How to Cite ChatGPT in APA, MLA, and Chicago Style (2026 Update)

Your professor said you can use ChatGPT, but only if you cite it. Great. Now you’re staring at the citation box and realizing nobody ever taught you how to cite a chatbot, and every blog post you find contradicts the last one. Here’s the part that matters: APA, MLA, and Chicago have all weighed in, the formats really are different, and getting them right takes about two minutes once you see the pattern.

Key takeaways

• All three major style guides now treat ChatGPT as a citable source, but each formats it differently, and none of them treats the AI like a human author.
• APA names the company (OpenAI) as author. MLA puts your prompt front and center. Chicago usually keeps the whole thing in a note instead of the bibliography.
• Record the exact date you generated the text. AI output isn’t reproducible, so the date and model version are your only stable reference points.
• Citing is about transparency, not format alone. Check your instructor’s or journal’s AI policy first; some ban AI entirely, others want a disclosure statement.
• A citation tells the reader the words came from a tool. It does nothing to make the writing sound like you, and it won’t change how an AI detector scores your text.

First, a reality check on what a citation does (and doesn’t)

A citation is an honesty move. It tells your reader: this sentence, this idea, this draft of an argument came from a tool, not from my own head. That’s it.

What a citation does not do is launder the text. Paste raw ChatGPT output into a paper, slap a citation on it, and hand it in, and you’ve been honest about the source, sure, but the writing is still the model’s writing. Some instructors are fine with that as long as you’ve disclosed it. Others expect you to use the AI as a starting point and make the words your own. Read the assignment. When the rules are vague, ask.

One more thing, because it trips people up constantly: a citation has nothing to do with AI detectors. Detectors don’t read your reference list. They score the statistical patterns in the prose itself. So citing ChatGPT won’t lower your detector score, and skipping the citation won’t raise it. Those are two separate problems, and we’ll come back to the second one at the end.

How to cite ChatGPT in APA style

APA (7th edition, per the American Psychological Association’s official guidance) treats the AI tool like software. The company that built the model is the author. The model name and version is the title.

Reference list format:

> OpenAI. (2026). *ChatGPT* (Mar 14 version) [Large language model]. https://chat.openai.com/chat

Break that down: the author is OpenAI, not “ChatGPT.” The year is when you used it. The italicized title is the model name, with the version in parentheses. The bracketed description tells the reader what kind of thing this is. Then the URL.

In-text citation:

> (OpenAI, 2026)

Again, that’s *OpenAI*, not ChatGPT, in the parenthetical. It trips up almost everyone the first time.

APA also recommends something a lot of students skip. Because the output isn’t reproducible, you’re supposed to include the full prompt and the response in an appendix, or describe how you used the tool in your methods. A reader can’t go look up your exact conversation, so you show your work instead.

Mini-scenario. You’re writing a psych paper and you asked ChatGPT for everyday examples of cognitive dissonance. You might write: *When prompted for everyday examples of cognitive dissonance, the model returned three scenarios (OpenAI, 2026; see Appendix A).* The appendix holds the actual prompt and reply.

How to cite ChatGPT in MLA style

MLA’s approach, from the MLA Style Center, is built around the Works Cited “template of core elements.” It puts your prompt in the spotlight, because MLA treats a description of the prompt as the title of the source.

Works Cited format:

> “Examples of cognitive dissonance in daily life” prompt. *ChatGPT*, 14 Mar. version, OpenAI, 14 Mar. 2026, chat.openai.com/chat.

The “Title of Source” slot is a description of your prompt, in quotation marks. *ChatGPT* is the title of the container. OpenAI is the publisher. Then the version, the date you generated it, and the URL.

In-text citation:

MLA usually points to the title, shortened. Since the title here is your prompt, you’d use a short version of it:

> (“Examples of cognitive dissonance”)

MLA is explicit that you should not list the AI tool as an author, because it can’t take responsibility for what it produces. You name the tool as the container title instead. MLA also tells you to vet anything the model hands you, since it can invent sources, and to cite the original work directly if the AI pointed you to a real one.

How to cite ChatGPT in Chicago style

Chicago (per The Chicago Manual of Style) handles this differently from the other two. Its general recommendation is to cite AI tools in a footnote or endnote, and in most cases *not* to list them in your bibliography. The reasoning: the content isn’t recoverable by your reader, so it behaves more like a personal communication.

Note format:

> 1. Text generated by ChatGPT, OpenAI, March 14, 2026, chat.openai.com/chat.

You name the tool, the company, the date, and add the URL. If you’ve already named ChatGPT and described your prompt in the body text, Chicago says the note can be shorter, since you’re not repeating what’s already there. And if you edited the output, say so in the note.

In the text itself, Chicago is happy to let you fold the attribution into a sentence: *In response to my prompt, ChatGPT suggested…* followed by the note number. That’s often cleaner than a parenthetical.

If your instructor or publisher *does* want a bibliography entry, you can adapt the note format, but check first. The default Chicago position is notes only.

A side-by-side cheat sheet

Same scenario, all three styles, so you can see how they diverge.

• APA in-text: (OpenAI, 2026)
• MLA in-text: (“Examples of cognitive dissonance”)
• Chicago: a numbered footnote naming ChatGPT, OpenAI, and the date

The pattern underneath all three: name the maker, name or describe the prompt, mark the version, and pin the date. Learn that and you can rebuild any of these from scratch.

What’s actually new in 2026

The big shift over the last couple of years isn’t the citation mechanics. It’s policy. Style guides settled on formats fairly quickly. Institutions are still arguing about whether AI use is allowed at all.

So the citation question now sits downstream of a bigger one: *am I even allowed to use this?* A growing number of syllabi ask for a short AI use statement, separate from any citation, where you describe in plain language what you used the tool for. Many academic journals have added similar disclosure requirements, and most won’t let you list an AI as a co-author. When your course or publication has a policy like this, the disclosure statement matters more than the citation format, and skipping it is the thing that actually gets people in trouble.

Worth knowing too: the same logic carries over to every other AI writing tool. Cite Claude with Anthropic as the developer. Cite Gemini with Google. The structure stays the same; only the names change.

After you cite: make sure the writing is yours

Say you’ve cited everything perfectly and your instructor allows AI assistance. You’re still not done, because a citation doesn’t change how the prose reads. Raw model output tends to sound flat and oddly even, and a careful reader can feel it. So can a detector, which, as we said up top, never even looks at your reference list. Here’s why that detector flags the patterns it flags if you want the underlying mechanics.

If the goal is work that genuinely sounds like you, the citation is step one and revision is step two. Rewrite the ideas in your own sentences. Drop in a real example from your own life or reading. Vary your rhythm. Cut the parts that read like a press release. That’s the honest version of using AI well, and you can browse the rest of our student guides for more on getting there.

When you’ve got AI-drafted text that you’re allowed to use but it still reads like a robot wrote it, revise it into your own voice before you submit. Cite the source, then make the writing yours.