Loading...
Cite ChatGPT, Claude, Gemini, Copilot and other AI tools correctly in APA 7, MLA 9, Chicago 17, Harvard, IEEE and Vancouver. Pick the tool, add the version, date and prompt, and copy the reference entry with every in-text form, built on the guidance the style manuals published for generative AI. Ten model families are preset; any other tool takes a few keystrokes.
The version is in the product's model picker or settings. The prompt makes MLA, Chicago and Harvard entries complete; APA works without it.





Trusted by researchers publishing in
Open the model picker or settings in your AI tool and note the exact version you used. AI tools change weekly, so the version is what makes your citation reproducible, the same way an edition dates a book.
Record when the text was generated and what you asked. MLA and Chicago build the citation around the prompt itself, and every style dates the exchange somewhere.
Pick APA, MLA, Chicago, Harvard, IEEE or Vancouver and copy the entry with italics intact, plus the in-text forms and the note or disclosure wording the style expects.
APA, MLA and Chicago published real guidance for citing AI in 2023, and each output here follows it: company as author for APA, prompt as title for MLA, notes-only for Chicago.
ChatGPT, Claude, Gemini, Meta AI, Copilot, Grok, DeepSeek, Le Chat, Perplexity and Qwen come with company and URL filled in. You add only what is yours: the version, date and prompt.
Chicago wants no bibliography entry, MLA wants no author, Harvard wants an appendix and IEEE wants an acknowledgment. The generator changes shape per style instead of forcing one template on seven rulebooks.
Copy puts rich and plain text on the clipboard, so model names arrive in Word or Google Docs already italicized where the style wants them.
For a year after ChatGPT's release, citing it was guesswork. Then the manuals answered, and they answered differently. APA reasoned that the output of an algorithm belongs to the company that made it, so OpenAI or Anthropic takes the author position, with the model italicized as the title and the version in parentheses. MLA reached the opposite conclusion: an AI cannot be an author at all, so the entry is built around your prompt, with the tool as the container it appeared in. Chicago looked at the same problem and filed it under personal communication, cited in a footnote or in the text and never in the bibliography.
What all three agree on is the details that make the citation reproducible: which version, and when. A model's answer to the same prompt differs between releases and even between days, which is why every pattern here dates the exchange and why the version field matters more than it does for any book. It is also why the guides point readers at the tool's URL rather than a chat link, and suggest keeping the full transcript in an appendix.
One line has hardened across academic publishing: an AI model is never an author. The ICMJE and COPE both hold that AI cannot meet authorship criteria because it cannot take responsibility for the work, and publishers including Nature and Science have written the same rule into their policies. In practice that means two obligations that travel together: cite AI output where it appears in your text, and disclose AI assistance where your journal or university asks for it, usually in the methods or acknowledgments.
The other half of AI citation hygiene concerns the sources a model recommends. Language models invent plausible references, so any paper an AI suggests needs verifying against a real database before it enters your reference list; our hallucinated-citation audit walks through the check, and the main citation generator formats real records straight from Crossref. For how we think about responsible AI use in academic writing more broadly, see our AI use policy.
One ChatGPT conversation, taken apart in six styles. Hover any coloured part of the reference entry or the in-text variations to see what it is and the rule behind it.
OpenAI. (2026). ChatGPT (GPT-5 version) [Large language model]. https://chatgpt.com/
The pattern the APA Style blog published for ChatGPT, applied to any AI model. Alphabetize under the company's name in the reference list.
(OpenAI, 2026)
OpenAI (2026)...
When prompted, ChatGPT generated... (OpenAI, 2026)
APA recommends describing the prompt in the text or method section; a full transcript can go in an appendix.
(OpenAI, 2026)
AI output has no page numbers, so no p. locator is possible; make clear in the sentence that the words are the model's.
Hover or tap any coloured part for what it is and the rule behind it.
Seven outputs, four philosophies. These are the decisions each manual made, and the generator applies.
OpenAI, Anthropic or Google takes the author position, credited with the algorithm's output. The model is italicized as the title, the version sits in parentheses, and a bracketed [Large language model] tells readers what the source is.
MLA declines to treat the AI as an author. The entry opens with your prompt in quotation marks, the tool becomes the italicized container, and the in-text citation is a shortened form of that prompt.
The Chicago Manual treats AI output like personal communication: a numbered footnote carries the model, prompt, company and full date, and nothing appears in the bibliography, in either Chicago system.
Cite Them Right files generative AI under personal communications because your reader cannot retrieve the exchange. The maker and tool are the originator, and the transcript belongs in an appendix.
IEEE has no official AI template, so the software pattern stands in. What IEEE does mandate is disclosure: AI-generated content must be declared in the acknowledgments, alongside any reference.
Medical journals follow ICMJE, which requires AI use to be described in the methods or acknowledgments and bars AI from authorship. The reference itself adapts NLM's software pattern.
Each preset carries the real company and URL. Shown in APA 7; the generator formats every one in all seven outputs.
OpenAI
OpenAI. (2026). ChatGPT (GPT-5 version) [Large language model]. https://chatgpt.com/
Anthropic
Anthropic. (2026). Claude (Sonnet 4.5 version) [Large language model]. https://claude.ai/
Google. (2026). Gemini (2.5 Pro version) [Large language model]. https://gemini.google.com/
Meta
Meta. (2026). Meta AI (Llama 4 version) [Large language model]. https://www.meta.ai/
Microsoft
Microsoft. (2026). Copilot (GPT-5 version) [Large language model]. https://copilot.microsoft.com/
xAI
xAI. (2026). Grok (Grok 4 version) [Large language model]. https://grok.com/
DeepSeek
DeepSeek. (2026). DeepSeek (DeepSeek-V3 version) [Large language model]. https://chat.deepseek.com/
Mistral AI
Mistral AI. (2026). Le Chat (Mistral Large version) [Large language model]. https://chat.mistral.ai/
Perplexity AI
Perplexity AI. (2026). Perplexity (Sonar version) [Large language model]. https://www.perplexity.ai/
Alibaba Cloud
Alibaba Cloud. (2026). Qwen (Qwen3 version) [Large language model]. https://chat.qwen.ai/
APA 7 shown; the generator formats each tool in all seven outputs. The version shown is a current example, so swap in the one you used.
The same ChatGPT exchange formatted for every style the generator supports. Note where the prompt appears, where it does not, and which styles keep AI out of the reference list entirely.
Reference list entry
OpenAI. (2026). ChatGPT (GPT-5 version) [Large language model]. https://chatgpt.com/
Works Cited entry
"Explain the difference between correlation and causation for a first-year statistics student" prompt. ChatGPT, GPT-5 version, OpenAI, 14 July 2026, chatgpt.com.
Footnote (Chicago cites AI in notes only)
ChatGPT, response to "Explain the difference between correlation and causation for a first-year statistics student," OpenAI, 14 July 2026.
In-text citation (no reference list entry)
(ChatGPT, 14 July 2026)
Reference list entry
OpenAI ChatGPT (2026) ChatGPT response to prompt 'Explain the difference between correlation and causation for a first-year statistics student', 14 July.
Reference list entry
[1] OpenAI, ChatGPT (GPT-5 version) [Large language model], 2026. [Online]. Available: https://chatgpt.com/ (accessed 14 July 2026).
Reference list entry
1. ChatGPT (GPT-5 version) [Large language model]. OpenAI; 2026 [cited 14 July 2026]. Available from: https://chatgpt.com/
All seven come from the generator above. The prompt and dates are examples; substitute the details of your own conversation.
ProofreaderPro proofreads your full manuscript with tracked changes and never touches your citations: grammar, clarity and academic register, reviewable line by line. Built for researchers, free to try.
Proofread my writing free