How to Cite AI in an Essay: APA, MLA, and Chicago, With Real Examples

Each major style handles AI differently. APA treats the company as author — (OpenAI, 2023) in text, full entry in the reference list. MLA refuses to treat AI as an author and cites your prompt as the title. Chicago cites AI in a footnote only and keeps it out of the bibliography. Exact templates for all three are below.

Two things before the templates. First, a citation is not a permission slip: citing ChatGPT beautifully in a course that bans ChatGPT is still a violation, so confirm what your course policy actually allows before formatting anything. Second, all three styles published their AI guidance as interim blog-and-FAQ guidance in 2023 and have updated it since — formats can shift, so we note what we verified and when, and your instructor's version wins any conflict.

Why AI citation is genuinely awkward

Citation systems were built on a stable assumption: a source is a fixed thing someone can go find. AI output breaks the assumption three ways. There's no human author to credit. The "source" is often unretrievable — ask ChatGPT the same question twice and you get two different answers, so a reader can never check your quote against the original. And the thing you're citing may itself be wrong, or cite sources that don't exist.

The three styles resolved this differently, and their disagreements are instructive. APA decided the software's maker is the closest thing to an author. MLA decided nothing about an LLM deserves the author slot and organized the citation around what you did — the prompt. Chicago decided AI output is closest to a personal communication: real, citable, but not a recoverable source, so it lives in the notes and stays out of the bibliography. None of these is silly. All of them are workarounds, and each style says so with unusual candor.

Here's the quick-reference version, then each style in detail.

APA: the company is the author

APA's guidance comes from Timothy McAdoo's APA Style blog post "How to Cite ChatGPT" (first published April 7, 2023, since updated). The model: the organization that made the tool is the author, the model name is the title, the version goes in parentheses, and a bracketed description tells the reader what kind of thing this is.

Reference list template:

Company. (Year). Name of model (Version) [Large language model]. URL

APA's own published example, verbatim:

OpenAI. (2023). ChatGPT (Mar 14 version) [Large language model]. https://chat.openai.com/chat

In-text: parenthetical (OpenAI, 2023) or narrative OpenAI (2023).

Example entries for other tools, following the same pattern:

(The ChatGPT entry is APA's published example; the Claude and Gemini entries apply APA's stated pattern to those tools — swap in the version identifier and URL from your own session.)

Three pieces of APA guidance students routinely miss. Describe your prompt in the text itself — APA expects something like: When given the prompt "Is left-handedness heritable?", ChatGPT responded that… (OpenAI, 2023). Put long AI responses in an appendix, since your reader can't retrieve the original chat, and call out each appendix at least once in the body. And treat any sources the AI "cites" as radioactive until you've held them in your hands — models fabricate plausible-looking references, and citing a fabricated study is a much worse day than citing ChatGPT.

MLA: cite the prompt, never the "author"

The MLA Style Center's guidance ("How do I cite generative AI in MLA style?") starts from a firm principle: do not treat AI as an author. MLA instead runs AI output through its standard container system, with your prompt serving as the title of the source.

Works-cited template:

"Prompt you used" prompt. Name of Tool, version, Company, date generated, URL.

MLA's own published example, verbatim:

"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.

In-text: a shortened form of the title — for the entry above, ("Describe the symbolism").

MLA's scope rule is broader than most students expect: cite the tool "whenever you paraphrase, quote, or incorporate into your own work any content" it created — that includes AI-generated ideas you paraphrased, not just copied sentences. For functional uses that don't put AI content into your text — editing, translation, brainstorming — MLA says to acknowledge the use "in a note, your text, or another suitable location" rather than build a works-cited entry. That acknowledgment is what the disclosure templates later in this post are for.

And MLA repeats the warning that bears repeating: vet every source an AI names before you rely on it, because the tools sometimes invent them.

Chicago: a footnote, not a bibliography entry

The Chicago Manual of Style's online Q&A treats AI output like a personal communication — you cite it where you use it, but it doesn't earn a bibliography entry, because your reader can't go retrieve it. (Chicago's stated exception: if you can provide a publicly accessible link to the conversation, a formal entry becomes reasonable.)

Footnote, when the prompt already appears in your text:

1. Text generated by ChatGPT, OpenAI, March 7, 2023, https://chat.openai.com/chat.

Footnote, when you need to include the prompt:

1. ChatGPT, response to "Explain how to make pizza dough from common household ingredients," OpenAI, March 7, 2023.

Author-date style: put the details in parentheses — (ChatGPT, March 7, 2023).

Chicago adds one touch the other styles skip, and it's a good one: if you edited the AI's text before using it, say so — a note can end with "edited for style and content." Minor silent cleanup of formatting doesn't need a flag, but substantive editing does. That's the citation system quietly agreeing with a point we make throughout this site: the difference between honest AI-assisted work and misrepresentation is whether the reader knows what happened.

When the assignment just wants an acknowledgment

Plenty of instructors don't want formal citations for AI at all — they want a disclosure statement: a few sentences at the end of the paper saying what you used and for what. This is also the right instrument for uses that never put AI text into your essay (feedback, brainstorming, grammar checking), which is why MLA points such uses to "a note" rather than the works-cited list.

A good disclosure names the tool, the task, the stage, and what you did with the output. Templates you can adapt:

Brainstorming/outlining only:

I used ChatGPT (OpenAI) to brainstorm potential essay topics and to generate a rough outline, which I substantially reorganized. All research, drafting, and final wording are my own.

Feedback on a self-written draft:

After completing my draft, I used Claude (Anthropic) to critique its structure and identify unclear passages. I made all revisions myself; no AI-generated text appears in this essay.

Editing/language help:

I wrote this essay myself and used a generative AI tool to suggest grammar and clarity edits, accepting or rejecting each suggestion individually. As a non-native English speaker, I used it in place of a human proofreader.

Permitted drafting assistance:

As permitted by the course policy, I used ChatGPT (OpenAI, March 2026 version) to generate a first draft from my own outline. I then fact-checked, restructured, and rewrote the draft; prompts and original outputs are available on request.

No use (increasingly requested):

No generative AI tools were used at any stage of this assignment.

Keep the receipts behind every statement — prompt logs, chat links, version history. A disclosure you can document is protection; a disclosure you can't is just a claim. This matters more than it should, because AI detectors misfire on honest work, and a documented process is the strongest answer to a wrongly flagged score. Some students run drafts through an AI detector before submitting for exactly this reason — to see what a screening tool will claim about their writing in advance. Ours shows sentence-level results on the same probability logic Turnitin uses; it doesn't promise to predict any specific detector's verdict, because the vendors' models differ and nobody can honestly promise that.

When your style guide says nothing

Engineering courses with bespoke citation rules, IEEE, AMA, Bluebook, a professor's homegrown format — sooner or later you'll face an AI-use situation with no published rule. Fall back on the principles the big three styles agree on, because they're really one principle wearing three formats:

1. Say what tool, from what company, on what date. Those three facts appear in every style's format. Version information helps when you have it.
2. Preserve the prompt. MLA makes it the title; APA wants it in your text; Chicago puts it in the note. All three converged on the same insight — with a nondeterministic source, the prompt is the closest thing to a page number that exists.
3. Distinguish quoting from using. AI text that appears in your essay needs a citation at the point of use. AI help that shaped the essay without appearing in it needs an acknowledgment note. Don't let the second category go unmentioned just because there's no obvious place to put it.
4. Ask the instructor which format they want — and if they have no preference, borrow the closest major style's format and apply it consistently. Consistency plus honesty satisfies almost every grader; silence satisfies none of them.

When in doubt, over-disclose. In the entire history of academic integrity proceedings, "the student acknowledged too much" has never been the charge.

What citing AI does and doesn't do for you

Worth being precise here, because students sometimes treat a citation as a talisman.

A proper citation converts plagiarism questions into policy questions. Cambridge's rule — unacknowledged AI content in assessed work is misconduct — shows the mechanism: acknowledgment is exactly what separates a violation from permitted use. Under any disclose-and-cite policy, the citation is doing real protective work.

What a citation cannot do: make forbidden use permitted, or change what a detector reports. If your course bans generation, a beautifully formatted APA entry documents the violation rather than curing it — the honest treatment of that boundary is in our post on whether AI drafting is cheating. And Turnitin's AI indicator doesn't read your citations at all; it scores prose on statistical signals like perplexity and burstiness. Cited AI text will still raise the score, which is fine — expected, even — if your instructor knows the use was permitted and disclosed. This is why disclosure and citation travel together: the citation formats the record, the disclosure makes sure a human reads the score in context.

One last note on scale, because it explains why instructors have gotten pickier about this. Turnitin screened over 200 million papers in the detector's first year and flagged about 11% as containing substantial AI writing. Against that backdrop, a student who cites and discloses is doing something statistically unusual — and graders notice. Transparent AI use, in courses that allow it, reads as exactly what it is: someone following the rules in a semester when many people aren't.

FAQ

How do I cite ChatGPT in APA format? Reference list: OpenAI. (2023). ChatGPT (Mar 14 version) [Large language model]. https://chat.openai.com/chat — swapping in your session's version and date. In text, cite (OpenAI, 2023), describe the prompt you used in your prose, and put long AI responses in an appendix.

How do I cite ChatGPT in MLA format? Use your prompt as the title and never list AI as an author: "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. In text, use a shortened title: ("Describe the symbolism").

Does ChatGPT go in a Chicago-style bibliography? No — Chicago treats AI output like a personal communication: cite it in a numbered footnote (Text generated by ChatGPT, OpenAI, March 7, 2023), optionally including your prompt, but leave it out of the bibliography unless you can link to a publicly accessible version of the conversation.

Do I need to cite AI if I only used it for brainstorming or editing? You don't build a works-cited entry for it, but the major styles still expect acknowledgment — MLA says to note such uses "in a note, your text, or another suitable location." A one-to-two sentence disclosure statement at the end of the paper is the standard instrument, and some instructors now require one either way.

If I cite AI properly, can I still get in trouble? Yes, if the use itself wasn't allowed — citation documents use, it doesn't authorize it. Check your syllabus and our guide to university AI policies first. Under disclose-and-cite policies, though, proper citation is precisely what keeps permitted use from becoming misconduct.

Will citing AI stop Turnitin from flagging my paper? No. Turnitin's detector scores the statistical texture of prose; it doesn't parse citations or read disclosure statements. Cited, permitted AI text can and will raise an AI score — which is why the disclosure exists: so the human reading the score already knows why. See our Turnitin guide for how instructors are told to interpret those scores.

How do I cite Claude or Gemini instead of ChatGPT? Apply each style's pattern with the maker's name: APA — Anthropic. (2026). Claude (July 2026 version) [Large language model]. https://claude.ai; MLA — your prompt as title, Claude, version, Anthropic, date, URL; Chicago — Text generated by Claude, Anthropic, [date]. The structure carries over; only the proper nouns change.

What if my AI conversation has no shareable link? That's the normal case, and the styles have planned for it: APA has you reproduce the relevant output (appendix for long passages), MLA anchors the entry to your prompt, and Chicago keeps the citation in the notes precisely because the source isn't retrievable. Save your own copy of the conversation regardless — it's your evidence if questions come later.

Key facts

• APA's published example (APA Style blog, "How to Cite ChatGPT," Timothy McAdoo, first published April 7, 2023): OpenAI. (2023). ChatGPT (Mar 14 version) [Large language model]. https://chat.openai.com/chat; in-text (OpenAI, 2023).
• MLA's published example (MLA Style Center): "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 — and AI is never listed as an author.
• Chicago's published example (CMOS online Q&A): footnote "Text generated by ChatGPT, OpenAI, March 7, 2023, https://chat.openai.com/chat"; no bibliography entry unless a publicly accessible link exists; note edits with "edited for style and content."
• All three styles warn that AI tools fabricate sources — verify every reference an AI provides before using it.
• University of Cambridge policy makes unacknowledged AI content in summative assessment academic misconduct — acknowledgment/citation is the dividing line, not tool use itself.
• Turnitin's AI indicator (launched April 4, 2023) scored 200M+ papers in its first year, ~11% with ≥20% AI writing; it evaluates prose statistically and does not read citations or disclosure statements.

Sources

1. APA Style blog — "How to Cite ChatGPT," Timothy McAdoo (apastyle.apa.org, first published April 7, 2023)
2. MLA Style Center — "How do I cite generative AI in MLA style?" (style.mla.org)
3. Chicago Manual of Style Online — Q&A on citing ChatGPT and generative AI (chicagomanualofstyle.org)
4. University of Cambridge, Blended Learning Service — Generative AI and Assessment (blendedlearning.cam.ac.uk)
5. American Public University System — Generative AI Policy, Student Code of Conduct (apu.apus.edu)
6. Turnitin — AI Writing Detection FAQ; first-anniversary data release (April 2024)