AI Disclosure for Humanities Research

Humanities researchers need a different kind of AI disclosure

A historian who asks an AI tool to summarize a letter from 1892 has a different problem than a chemist who asks one to polish an abstract.

A literary scholar who asks for alternate readings of a sonnet has a different problem again.

Humanities research often depends on voice, interpretation, translation, archive work, and judgment. That makes [[AI disclosure](/ai-disclosure-in-systematic-reviews-and-meta-analyses/)](/how-to-disclose-ai-use-for-neurips-icml-and-acl-submissions/) feel awkward. You may not have a Methods section. You may not have data, code, or a lab workflow. You may have a close reading, a textual note, a translation choice, or a critical argument.

You still need to tell readers where AI entered the work.

An [[AI Usage Card](/ai-disclosure-for-social-science-research/)](/what-are-ai-usage-cards/) gives you a short record for that. It names the tool, the task, the stage of work, the human checks, and the limits. You can add it to an acknowledgment, an appendix, a submission form, or your own project notes.

Start with the reader's question

Readers do not need a confession. They need a map.

A good disclosure answers one plain question: where could AI have shaped what I am reading?

That question works well in humanities fields because it avoids fake precision. You do not need to estimate that AI affected "12 percent" of a paper. You need to say whether AI touched translation, transcription, search terms, prose, summaries, images, annotations, citations, or argument development.

The MLA-CCCC Joint Task Force on Writing and AI has published work for writing, literature, and language programs, including guidance on policy development for AI in instructional settings. That matters for humanities scholars because the field treats writing as part of knowledge, not as packaging after the "real" research ends. (cccc.ncte.org)

If an AI tool helped shape the words, it may have helped shape the thought. That does not make the work invalid. It makes the process worth naming.

For a general decision tree, start with Do I Need to Disclose AI Usage in My Paper?. For prose drafted with ChatGPT, see How to Disclose ChatGPT Usage in Academic Papers.

What counts as AI use in humanities work

Humanities scholars use AI in small ways that can look harmless in isolation. The trouble comes when those small choices affect interpretation.

Consider a few common cases.

You ask Gemini to summarize a long review in a nineteenth century newspaper. You later quote the review, but the summary shaped what you noticed.

You ask Claude to improve the tone of a paragraph on colonial archives. The tool softens a claim. You accept the wording.

You ask ChatGPT for "possible theoretical lenses" for a chapter on memory. One suggestion sends you toward a body of scholarship that you had not planned to cite.

You ask an image tool to create a diagram of manuscript transmission. The figure looks clean, but it hides uncertainty in the stemma.

You ask an OCR correction tool to clean a handwritten diary transcription. It changes a name.

None of these cases mean you did something wrong. They mean your disclosure should name the task.

The MLA Style Center advises writers to go past citation when they describe AI use. It notes that AI output itself does not count as a source and that writers should click through, read the source text, and cite the source directly when AI points them toward material. (style.mla.org)

That advice fits humanities research well. If AI helps you find a passage, cite the passage, not the chatbot. If AI gives you an interpretation, do not treat it as a scholarly source. Tell readers that you used it for brainstorming or drafting, then take responsibility for the claim.

Disclosure is not the same as citation

Humanities researchers live inside citation systems. That can make AI disclosure confusing.

Citation tells readers where a claim, quotation, image, or idea came from.

Disclosure tells readers how you made the work.

Sometimes you need both.

If you quote AI-generated text, cite it according to the style guide you use. The Chicago Manual of Style says writers should credit ChatGPT and similar tools when they use generated text, and it gives examples for notes that include the tool, the company, the date, and the prompt when needed. Chicago also says that if you edited AI-generated text, you should say so. (chicagomanualofstyle.org)

Most humanities articles will not quote AI-generated text. They will use AI for search, revision, translation support, or draft feedback. In those cases, a disclosure statement or AI Usage Card often works better than a footnote.

For journals, match the venue first. Our guide to AI Transparency Requirements for Journal Submissions explains how authors can align a disclosure with submission rules.

A field-sensitive disclosure checklist

You do not need a long statement. You need a specific one.

Before submission, ask:

Did AI generate or revise prose that appears in the paper?
Did AI summarize primary or secondary sources?
Did AI translate, paraphrase, transcribe, annotate, classify, or code material?
Did AI suggest sources, concepts, examples, titles, or structure?
Did AI create or modify an image, table, diagram, edition, or teaching artifact?
Did any AI output affect the argument?

If you answer yes to any question, write it down.

Then add human checks. Humanities readers care about this because interpretive work turns on detail: a word choice in a poem, a marginal note in an archive, a translation of a legal term, the date attached to an unsigned pamphlet.

A weak disclosure says:

"AI tools were used during preparation of this manuscript."

A better disclosure says:

"I used Claude 3.5 Sonnet in February 2026 to test alternate outlines for section 2 and to identify unclear transitions in a draft. I did not use AI to generate primary source analysis. I checked all citations and quotations against the cited editions."

The second version gives readers something they can assess.

For more examples, use AI Usage Cards Examples and Templates.

Where to put the disclosure

Humanities papers often have fewer standard reporting sections than empirical articles. Choose the location that readers and editors will see.

For a journal article, place a short AI disclosure before the acknowledgments, in the acknowledgments, or in a separate "AI use statement" if the journal allows it.

For a book chapter, put it in the acknowledgments or a note on method.

For a dissertation or thesis, place the statement near the preface, acknowledgments, or methods chapter. If your graduate school asks for a declaration, follow that format and attach a fuller card as an appendix. See How to Disclose AI Usage in Your Thesis for thesis-specific wording.

For a digital humanities project, place the disclosure in the project documentation, repository README, or "About this project" page.

For peer review, do not upload confidential manuscripts into public AI systems unless the journal or editor allows it. Nature Portfolio asks peer reviewers not to upload manuscripts into generative AI tools, and it asks reviewers to declare any AI support used in evaluating claims. (nature.com)

Sample AI disclosure statements for humanities research

Use these examples as starting points. Replace the tool names, dates, tasks, and checks with your own record.

Language editing only

"I used DeepL Write and Grammarly in March 2026 to revise grammar and sentence clarity in the final manuscript. I reviewed all suggested changes and accepted only edits that preserved my meaning. I did not use AI to generate arguments, source interpretations, translations, or citations."

Some publishers do not require disclosure for narrow copy editing. Nature Portfolio, for example, says authors do not need to declare LLM use for "AI assisted copy editing" when it only improves human-generated text for readability, style, grammar, spelling, punctuation, and tone. It still requires human accountability for the final text. (nature.com)

Check your venue before you omit the disclosure. When in doubt, add one short sentence.

Archival summary support

"I used ChatGPT-4o in April 2026 to create preliminary summaries of digitized newspaper articles from the Chronicling America collection. I treated those summaries as research notes only. I read each cited article myself, checked all quotations against the original scans, and cited only the primary sources."

Translation support

"I used DeepL and Claude 3.5 Sonnet in January 2026 to compare draft translations of selected German passages. I made all final translation choices. I checked idioms, dates, and names against the source text and standard dictionaries. The article marks all translations as my own unless otherwise noted."

Argument feedback

"I used Claude 3.5 Sonnet in May 2026 to ask where a draft section on memory and testimony lacked clear transitions. The tool did not generate source analysis or final prose. I revised the section myself and verified all citations."

Digital humanities classification

"I used a local transformer model to cluster 1,240 pamphlet titles by recurring themes during exploratory analysis. I manually reviewed the clusters used in the article and corrected misclassified titles before drawing conclusions."

This last example belongs near a method note because the tool shaped analysis. For related ethical questions, read AI Ethics and Documentation in Academic Research.

Add an AI Usage Card to a humanities paper

An AI Usage Card works because it gives your disclosure a stable shape. You can keep it in your notes during drafting, then copy the short text into the manuscript.

A card for a humanities article might include:

Tool: Claude 3.5 Sonnet
Date: May 2026
Stage: outlining and revision
Purpose: identify unclear transitions
Exclusions: no AI-generated close readings, quotations, translations, or citations
Human checks: author reviewed all suggestions and verified all references

That is more useful than a vague sentence. It lets an editor see the boundary.

You can generate a card at ai-cards.org and paste the text into your acknowledgment. If you write in LaTeX, you can add a short statement like this:

\section*{AI use statement}
 
The author used Claude 3.5 Sonnet in May 2026 to identify unclear
transitions in an early draft and to test alternate section headings.
The tool did not generate primary source analysis, translations,
citations, or final prose. The author reviewed all suggestions and
checked all quotations and references against the cited sources.

If your journal allows appendices, add the fuller card there. For formatting help, see the LaTeX Tutorial for AI Usage Cards or the Overleaf guide.

Do not list an AI tool as an author

Humanities scholars may talk about authorship in richer ways than journal policies do. A literary theorist can spend a career troubling the idea of the author.

Submission systems will still ask for accountable humans.

COPE-based guidance states that AI tools cannot appear as authors and that authors who use AI in writing, figures, data collection, or analysis should disclose how they used the tool and which tool they used. (iodp.tamu.edu)

That rule protects more than bureaucracy. It tells editors who can answer questions, approve changes, handle errors, and take responsibility for the work.

If AI shaped your paper, disclose the help. Do not give the tool credit as a co-author. For a fuller explanation, see Can AI Be a Co-Author on a Research Paper?.

Keep the disclosure close to the scholarly risk

The best humanities AI disclosure does not sound dramatic. It sounds exact.

If AI cleaned grammar, say that.

If AI helped translate primary sources, say that and name your checks.

If AI summarized archives, say that and explain how you verified the sources.

If AI generated draft prose that remains in the paper, say where.

If AI shaped interpretation, do not hide that under "editing." Readers can handle the truth. They cannot assess a blank space.

Generate an AI Usage Card at ai-cards.org before you submit. Then copy the short disclosure into your manuscript, appendix, or cover letter. A clear card tells readers what you did, what you checked, and where your judgment begins.