you use it?
What did you ask it to do?
Which parts of the paper or project did it affect?
How did you check the output?

You do not need to save every prompt in the paper itself. For sensitive projects, you may not want to publish prompts at all. But you should keep enough detail to reconstruct the role of the tool.

For example:

"Google Gemini 2.5 Pro in the Gemini app was used in June 2026 to suggest clearer wording for parts of the introduction and discussion. The authors reviewed all suggestions, checked terminology against the source literature, and wrote the final manuscript."

Or:

"Google AI Studio with Gemini 2.5 Flash was used in May 2026 to prototype Python scripts for data cleaning. The authors inspected, revised, and tested the code before analysis."

For Gemini API and AI Studio, account setup can affect data handling. Google's Gemini API terms say that when users activate Cloud Billing, Gemini API and Google AI Studio use counts as a "Paid Service" for data use, and Google says it does not use prompts or responses from Paid Services to improve its products. Google's zero data retention page also notes that for Paid Services, Google logs prompts and responses for a limited period for abuse monitoring unless zero data retention applies. (ai.google.dev)

That does not mean your disclosure should become a privacy essay. It means you should record the service tier if it matters for your institution or ethics file.

Do not paste confidential research data into consumer Gemini

This sounds obvious until a deadline gets close.

Do not paste identifiable interview transcripts, unpublished grant reviews, patient data, student records, proprietary datasets, or confidential peer review text into a consumer AI tool unless your institution has approved that workflow. If you need Gemini for sensitive work, ask your data protection officer, IT team, ethics board, or project PI which account type and settings you may use.

Google's own consumer Gemini guidance warns users not to enter information they would not want reviewed or used. Workspace and API terms may offer different protections in some account settings, but you still need project approval before you process restricted data. (support.google.com)

If you work with interviews or field notes, read the guide on [AI disclosure for qualitative research](/ai-disclosure-for-qualitative-research/). If your work involves journals, check AI transparency requirements for journal submissions before you upload files to any tool.

Example disclosure statements for Gemini

Use these examples as starting points. Change the wording so it matches what you did.

Minor language editing

"Google Gemini was used to suggest grammar and clarity edits for selected paragraphs in the introduction and discussion. The authors reviewed all suggestions and approved the final text."

This fits low risk writing support. It does not fit analysis, coding, screening, or interpretation.

Drafting support

"Google Gemini was used to generate alternative phrasings for parts of the introduction. The authors selected, revised, and verified the final wording. Gemini was not used to generate results, interpret findings, or write the final manuscript without author review."

Use this when Gemini shaped some prose but did not create evidence or analysis.

Literature search support

"Google Gemini was used to generate preliminary search terms and to summarize candidate articles during scoping. The authors verified all bibliographic details, read the cited works, and made all inclusion decisions."

This disclosure works better than "used for literature review" because it names the limited task. For systematic reviews, you need tighter records. See AI disclosure in systematic reviews and meta-analyses.

Code assistance

"Google Gemini was used to draft Python code for data cleaning and figure formatting. The authors reviewed, revised, and tested the code. The final analysis scripts are available in the project repository."

If you used Gemini inside an IDE or through the Gemini API, name that. If the code affects analysis, keep versioned scripts.

Translation or multilingual writing

"Google Gemini was used to draft English translations of selected German text passages. The authors, who are fluent in both languages, checked the translations against the source text and revised them before submission."

If you translate participant quotes, check consent language and anonymization before using any AI tool. Humanities researchers may also want the field specific advice in AI disclosure for humanities research.

Add an AI Usage Card in LaTeX

Some papers only need a short note. Theses, preprints, supplementary files, and methods appendices often benefit from a table.

You can generate a card at ai-cards.org, copy the text, and paste it into your manuscript. If you write in LaTeX, you can use a compact version like this:

\section*{AI Usage Disclosure}
 
\begin{table}[h]
\centering
\begin{tabular}{p{0.28\linewidth} p{0.62\linewidth}}
\hline
Tool & Google Gemini \\
Access mode & Gemini app through university Google Workspace account \\
Date of use & June 2026 \\
Purpose & Language revision for the introduction and discussion \\
Research data entered & No participant data, unpublished data, or confidential material \\
Human oversight & The authors reviewed all suggestions, checked terminology, and wrote the final text \\
Affected sections & Introduction, Discussion \\
\hline
\end{tabular}
\caption{AI Usage Card for Gemini-assisted writing.}
\end{table}

If your journal prefers a paragraph, convert the same card into prose:

\section*{Declaration of AI-assisted work}
 
Google Gemini, accessed through a university Google Workspace account,
was used in June 2026 to suggest language revisions for the introduction
and discussion. No participant data, unpublished data, or confidential
material was entered. The authors reviewed all suggestions, checked
terminology against the source literature, and wrote the final text.

For more formatting options, use the [LaTeX tutorial for AI Usage Cards](/latex-ai-usage-cards-tutorial/) or the Overleaf AI Usage Cards guide.

Match the disclosure to the submission venue

Some journals ask for an AI declaration in a specific box. Others expect a note in the acknowledgments, methods, cover letter, or supplementary material. Conference systems may ask separate questions about AI tools.

Do not bury substantive Gemini use in a cover letter only. If Gemini helped with methods, data handling, analysis, coding, screening, or interpretation, put the disclosure where readers can find it. Usually that means the methods section, a declaration section, or a supplement.

If Gemini only helped with copyediting, a short declaration often works. Check the journal rules before you submit. The article on AI disclosure policies by major journals can help you compare common requirements, and do I need to disclose AI usage in my paper? gives a simple decision path.

Keep a private record even if the paper uses a short note

Your published disclosure should stay readable. Your private record can carry more detail.

Save the tool name, access mode, date range, major prompts, output role, files uploaded, data safeguards, and verification steps. If a reviewer asks how Gemini affected your work, you will not have to reconstruct the answer from memory.

An AI Usage Card helps because it keeps the record in one place. The AI Usage Cards examples and templates page shows formats for short declarations, methods notes, and appendices.

A practical rule for Gemini disclosure

If Gemini changed only style, disclose style support.

If Gemini touched evidence, code, data, analysis, or interpretation, disclose the task and your checks.

If Gemini saw sensitive data, stop and review your ethics approval, consent language, and institutional policy before you go further.

Generate an AI Usage Card at ai-cards.org before submission. Use it as your working record, then adapt it into the journal's AI declaration box, acknowledgments section, methods note, LaTeX appendix, or supplementary file.