How to Cite ChatGPT and Other AI Tools in Turabian
Turabian 9 has no AI rule, so Chicago's official guidance applies: cite ChatGPT in a note, not the bibliography. Templates, examples and edition notes.
Turabian’s 9th edition contains no rule for citing ChatGPT, because it was published in 2018. The official fallback is the University of Chicago Press guidance for Chicago style, which Turabian follows: treat AI output like an unpublished personal communication, cite it in a note (or in your own sentence), and keep it out of the bibliography unless you can give readers a link they can actually open.
First footnote, no prompt in the note:
- Text generated by ChatGPT, OpenAI, September 3, 2026, https://chatgpt.com/.
First footnote, recording the prompt:
- ChatGPT, response to “Summarize the main arguments for and against the New Deal coalition thesis,” OpenAI, September 3, 2026.
Shortened note:
- ChatGPT, response to “Summarize the main arguments.”
That is the whole answer for most papers. The rest of this page explains where each element comes from, what changes when you have a shareable link, how the author-date system handles the same source, how the pattern carries over to Claude, Gemini and Copilot, and why citing an AI tool and being permitted to use one are two separate questions.
What to record before you write the note
Gather these details while the conversation is still open. Most of them disappear the moment you close the tab.
| Element | What to record | Example |
|---|---|---|
| Tool | The product name as the maker writes it | ChatGPT |
| Company | The organization behind the tool | OpenAI |
| Model, if it matters | The version you actually used | GPT-5 |
| Date | The date the response was generated | September 3, 2026 |
| Prompt | Your wording, exactly | ”Summarize the main arguments for and against…” |
| Link | A share URL that opens without a login, if one exists | https://chatgpt.com/share/… |
The date matters more here than in almost any other citation. AI tools are updated constantly, and the same prompt can return a different answer a week later, so the date is what tells your reader which version of the tool you were talking to.
Why Turabian 9 sends you to Chicago for this
Turabian is the student-facing version of Chicago style. The 9th edition (2018) is built on the 17th edition of The Chicago Manual of Style, and where Turabian is silent on a source type, Chicago’s guidance is the accepted authority.
Generative AI arrived after both books went to press. The University of Chicago Press answered the gap in its Style Q&A, and that answer is still the operative guidance for anyone writing in Turabian 9: AI output is not a published source with a stable location, so it behaves like an interview or an email. You cite it where the reader meets it, and you do not build a bibliography entry that sends them somewhere they cannot go.
Chicago’s 18th edition (2024) has since added a numbered section on citing AI-generated content, and the forthcoming Turabian 10th edition is described by the Press as including extensive coverage of the appropriate use of generative AI. Neither of those changes the advice above for a paper written in Turabian 9 today. Our Turabian 10th edition guide tracks what has actually been confirmed about the new edition.
The note, element by element
A full AI note has four moving parts, in this order:
- What produced the text. Either “Text generated by ChatGPT” or the tool name followed by “response to” and your prompt.
- The company. OpenAI, Anthropic, Google, Microsoft.
- The date the response was generated, written out in full: September 3, 2026.
- A URL, only if it leads somewhere your reader can actually reach.
Notice what is not there. No italics: the tool is not a book or a journal. No place of publication. No access date, because the date you generated the text already does that job. And no author in the human sense, which is exactly why the citation looks unusual.
Naming the tool in your own sentence
Chicago’s first suggestion for most writing is not a note at all. If you say in your text that ChatGPT produced something, you have already credited it, and a note is only needed when your instructor wants one or when the details matter:
The following outline of the debate was generated by ChatGPT (OpenAI) on September 3, 2026.
This reads better than a note in a short paper, and it keeps the AI’s role visible to the reader rather than buried at the bottom of the page.
Recording the prompt
Include the prompt when the exact wording shaped the output and your sentence does not already state it. Put the prompt in quotation marks inside the note:
- ChatGPT, response to “List the standard objections to Braudel’s long-duration model of history,” OpenAI, August 28, 2026.
If your sentence already says what you asked, the note can drop the prompt and use the shorter “Text generated by” form. Do not do both; repeating the prompt in the sentence and the note wastes the reader’s attention.
Shortening the note after the first use
Repeat citations of the same conversation follow the ordinary Turabian pattern: enough to identify the source, nothing more.
- ChatGPT, response to “List the standard objections.”
If you are citing several distinct AI responses in one paper, keep the prompts distinguishable in the shortened forms so a reader can tell them apart. For the general rules behind shortened notes, see our guide to the Turabian shortened footnote and to citing the same source multiple times.
When a shareable link exists
Most AI chats live behind a login. Some tools now let you publish a conversation to a public URL, and that changes the calculation, because your reader can verify what you saw.
When you have a genuinely public link, two things become reasonable:
A note with the link:
- ChatGPT, response to “Explain the Whig interpretation of history in two paragraphs,” OpenAI, September 3, 2026, https://chatgpt.com/share/example-link.
A bibliography entry, if your instructor wants one:
OpenAI. ChatGPT response to “Explain the Whig interpretation of history in two paragraphs.” September 3, 2026. https://chatgpt.com/share/example-link.
Two cautions. First, test the link in a private browser window before you submit. If it asks for a login, it is not a public link and the entry should come out. Second, official guidance on bibliography treatment for AI is thinner than it is for books and articles, and practice varies between departments. If your instructor has told you what they want, follow that instruction over anything on this page.
| Situation | Cite in text | Note | Bibliography |
|---|---|---|---|
| Private chat, no link | Yes | Yes | No |
| Public share link | Yes | Yes, with URL | Defensible, ask first |
| AI used only to brainstorm, nothing quoted | Disclosure, if required | Usually not | No |
| AI-generated image used as a figure | Figure caption | Yes | Rarely |
Author-date treatment
If your paper uses Turabian’s author-date system rather than notes, the same logic applies: AI output behaves like a personal communication, so it belongs in the text and not in the reference list.
In-text citation:
(ChatGPT, September 3, 2026)
With the prompt named in the sentence:
When asked to summarize the objections to the thesis, ChatGPT produced a three-part answer (ChatGPT, September 3, 2026).
A reference list entry only makes sense when a public link exists, and even then it is worth asking first. For the full rules on parentheticals, page numbers and reference-list order, see our Turabian author-date guide.
Claude, Gemini, Copilot and other tools
The pattern is about the kind of source, not the brand, so it carries over directly. Swap the tool name and the company:
- Text generated by Claude, Anthropic, September 5, 2026.
- Gemini, response to “Compare the two standard datings of the Gospel of Mark,” Google, September 5, 2026.
- Text generated by Microsoft Copilot, Microsoft, September 5, 2026.
Two situations need more care.
Tools that cite sources back to you. Some assistants return an answer with links attached. If you use a fact from one of those links, cite the linked source itself, not the assistant. The AI found the source; it did not write it.
Tools built into other software. When an AI feature inside a search engine, a word processor or a library database produces text you use, name the feature and its owner the same way, and say in your sentence where it appeared. Guidance here is genuinely unsettled, so describe what you did plainly rather than forcing it into a template that does not fit.
Citing AI is not the same as being allowed to use AI
These get confused constantly, and the distinction matters more than the punctuation.
Citation is a formatting question. It answers: where did this text come from, and how do I show that to my reader?
Permission is a policy question. It answers: was I allowed to use this tool for this assignment at all?
A correctly formatted note does not make unauthorized AI use acceptable. Many courses prohibit generative AI outright, some allow it for specific stages such as brainstorming or proofreading, and some allow it broadly but require a disclosure statement separate from your citations. A few departments ask for the full prompt log in an appendix.
Disclosure is a third thing again. It is a statement, usually near the start or the end of the paper, describing how you used AI across the whole project. A disclosure statement does not replace citations for specific quoted or paraphrased output, and citations do not replace a required disclosure.
Check the syllabus, the assignment brief and any departmental academic-integrity policy before you use a tool, not after. Nothing on this page is a judgment about whether your particular use is permitted; that decision belongs to your instructor and your institution.
Quoting, paraphrasing and keeping a record
Quoted AI output follows the same mechanics as any other quotation. Short runs stay in quotation marks in your sentence with a note number after the closing punctuation. If the passage runs to five or more lines of prose, it becomes a block quotation, set off and indented, with no quotation marks; our guide to Turabian block quotes covers the thresholds and the indentation.
Paraphrased output still needs the note. The test is not whether the words are yours but whether the idea, the structure or the selection of material came from the tool. A paragraph you rewrote from an AI outline is a paraphrase, and it gets a citation.
Because a chat can be deleted, edited or lost when an account changes, keep your own record while you work:
- Save the full prompt and the full response in a plain text file or a research log.
- Note the date and, if the tool shows it, the model version.
- Take a screenshot if the output includes formatting you plan to reproduce.
- Copy the share link, if one exists, and test that it opens in a private window.
That record costs a minute and answers the only question likely to follow later, which is what exactly the tool said and when. Some departments ask for it as an appendix, and having it already assembled is easier than reconstructing a conversation you no longer have.
Common mistakes
Treating the AI as an author. Writing “OpenAI, Kate. ChatGPT.” misreads the source. There is no human author and no title, which is precisely why the citation looks different from a book.
Italicizing the tool name. ChatGPT is not a titled work. Leave it in roman type.
Adding an access date. The date the response was generated already tells the reader when you saw it.
Pasting the browser URL from a private chat. A URL your reader cannot open adds nothing and suggests a verifiability you cannot deliver.
Citing the AI for a fact it reported. If the claim came from a study, cite the study. Verify that it exists first; AI tools sometimes produce references to sources that were never published.
Skipping the citation because you edited the text. If the structure or the substance came from the tool, credit it, even after you rewrote the wording.
Frequently asked questions
Does Turabian 9th edition have a rule for citing ChatGPT? No. Turabian 9 was published in 2018, before generative AI tools were widely used, so it contains no AI section. Writers using Turabian 9 follow the official Chicago guidance instead, which treats AI output like a personal communication: cite it in the text or in a note, and leave it out of the bibliography unless you can supply a publicly readable link.
Do I put ChatGPT in my bibliography? Usually not. A private chat cannot be opened by your reader, so it stays in a note only. If you have a share link that anyone can open without logging in, a bibliography or reference list entry becomes defensible, and some instructors ask for one.
Should I name the prompt I used? Include it when the prompt matters to your argument and is not already quoted in your sentence. If your own sentence already states what you asked, the note does not need to repeat it.
Is citing ChatGPT the same as being allowed to use it? No. Citation is a formatting question; permission is a policy question. Many courses restrict or ban generative AI, and some require a separate disclosure statement even when use is allowed.
Will the Turabian 10th edition change this? It is expected to. The Press describes the forthcoming tenth edition as including extensive coverage of the appropriate use of generative AI, and Chicago 18 already added a section on citing AI-generated content. Until the tenth edition is published and the actual rules can be read, keep following the current official guidance.
Before you submit
Write the note the way you would write one for an interview: name what produced the text, name the organization, give the date, and add a link only if it works for someone who is not you. Then check the syllabus, because the formatting is the easy half of this question.
For the source types Turabian does cover in detail, you can collect the author, title, date and URL and let the Turabian Citation Generator assemble the note and bibliography entry for you.