Quoting AI Output in an Academic Paper: Rules and Examples

You asked ChatGPT a question, it gave you a useful answer, and now you want to put some of that text in your paper. Citing it instead of pasting it in silently is the right instinct. The tricky part is that AI output doesn’t behave like a normal source, and the style guides only recently caught up.

This guide walks through the actual rules for APA, MLA, and Chicago, with concrete examples you can copy. It also covers the one thing most students miss: citing the AI correctly doesn’t mean you’re allowed to use it, and it doesn’t stop the text from reading like a machine wrote it.

Key takeaways

• APA, MLA, and Chicago all have specific rules now, and each treats a chatbot reply more like personal communication or software than a published article.
• Quote verbatim inside quotation marks when exact wording matters; paraphrase and attribute when you’re summarizing.
• Name the model and version (ChatGPT, GPT-4, OpenAI) and the date — AI replies aren’t reproducible and can’t be retrieved later.
• Citing AI is not the same as being permitted to use it. Check your course policy first.
• Even a properly cited block of raw AI text can still trip detectors and read as machine-written.

First: are you even allowed to use it?

Before you worry about commas and italics, answer the policy question. Some professors welcome AI as a brainstorming tool. Others ban it for specific assignments. A few ban it outright. Citing ChatGPT perfectly won’t help you if the syllabus says “no generative AI on this essay” — you’d just be documenting a violation in tidy formatting.

So step one is reading the assignment and the honor code, and asking if you’re unsure. Citation is about attribution. Permission is a separate thing. Don’t confuse the two.

Why AI output is awkward to cite

A journal article has an author, a title, a date, and a stable URL. Anyone can go find it and read the same words. AI output has none of that. The “author” is a statistical model. The same prompt can spit out a different answer five minutes later. There’s no public page someone else can visit to check what you saw.

Because of that, the style guides borrow from categories they already had — personal communication, software, unpublished material — instead of inventing something brand new. That’s why the three big formats handle it so differently.

How to quote ChatGPT in APA

APA treats the AI tool as software. The company that made the model is the author, the model name is the title (italicized), and you include the version date.

Reference list entry:

> OpenAI. (2026). *ChatGPT* (Jun 17 version) [Large language model]. https://chat.openai.com

In-text, direct quote:

> When prompted to define academic integrity, the model responded, “Academic integrity is the commitment to honesty and responsibility in scholarship” (OpenAI, 2026).

APA also gives advice worth following: because the response isn’t retrievable, drop the full prompt and the full response into an appendix or online supplemental materials. That way your reader can see exactly what you’re citing instead of taking your word for it.

If you’re paraphrasing rather than quoting, the same (OpenAI, 2026) citation applies — you just lose the quotation marks and put the idea in your own words.

How to quote ChatGPT in MLA

MLA goes the other way. It leaves the author blank and starts the entry with the prompt you used, treated as the title of the source. You then name the AI tool as the container, give the version and date, and end with the URL.

Works Cited entry:

> “Explain the causes of the French Revolution in two sentences” prompt. *ChatGPT*, GPT-4 version, OpenAI, 17 Jun. 2026, chat.openai.com.

In your text:

> One summary of the period’s tensions describes “a collision of fiscal crisis, food scarcity, and Enlightenment ideas about individual rights” (“Explain the causes”).

MLA’s reasoning is that you, the writer, produced the output by prompting it, so your prompt is the closest thing to a stable identifier. Keep your prompts specific and honest, because the prompt becomes part of the public-facing citation. MLA also leans on transparency over a fixed parenthetical format, so check whether your instructor wants the attribution in a note instead.

How to quote ChatGPT in Chicago

Chicago treats AI like personal communication — something not everyone can access. In most cases you cite it in a note or in your text rather than the bibliography, and you say who prompted it.

Footnote:

> 1. Text generated by ChatGPT, OpenAI, June 17, 2026, https://chat.openai.com.

If your prompt isn’t clear from the surrounding sentence, include it in the note, and note any edits you made to the output. Chicago doesn’t usually want a bibliography entry for AI text unless you can provide a publicly accessible link — another reason to check your instructor’s local rules.

Quote, paraphrase, or skip it: a mini-scenario

Say you’re writing a paper on study habits, and you asked an AI tool to list common procrastination triggers. It hands you a clean paragraph. You’ve got three honest moves:

1. Quote it. If the exact phrasing is the point — maybe you’re analyzing how the model frames the issue — put it in quotation marks and cite it. “The model described procrastination as ‘an emotional regulation problem disguised as a time-management one’ (OpenAI, 2026).”

1. Paraphrase it. If you just want the idea, restate it in your own voice and still attribute it. “When asked, the model framed procrastination less as poor time management and more as emotional self-regulation (OpenAI, 2026).”

1. Skip the AI entirely. Use the idea as a jumping-off point, then go find a real psychology source that says the same thing and cite that. For most academic work this is the strongest move, because a chatbot isn’t a peer-reviewed authority on anything.

Notice that quoting AI is rarely the best choice for a substantive claim. It’s most defensible when the AI output itself is the object of study — a writing class analyzing AI prose, a paper about the technology, or a methods section documenting how you used a tool.

The part most guides leave out: detection

Here’s the catch nobody warns you about. A citation is for human readers. An AI detector doesn’t read your citations or your quotation marks — it analyzes the statistical fingerprint of the words. Drop a long, unedited block of model output into your paper and that block still has the smooth, low-variation phrasing detectors associate with machines, citation or not.

Two technical things drive this. AI text tends to have low *perplexity* (the words are highly predictable to a language model) and low *burstiness* (sentence length and rhythm stay flat). Human writing jumps around more. If you want the underlying mechanics, here’s why a block of AI text still reads as machine-written.

This is why a properly cited AI quote can still push your paper’s overall AI score up, and detectors give you a probability, not proof. An instructor who sees a high score might not stop to notice the quotation marks. None of these tools are reliable, either — OpenAI retired its own AI-text classifier in July 2023 for low accuracy, and a 2023 study by Liang and colleagues found that detectors flag writing by non-native English speakers as AI-generated far more often than native writing. But you still don’t want to hand a flawed tool an easy reason to flag you. If you’re curious how a humanizer reworks that flat phrasing, you can see how PaperBleach handles it.

Practical takeaway: keep quoted AI blocks short, lead with and surround them with your own analysis, and read the result aloud so your voice clearly dominates. For more on disclosing and defending AI use, browse more guides for students and writers.

A quick checklist before you submit

• Confirm AI is allowed for this specific assignment.
• Match the citation to your required style — APA, MLA, and Chicago are not interchangeable.
• Name the model, version, and date every time.
• Keep verbatim AI quotes short and inside quotation marks.
• Save the prompt and full response in case you’re asked.
• Read the surrounding paragraphs aloud; your voice should dominate.

Frequently asked questions

Do I cite ChatGPT as the author? In APA, you list the company (OpenAI), not “ChatGPT,” and the model name becomes the title. MLA leaves the author blank and starts with your prompt. So it depends on your required style.

Can I quote ChatGPT word for word in my paper? Yes, if AI use is permitted. Put the text in quotation marks and attribute it. The real risk is using AI when it’s banned, or quoting so much that the paper stops being yours.

How do I cite an AI response nobody else can see? Treat it as non-retrievable. APA suggests putting the full prompt and response in an appendix. Saving a transcript or screenshot for your own records is wise too.

Will quoting AI make my paper get flagged by a detector? It can. Detectors analyze word patterns, not citations, so unedited model output still carries machine-like phrasing that can raise your score.

Is citing AI enough to stay out of trouble? No. Citation handles attribution, not permission. If your policy bans AI, citing it correctly won’t save you.

Closing

Quoting AI in a paper isn’t complicated once you separate the three jobs: get permission, format the citation for your style, and make sure the text still reads like you wrote it. The first two are rule-following. The third is where students slip. If AI touched your draft anywhere, run it through PaperBleach before you submit so you can see — and fix — what a detector would flag.