How to Quote and Cite an AI Chatbot Answer

Students, journalists and researchers increasingly want to quote what a chatbot said: as an example of a mistake, as a draft they then revised, or as the thing their essay is analyzing. The major style guides now say how to cite such text, and they agree on the essentials. What they cannot do is give the quote a permanent home, because the same prompt produces a different answer tomorrow. So the method on this page has two halves: cite the answer in the style your reader expects, and keep a record that shows exactly which words you are quoting.

Quote the words you received, cite the tool that produced them, and verify the facts elsewhere.

What the style guides agree on

All three treat a chatbot as a tool that produced text for you, never as an author, and all three ask for the date, because answers change. None of them treats the chatbot as a source for facts. If the answer states that a law passed in 2019 or that a study found something, you cite the law or the study, having found it yourself; the chatbot at most appears as the place the idea came from.

Capture, mark, cite: the method

1. Save the whole exchange the moment you get it

   Copy the prompt and the full answer into a document or a PDF, with the date, the tool's name and the version if it shows one. Many tools offer a share link for a conversation; save it too, but do not rely on it, because shared conversations can be deleted or changed.

2. Mark the exact words you will quote

   Highlight the passage in your saved copy, in one color, and nothing else. If you later shorten the quote, mark the omission in your saved copy as well, so the version you quote can always be compared with the version you received. Our guide to annotating a chatbot answer claim by claim explains how to mark the factual claims separately.

3. Check every factual claim outside the chatbot

   For each claim you rely on, find a source a reader could open and confirm it there. When you quote a person or a document the chatbot mentioned, go to that person or document, the way tracing a quotation back to the original describes. Chatbots misattribute quotations often, and a quotation you have not seen in its source should not appear in your work.

4. Write the citation in the house style

   Use the pattern in the table for APA, MLA or Chicago. Put the prompt where the style asks for it. If your teacher or editor has their own rule, theirs wins; ask before you submit.

5. Attach the record

   For coursework, add the saved exchange as an appendix or share the file with your teacher. For published work, keep it on file. When someone asks what the tool actually said, you can show the marked passage in seconds.

How to cite a chatbot answer: a worked example

A student is writing an essay on how well chatbots handle simple factual questions. On 15 September 2026 she asks a chatbot whether 1900 was a leap year, and the tool replies: "Yes, 1900 was a leap year, because 1900 is divisible by four." The exchange is invented for this example, but it is exactly the kind of confident error worth quoting.

1. Save. She copies the prompt and the whole reply into a document headed with the date, the tool's name and the version shown in its settings.
2. Mark. She highlights the one sentence she will quote and nothing else.
3. Check. The Royal Museums Greenwich page on leap years, from the home of the Royal Observatory, gives the rule: a year must be divisible by four, except end-of-century years, which must be divisible by 400, so 2000 was a leap year and 1900 was not. The chatbot's claim is wrong, which is the point of quoting it.
4. Cite. In APA style the reference names the tool's developer as author, the date of the chat, the chat's title with [Generative AI chat] after it, the tool or model name and the shared chat's URL, with a sentence in the text saying the prompt was "Was 1900 a leap year?". In MLA style that prompt becomes the title of the works-cited entry. The museum page gets an ordinary citation of its own.
5. Attach. The saved exchange goes in an appendix with the quoted sentence marked, so the marker can see the full context.

How to cite ChatGPT in APA, MLA and Chicago

ChatGPT is the chatbot students cite most, and all three major style guides now use it in their own examples. The pattern is the same for any chatbot: name the tool, give the date, describe or title the exchange, and point to a saved or shared copy. The templates below follow APA Style's September 2025 guidance on citing AI chats, the MLA Style Center's updated post on citing generative AI and the Chicago Manual of Style Q&A on AI-generated text.

Four details trip people up. First, none of the three guides treats ChatGPT as an author: APA puts the developer in the author position, MLA leaves the author element empty, and Chicago names the tool inside the note. Second, APA's 2025 update stopped asking for a version number by default; give the model name if the tool shows one, or just "ChatGPT". MLA, by contrast, asks for the model in its Version element. Third, use the Share link if the tool offers one: APA and MLA both prefer a URL that opens the exact conversation, and MLA says to fall back to the tool's general address only when no shareable link exists. Fourth, Chicago says to note it if you edited the generated text, for example "edited for style and content", though silent changes such as smart quotes need no mention.

If ChatGPT gave you a fact with a source attached, the MLA guidance says to click through and cite that source directly, because AI tools can invent sources or summarize them wrongly. That is the same habit as the Check step in the worked example above: the chatbot is cited only for its own words, never as the authority for a fact.

How to cite AI in APA: a chat, a tool, and where the prompt goes

The templates above cover ChatGPT, but APA's rule applies to any generative AI tool, and its first question is what you are citing. APA Style's September 9, 2025 post on reference formats for generative AI separates two cases: a specific chat, and the tool in general. In both, the author is the developer, the company responsible for the tool, because APA says AI itself cannot be an author.

A specific chat

Cite the chat when readers will benefit from seeing what the tool produced, for example when you quote it or discuss its output. APA defines the four elements this way: the date is the day the chat occurred or concluded; the title is the chat's title in italic sentence case, followed by a bracketed description such as [Generative AI chat]; and the source is the tool or model name, then the URL of the shared chat. APA suggests editing the chat's title in the tool before you write the reference, so it describes the chat. It also notes that readers who open a shared link see the original conversation, not later turns someone else added.

The tool in general

Cite the tool when citing a specific chat would be unhelpful, unethical or inappropriate. APA's examples include using AI to edit or refine your own writing, to translate your own writing, to create or refine code, to make tables or figures (say so in the table or figure note), and as part of a study's method, where citing every participant's chat could compromise confidentiality. The reference gives the developer, the year the tool was most recently updated (or the copyright date on its site or app if no update date is shown), the tool or model name in italics, a description such as [Large language model], and the tool's URL. If you used AI only to translate a source so you could understand it, cite the original source, not the tool.

The prompt and the disclosure

The prompt never goes in an APA reference. APA says the reference is for the chat, not the prompt; describe the prompt in your text, such as the Method section, or put it in an appendix or supplemental material when it helps readers. Describe how you used the tool in the body of the paper with an in-text citation. You may also mention AI use in the author note, but APA says the in-text citation should still appear in the body, and that author note policies vary by publisher.

Quoting a chatbot answer well

• Quote short. A sentence or two is enough to show what the tool said. Long pasted answers make it hard to tell your thinking from the tool's text.
• Never tidy the words. If the answer contains an error, quote the error and say so. Silent corrections turn a record into a paraphrase.
• Say what the quote is for. "The chatbot's first answer confused the two treaties" tells the reader why the quote is there; an unexplained block of generated text does not.
• Keep your analysis separate. Your own sentences should carry your argument, with the quoted text clearly marked as quoted.

These are the same habits that apply to quoting any source, which the general rules on quotation from university writing centers cover in detail. The difference is only that the chatbot's words cannot be looked up later by your reader, so your saved copy has to do that job. If you used a tool to help write the work itself rather than to supply a quotation, that is a disclosure question, covered in saying honestly what a writing tool did.

Pitfalls when quoting generated text

• Quoting from memory. Answers cannot be fetched again. If you did not save it, you cannot quote it; describe it instead and say that the record was not kept.
• Citing the share link alone. A link that may stop working is not a record. Cite it if the style asks, but keep the saved copy.
• Letting the quote carry a fact. If the chatbot said something true, the citation is still to the real source, found by you. The chatbot appears only when its words are the subject.
• Mixing regenerated answers. If you pressed regenerate, you have two answers. Quote one, say which, and keep both.
• Forgetting the version. Tools change models under the same name. Note the version string shown in the interface, or the date alone if none is shown.

When not to quote a chatbot at all

If you need a definition, a statistic or a quotation from a real person, quote the real source instead; a chatbot is a poor intermediary for facts because it gives no reliable way to trace where a sentence came from. And if your course or publication does not allow generated text, the citation does not make it allowed. For everything else, the more precisely you capture what the tool said, the easier it is to discuss honestly. The rest of the highlighting guides cover the reading skills behind this method, and a claim-by-claim source check is the general version of step three.

The saved exchange as a record of your work

The dated copy with the quoted sentence marked is a small authenticity record. It shows what you asked, what came back, which words you used and what you checked. Keep it with your drafts and your other marked sources, and it answers the disclosure question before it is asked: here is what the tool said, and here is what I did with it.

The same habit runs through the guides on highlights kept as a record of your sources. For quotations from people rather than tools, citing a quotation from a human author uses the same order of work: highlight the line in the source first, then write the citation from the highlight.

Citing the chatbot is one rule among several: where AI use becomes plagiarism covers the rest, and how to paraphrase covers rewording a source in your own words. The same styles cover other online sources; citing a YouTube video uses a similar pattern with a time stamp.