APA files ChatGPT under software. That makes OpenAI the author. The date? Just the year you used it. ChatGPT is the italicized title, the version goes in parentheses, and you can't drop that bracketed [Large language model] descriptor. MLA does the reverse. Here your prompt is the title of the source, and ChatGPT slides into the container slot. Both formats sit below, ready to copy, next to the in-text forms and the little variations that snag students.
How to cite ChatGPT in APA
APA's format comes from its official blog post on citing ChatGPT, which treats the tool under the rules for software with a company author. The published example looks like this:
OpenAI. (2023). ChatGPT (Mar 14 version) [Large language model]. https://chat.openai.com/chat
Swap in your own details. If you used ChatGPT this month, your entry would read:
OpenAI. (2026). ChatGPT (June 2026 version) [Large language model]. https://chat.openai.com/chat
Three details matter here. The date is only the year; the specific version date lives in the parentheses, written however OpenAI labels it in the interface. The bracketed [Large language model] is the source-type descriptor and cannot be deleted, the same way you would keep [Film] or [Mobile app] in other software references. And the URL is the tool's address as you accessed it: APA's example predates OpenAI's move to a shorter web address, so if your browser bar showed something different, write what you actually used.
APA in-text citation
In-text, the author is OpenAI, not ChatGPT. Both standard forms work:
Parenthetical: (OpenAI, 2026)
Narrative: OpenAI (2026)
What APA wants beyond the reference entry
The reference entry alone does not satisfy APA. The blog post tells you to describe how you used the tool in your Method section, or in the introduction if the paper has no Method section, the way a literature review or essay would not. That description is where you say what you asked and what you did with the answer.
When you quote a response, APA wants the prompt as well. The order runs prompt, then the generated text, then your in-text citation. ChatGPT spits out a different answer in every session, so APA also floats a second option: drop the full text of long responses into an appendix, or into online supplemental materials. That way a reader sees exactly what the model told you. Your grader can't open your chat, remember. The appendix ends up being the only trail.
How to cite ChatGPT in MLA
MLA won't name an AI tool as an author, full stop. The MLA Style Center's guidance on citing generative AI feeds the tool through the standard core-elements template instead: your prompt is the Title of Source, ChatGPT is the Title of Container, and then come version, publisher, date of generation, and the tool's general URL. Here is MLA's own example for paraphrased content:
"Describe the symbolism of the green light in the book The Great Gatsby" prompt. ChatGPT, 13 Feb. version, OpenAI, 8 Mar. 2023, chat.openai.com/chat.
Quoting works the same way. MLA's quoting example also shows how to handle a follow-up prompt in the same chat:
"In 200 words, describe the symbolism of the green light in The Great Gatsby" follow-up prompt to list sources. ChatGPT, 13 Feb. version, OpenAI, 9 Mar. 2023, chat.openai.com/chat.
Notice the location element in those entries: MLA drops the https scheme and prints only the domain path. That is deliberate MLA style for URLs, not a shortcut.
MLA in-text citation
In-text, MLA points to the first element of the works-cited entry, which is your prompt. Prompts run long, so shorten to the first phrase:
("Describe the symbolism")
Citing an AI-generated image in MLA
Images follow the same template with the image tool as container. MLA's example uses DALL-E and labels the image as a figure:
Fig. 1. "Pointillist painting of a sheep in a sunny field of blue flowers" prompt, DALL-E, version 2, OpenAI, 8 Mar. 2023, labs.openai.com/.
APA vs MLA at a glance
| Element | APA | MLA |
|---|---|---|
| Author position | OpenAI (the company) | None; entry starts with your prompt |
| Title of source | ChatGPT, italicized | Your prompt, in quotation marks |
| Container | Not used | ChatGPT, italicized |
| Version | In parentheses after the title | After the container, as specific as possible |
| Descriptor | [Large language model], required | Not used |
| Date | Year only | Full date you generated the content |
| URL | Full URL with https scheme | Domain form, no scheme |
| In-text form | (OpenAI, 2026) | ("First words of prompt") |
Which situations need a citation at all
Quoting or paraphrasing a response always gets a full entry in both styles. The gray zone is everything short of that, and the two guides split it slightly differently.
- You quote a response. Full citation in either style, plus the prompt. APA wants the prompt in your text and long outputs in an appendix; MLA folds the prompt into the works-cited entry itself.
- You paraphrase or borrow an idea. Still a full citation. MLA says to cite whenever AI-generated content is paraphrased or incorporated, not only when quoted verbatim.
- You used it to brainstorm or outline. No response text appears in the paper, so a reference entry misleads more than it helps. APA's route is the Method or introduction description; MLA says to acknowledge functional uses, editing or translating included, in a note or in the text.
- The model cited a source for you. Do not cite it secondhand. MLA is explicit that you must vet the sources an AI names, and models still fabricate references. Find and read the real source, then cite that.
A quick word on Chicago
Chicago handles ChatGPT in a numbered footnote rather than the bibliography, since a reader cannot retrieve the conversation. The format from the Chicago Manual of Style's Q&A on citing AI:
Text generated by ChatGPT, OpenAI, March 7, 2023, https://chat.openai.com/chat.
If you edited the model's output before using it, Chicago asks you to say so at the end of the note, for example "edited for style and content." Keep the entry out of the bibliography unless you have a publicly shareable URL for the exchange.
When citing is not enough
A perfect reference entry settles nothing about permission. Citation formats show you how to credit a source you were cleared to use. They say zero about whether your instructor, department, or journal lets you use it at all. Plenty of syllabi ban generative AI flat out, and a flawless APA entry under a banned tool isn't diligence, it's documented misconduct. So read the course policy before you reach for the style guide. If you want the fuller picture of where ChatGPT use crosses into plagiarism, know that the rules split between institutions more than students expect.
Disclosure is the second layer. APA and MLA both expect you to state how you used the tool now, not just tuck an entry at the back. One sentence does it, in the method section or an acknowledgments note: what you asked, what you kept, what you changed. When questions surface later, vague disclosure reads worse than saying nothing at all.
The third layer is the text itself. Models lift phrasing from their training data more than most people assume. That means a passage you tagged "generated by ChatGPT" might match a published source word for word, and pointing to the model does nothing to credit the person who wrote it first. Before you submit, push the drafted sections through the free plagiarism checker on TextToHuman. No account. No word limit. It even surfaces the real matching source URLs, so you cite the true origin and skip the middleman.
Cite the model, disclose the use, verify the text. Do those three in order and the bibliography stops being the risky part of the paper.
