Citation guidance assumes the use was permitted. If your syllabus, assignment sheet, or institutional policy prohibits generative AI on a given task, citing the model does not convert prohibited use into compliant use. The first question is always whether the tool was allowed for that specific assignment, not how to format the reference.
The short answer
APA 7th treats ChatGPT output as a software reference with author (OpenAI), year, model name and version, and a retrieval URL. MLA 9th treats it as a generative AI source and asks you to include the prompt, the tool, the version, the company, and the date. Chicago 17th treats it as software and recommends a footnote with the model, version, company, date, and the prompt used. In all three styles, include the date and the exact version, and disclose what the AI did for you.
The longer version covers when citation is appropriate in the first place, the exact format each style now uses, and what your disclosure note should contain.
When AI Should Be Cited (and When It Should Not Be Used at All)
Where AI use is permitted, the cases that typically require citation include any passage that paraphrases or quotes model output, any structural element (an outline, a thesis suggestion) that the model proposed and you adopted, and any factual claim that originated with the model rather than with a source you verified yourself. Harvard Business Publishing's editorial guidance on citing AI in your writing recommends disclosure even when the model's contribution is indirect.
Brainstorming a topic, summarizing your own draft back to you, or generating a list of possible counterarguments that you then rewrite are the use cases citation was designed for. Submitting model output as your own prose is not a citation problem; it is an authorship problem that no reference list can solve.
A citation documents permitted use. It does not convert prohibited use into compliant use.
Working Educators editorial
APA 7th Edition Format with Example
The American Psychological Association published direct guidance on how to cite ChatGPT and other large language models. APA treats the model as the author of the algorithm's output, with OpenAI, Anthropic, or the relevant company listed as the author element. The reference list entry uses the format: Author. (Year). Model name (Version) [Large language model]. URL.
A working example for ChatGPT: OpenAI. (2026). ChatGPT (June 4 version) [Large language model]. https://chat.openai.com. For Claude, the parallel entry: Anthropic. (2026). Claude (Sonnet 4.5) [Large language model]. https://claude.ai. The in-text citation follows standard APA author-date form: (OpenAI, 2026) or (Anthropic, 2026).
APA recommends that the methods or notes section of your paper describe the prompt you used and how you used the response. The reference list entry identifies the tool; the prose explains the interaction.
MLA 9th Edition Format with Example
The Modern Language Association's guidance on citing generative AI treats the tool as a source and structures the entry around the prompt. The MLA template uses: Prompt text prompt. Tool name, version, Company, Date, URL.
A working example: Summarize the main arguments in Federalist No. 10 prompt. ChatGPT, 4 June 2026 version, OpenAI, 4 June 2026, chat.openai.com. The in-text citation uses a shortened form of the prompt.
MLA's choice to put the prompt in the title position is deliberate. It signals that what makes an AI source citable is not the tool alone but the specific exchange between user and tool.
Chicago 17th Edition Format with Example
The Chicago Manual of Style's FAQ on citing generative AI treats AI tools as software and recommends documenting the use in a footnote rather than a bibliography entry.
A Chicago footnote example: 1. ChatGPT, response to Summarize the main arguments in Federalist No. 10, OpenAI, June 4, 2026, https://chat.openai.com.
Chicago's editors note that the bibliography entry is optional and that some publishers may require it; check your specific venue. For dissertations and theses, follow your institution's library guidance.
What to Include in the Disclosure
Across all three styles, the underlying expectation is the same: a reader should be able to tell what the model did, when it did it, and how. A complete disclosure includes the model name, the version or release date, the company, the date you accessed it, and the prompt or prompts you used.
Disclosure also means describing the task. I used ChatGPT to generate an initial outline, which I then revised is a different claim than I used ChatGPT to draft the introduction. Both may be permitted depending on the assignment; both should be stated plainly.
For teachers writing policy or designing assignments, our guide to AI policies faculty will actually follow covers how to specify permitted use in language students can act on. For students who want to document their writing process more thoroughly, tools like Proofademic's writing-process timeline can supplement a disclosure note with a record of how the draft developed.
Frequently Asked Questions
Do I need to cite ChatGPT if I only used it for brainstorming?▼
If the brainstorm produced ideas you used in the final paper, most style guides recommend a disclosure note even if a formal citation is not strictly required. When in doubt, disclose.
What if my professor has not addressed AI citation in the syllabus?▼
Ask before you submit. If you cannot reach the instructor, default to disclosure: include a methods note describing the tool, version, and use, and follow your discipline's dominant style guide.
Do I cite the same way for Claude, Gemini, or other models?▼
Yes. The template is identical; you swap the model name, company (Anthropic, Google), and access URL. APA, MLA, and Chicago all use the same structure regardless of which generative AI tool produced the output.
How do I cite an AI tool if I do not have a stable URL?▼
All three style guides accept the tool's home URL (chat.openai.com, claude.ai) when individual conversations are not shareable. The date and version are what allow a reader to understand which iteration of the model you used.
Does citing AI protect me from an academic integrity violation?▼
No. Citation documents use that was already permitted. If the assignment prohibited AI, citing the tool does not retroactively authorize the use.
The Bottom Line
The mechanics of citing AI are settled. APA treats it as a software reference. MLA treats it as a generative AI source organized around the prompt. Chicago treats it as software documented in a footnote. In all three, the load-bearing details are the model name, the version, the date, and the prompt.
The harder question, and the one no style guide can answer for you, is whether the use was permitted in the first place. Citation is a transparency tool, not a permission slip.
