How to Cite ChatGPT and AI in MLA, APA, and Five More Styles

By MLA Generator Editorial Team · Updated July 7, 2026

A ChatGPT conversation breaks every assumption a citation template makes. There is no human author — MLA explicitly says not to treat the AI as one. There is no fixed title, no publication, and no guarantee the same prompt will ever produce the same words again. On top of that, both official style bodies rewrote their guidance recently: MLA revised its generative AI page in August 2025, and APA replaced its widely copied 2023 format with a new chat-level format in September 2025. Most library guides still show the old APA entry with “(Mar 14 version)” and a chat.openai.com URL. This page uses the current templates.

The shortest answer: in MLA, the prompt is the title. Put your prompt in quotation marks followed by the word “prompt,” then ChatGPT in italics, the model, OpenAI, the date you generated the text, and the share link: “Describe the theme of nature in Jane Austen’s Mansfield Park” prompt. ChatGPT, model GPT-4o, OpenAI, 23 Sept. 2024, chatgpt.com/share/66f1b0a0-d704-8000-be9a-85f53c850607. In-text, use a shortened title: (“Describe the theme”).

What counts as “citing ChatGPT”

This page covers text you generated with a conversational AI tool — ChatGPT, Claude, Gemini, Copilot — and quoted, paraphrased, or otherwise used as a source in your writing. The same templates work for all of these tools; only the tool name, model, and company change.

Several things that look like citing an AI are actually citing something else:

• The chatbot pointed you to sources. MLA is blunt about this: “you should click through to the cited sources and use those directly.” Cite the web page or journal article the AI surfaced, not the AI’s summary of it. APA says the same about Google’s AI Overviews — click through and cite the underlying source.
• You used AI as a tool, not a source. If ChatGPT fixed your grammar, translated a passage, or tightened your prose, MLA says to acknowledge that in a note or in the text rather than build a citation. Nothing goes in the Works Cited.
• You are citing an AI response someone else published. A screenshot of a chatbot answer in a news article cites as the news article.
• You generated an image. MLA handles AI images with a figure caption rather than a Works Cited entry — see the MLA section below, and the image guide for images generally.

If you genuinely used the model’s own output as material in your paper, read on.

Information to collect before you cite

Before the conversation scrolls away, capture these fields:

• The tool — ChatGPT, Claude, Gemini. This is the container in MLA and part of the source line in APA.
• The model, as specifically as possible — GPT-4o, GPT-5, Claude Sonnet 4. MLA asks for “the specific AI model or model version as specifically as possible.”
• The company — OpenAI, Anthropic, Google, Microsoft. Publisher in MLA; the author in APA.
• Your prompt, verbatim — MLA builds the title from it; APA wants it in your text or an appendix when you quote output.
• The date you generated the text — not the date you later pasted it into your draft. Both styles date the generation.
• A stable share link — chatgpt.com/share/… or claude.ai/share/…. Both styles now prefer it over the tool’s homepage. Note the domain: current ChatGPT links live at chatgpt.com, not the old chat.openai.com.
• The chat’s title — the label in the sidebar (ChatGPT auto-generates one). APA’s 2025 format uses it as the reference title.

The generator at / accepts a pasted share link and extracts most of these fields automatically; double-check the model name, which share pages do not always expose.

The same chat, formatted in all seven styles

The source: the MLA Style Center’s own published example — a ChatGPT (model GPT-4o) response to the prompt “Describe the theme of nature in Jane Austen’s Mansfield Park,” generated by OpenAI’s tool on September 23, 2024, and saved at the share link chatgpt.com/share/66f1b0a0-d704-8000-be9a-85f53c850607. The chat’s sidebar title is “Nature Theme in Mansfield Park.” Accessed July 4, 2026.

One decision splits the styles more than any other: who takes the author slot. MLA leaves it empty and leads with the prompt, because “We do not recommend treating the AI tool as an author.” APA credits the company, OpenAI, exactly as it would a corporate author. Chicago credits the tool itself, ChatGPT, with OpenAI as publisher. AMA agrees with MLA that the AI is no author but reaches the opposite conclusion about the entry: its manual cites the tool like software — name, version, company, year — and keeps the prompt out of the reference entirely. Harvard, following Cite Them Right, credits the AI company (OpenAI) as the author. Vancouver and IEEE have no dedicated AI rule yet, so they treat the share page as an ordinary web document — here with OpenAI as its corporate author and the prompt as the title, as the table shows.

The second decision is where the prompt lives. MLA puts it in the citation itself, as the Title of Source. APA keeps it out of the reference (the chat title goes there instead) and asks you to reproduce the prompt in your text, your Method section, or an appendix.

The in-text citation for the same chat:

• MLA: (“Describe the theme”) — a shortened version of the prompt title in quotation marks
• APA: (OpenAI, 2024), or narratively: OpenAI (2024)
• Chicago author–date: (ChatGPT, September 23, 2024) — Chicago keeps the full date, as it does for personal communications
• Harvard: (OpenAI, 2024) — Cite Them Right puts the AI company in the author slot
• Vancouver, AMA: (1) or superscript ¹; IEEE: [1] — numeric styles cite by entry number

See the in-text citations guide for how these forms behave inside sentences.

How to cite ChatGPT in MLA

MLA 9 runs generative AI through its ordinary core-elements template. The mapping, straight from the MLA Style Center’s updated guidance:

“Your prompt” prompt. Tool Name, model Model Name, Company, Day Month Year, share URL.

• Title of Source — the prompt in quotation marks, followed by the lowercase word “prompt.” If the prompt is too long, substitute a short description of what was generated.
• Title of Container — the tool, italicized: ChatGPT, Claude, Gemini.
• Version — the model, “as specifically as possible”: model GPT-4o, model Claude Sonnet 4.
• Publisher — the company: OpenAI, Anthropic, Google.
• Publication date — the date the content was generated.
• Location — “a stable, shareable URL for accessing the generated content.” Only if the tool cannot produce one do you fall back to the general URL (chatgpt.com).

The official example:

“Describe the theme of nature in Jane Austen’s Mansfield Park” prompt. ChatGPT, model GPT-4o, OpenAI, 23 Sept. 2024, chatgpt.com/share/66f1b0a0-d704-8000-be9a-85f53c850607.

In-text, shorten the title the way you would any long title: (“Describe the theme”). Paraphrased AI output takes exactly the same citation as quoted output.

Three variants matter in practice:

The AI produced a titled work. If you asked for a poem or story and the output has a title, lead with that title and describe the genre: “The Oak Tree” free verse poem. ChatGPT, model GPT-4o, OpenAI, 23 Sept. 2024, followed by the URL.

You generated an image. MLA cites AI images in a figure caption, not the Works Cited list: Fig. 1. “Create an expressionist-style image of two people standing on a beach looking at the ocean” prompt, DALL-E, version 3, OpenAI, 23 Sept. 2024, then the share URL.

A different chatbot, same template. For Claude: “Your prompt” prompt. Claude, model Claude Sonnet 4, Anthropic, date, claude.ai/share/…. Swap tool, model, and company; nothing else changes.

How to cite ChatGPT in APA

APA 7 changed its recommendation on September 9, 2025, and the new guidance supersedes the 2023 “How to cite ChatGPT” post that most library guides still reproduce. The current rule: when you have a shareable link, cite the specific chat.

AI Company Name. (Year, Month Day). Title of chat [Generative AI chat]. Tool Name/Model. URL of the chat

The two official examples:

OpenAI. (2025, August 21). High school grammar concepts [Generative AI chat]. ChatGPT. https://chatgpt.com/share/68a77b60-0ee4-800c-9acc-cd3fd573c311

Anthropic. (2025, May 20). Essential grammar topics for high school graduates [Generative AI chat]. Claude Sonnet 4. https://claude.ai/share/329173b2-ec93-4663-ac68-4f65ea4f166d

The company is the author, the chat’s title (the sidebar label) is the reference title in italics and sentence case, and the tool name or the model as displayed fills the source slot. In-text: (OpenAI, 2025), or OpenAI (2025) in a narrative sentence.

No shareable link? Cite the tool generally and move the specifics into your paper:

OpenAI. (2025). ChatGPT-5 [Large language model]. https://chatgpt.com/

Then reproduce the prompt and the quoted output in your text, your Method section, or an appendix (an appendix is the place for long transcripts). APA also asks you to describe how you used the tool, whichever reference format you pick.

One warning. The 2023 format — OpenAI. (2023). ChatGPT (Mar 14 version) [Large language model]. https://chat.openai.com/chat — is outdated on both counts: the format is superseded and the old chat.openai.com address now just redirects to chatgpt.com. If a guide shows “(Mar 14 version),” it predates the September 2025 change.

Edge cases

The references ChatGPT gave you might be fake. Language models fabricate plausible-looking citations: real journals, real-sounding authors, articles that do not exist. Never move a citation from a chatbot into your Works Cited without opening the actual source and confirming it says what the AI claims. A fabricated reference in a submitted paper is worse than a formatting error; it reads as fraud.

The chat is gone. Share links can be revoked and conversations deleted, and regenerating the same prompt will not reproduce the same words. The citation still carries the original generation date; do not re-date it. Protect yourself the way you would with any unstable web source: export or screenshot the conversation before you cite it, so you can produce the transcript if a reader or instructor asks.

Your tool cannot make a share link. Enterprise Copilot, school-locked accounts, and some modes of consumer tools disable sharing. MLA’s fallback is the tool’s general URL; APA’s fallback is the tool-level reference plus the prompt and output reproduced in your text or an appendix. Both are covered above — the practical point is that the burden of evidence shifts to you, so keep the transcript.

Chicago and the remaining styles. Chicago’s official position (CMOS 18 and its FAQ) is to credit the AI in a note, treated like a personal communication: “Text generated by ChatGPT, OpenAI, March 7, 2023, https://chat.openai.com/chat.” ChatGPT stands as the “author,” OpenAI as publisher, and the entry only joins a bibliography when a publicly available link exists — which is why the table above can include one. Chicago also asks you to note it if you edited the AI’s text. AMA does have a rule, and it points the other way: the AMA Manual of Style says not to credit the AI as an author and cites the tool like software — ChatGPT. Version Mar 14. OpenAI; 2023. Accessed March 28, 2023. https://openai.com/ — with the prompt and how you used the tool described in the text or Methods, which is what the AMA row of the table follows. IEEE likewise has no reference format for AI but requires AI-generated content to be disclosed in the acknowledgments, and points authors to its software reference format. Vancouver has published no dedicated generative AI format as of mid-2026, so treating the share page as a web page — with OpenAI as corporate author and the prompt as its title, as the table does — is the defensible reading of its existing web-source rules. Harvard, as codified by Cite Them Right, does have guidance: cite AI content that is available online in the format for that medium with the AI company as author, which is why the Harvard row leads with OpenAI. For output only you can see, with no shareable link, Cite Them Right instead treats it as a personal communication: OpenAI ChatGPT (2024) ChatGPT response to [your name], 23 September.

You used AI but produced the words yourself. Brainstorming, outlining, critique of your draft: no citation exists for influence. MLA’s advice for functional uses applies — acknowledge the tool in your text or a note where the use was substantive. Silence about substantive AI help is the risk, not the missing citation.

Citing does not equal permission. A perfect citation documents AI use; it does not authorize it. Course policies, journal policies, and institutional rules on generative AI vary enormously, and passing off AI-generated analysis as your own thinking is misconduct with or without a reference. When in doubt, ask first — and see avoiding plagiarism for where the lines sit.

A final principle: a reader must be able to see what the machine actually said. Name the tool and the model, date the generation, and link the exact conversation whenever the tool allows it. When it does not, the transcript in your appendix or your files does the job the link would have done. The citation formats differ; the obligation does not.

Editorial methodology and sources