To cite ChatGPT, treat it as a non-recoverable source created by OpenAI: list OpenAI as the author, give the year and the model version (for example, ChatGPT, GPT-4o), add the phrase you used to describe it, and finish with the URL https://chat.openai.com/. The exact wording changes with your referencing style, but every style needs those same four ingredients — author, date, descriptive title and source location.
This guide shows you exactly how to cite ChatGPT in the six styles UK and international students use most — APA, MLA, Harvard, Chicago, Vancouver and IEEE — with copy-ready reference-list entries and in-text examples for each. It also covers when a citation is required, how to log the prompt and full response in an appendix, and how to stay on the right side of your university’s academic-integrity rules. Always check your own department’s policy first, because some courses restrict or ban generative AI altogether.
Why and When You Need to Cite ChatGPT
Citing is about transparency. If any words, ideas, structure or data in your work came from a generative AI tool rather than from your own thinking, your reader is entitled to know. Failing to declare AI use — and passing its output off as your own — is a form of plagiarism, and most universities now treat undeclared AI use as an academic-misconduct offence. Honest attribution protects you: it shows the marker exactly where the AI ends and your scholarship begins.
A short rule of thumb: if you would cite a human or a website for the same contribution, cite ChatGPT too. You need a citation when you quote ChatGPT’s wording, paraphrase a substantive idea it produced, reproduce a table or list it generated, or used it to draft a passage you kept. You usually do not need a formal citation for trivial mechanical help (such as fixing a typo), but many institutions still ask you to acknowledge any AI use in a methods note or declaration. When in doubt, declare it — over-disclosure is never penalised the way concealment is.
There is a second reason ChatGPT is awkward to cite: its output is non-recoverable. Two people who type the same prompt will get different answers, and your reader cannot click a link and retrieve the exact text you saw. The major style guides handle this in one of two ways — either by treating ChatGPT as a personal communication (something the reader cannot access), or by asking you to reproduce the full prompt and response in an appendix so the work can still be evaluated. Knowing which approach your style takes is the key to citing it correctly.
How to Cite ChatGPT: The Four Core Elements
Whatever style you use, a ChatGPT citation is built from the same four pieces of information. Get these right and adapting them to any referencing system is just a matter of punctuation and order.
| Element | What to use | Example |
|---|---|---|
| Author | The company that made the model, not “ChatGPT” itself | OpenAI |
| Date | The year you generated the response (plus full date in some styles) | 2026 |
| Title / descriptor | The model and a short description of the content, often the prompt | ChatGPT (GPT-4o) response to “Explain Porter’s Five Forces” |
| Source / location | The model version label and the tool’s URL | OpenAI. https://chat.openai.com/ |
Two practical tips before you start. First, record the model version (for example GPT-3.5, GPT-4 or GPT-4o) — styles increasingly ask for it because the model directly affects the output. Second, save your transcript. Because the response cannot be retrieved later, your appendix copy of the prompt and answer becomes the evidence trail your marker may want to see.
How to Cite ChatGPT in APA (7th Edition)
APA — the dominant style in psychology, education and the social sciences — treats ChatGPT’s output as the result of an algorithm and asks you to reference the software. Per the official APA guidance, OpenAI is the author, the year you used the tool is the date, the model name is the title (italicised) with the version in brackets, and “Large language model” describes the source.
Reference list:
OpenAI. (2026). ChatGPT (GPT-4o version) [Large language model]. https://chat.openai.com/
In-text (paraphrase):
When prompted to summarise the causes of the 2008 financial crisis, ChatGPT produced a four-point overview (OpenAI, 2026).
In-text (direct quote):
OpenAI (2026) described the crisis as “a systemic failure of credit risk pricing.”
Appendix note: APA recommends you include the full prompt and the response in an appendix (or in the body if short), so the non-recoverable text can still be reviewed.
Note the deliberate shift from the older, incorrect habit of writing “(ChatGPT, personal communication)” — APA does not classify ChatGPT as a personal communication, because the author (OpenAI) is recoverable and you can name the model. Use the software-reference format above instead.
How to Cite ChatGPT in MLA (9th Edition)
MLA — standard in English, the humanities and many arts subjects — builds references from its “core elements” template. MLA recommends you describe the AI act in the Title of Source slot using the prompt, name ChatGPT as the container’s title, credit the version and the publisher (OpenAI), give the date you generated the text, and end with the URL.
Works Cited:
“Explain the impact of artificial intelligence on the UK job market” prompt. ChatGPT, GPT-4o version, OpenAI, 14 June 2026, chat.openai.com/.
In-text:
One generated overview noted three sectors most exposed to automation (“Explain the impact”).
MLA shortens the in-text reference to the first words of the prompt title in quotation marks, because there is no page number to cite.
How to Cite ChatGPT in Harvard Style
Harvard is not a single fixed standard — it is a family of author–date formats, and your university will have its own house variant, so always check your course handbook. The widely used Cite Them Right interpretation of Harvard style treats AI tools much like APA: author, year, the tool and version, a descriptor, the generating date and the URL.
Reference list:
OpenAI (2026) ChatGPT (GPT-4o) [Large language model]. Available at: https://chat.openai.com/ (Accessed: 14 June 2026).
In-text:
The model summarised three competing definitions of sustainability (OpenAI, 2026).
If your handbook prefers “ChatGPT” as the in-text author, follow it — consistency with your institution’s guide always wins over a generic template.
How to Cite ChatGPT in Chicago Style
The Chicago Manual of Style treats ChatGPT output as something the reader cannot access, so a full bibliography entry is optional — the essential acknowledgement goes in a note or in the text. You credit OpenAI, name the tool, describe the content (ideally the prompt), and give the date.
Footnote:
1. Text generated by ChatGPT, OpenAI, June 14, 2026, https://chat.openai.com/, in response to the prompt “Summarise the causes of inflation.”
In-text alternative (no note):
The following summary was produced by ChatGPT (OpenAI) on June 14, 2026, in response to the author’s prompt.
Chicago advises that you reveal the prompt within the text or note, since the response itself is not retrievable.
How to Cite ChatGPT in Vancouver Style
Vancouver — the numbered style used across medicine and many sciences — assigns each source a number in the order it first appears, shown in the text and matched to a numbered reference list. Because there is no agreed standard for AI yet, the safest approach is to follow your journal or faculty guidance and cite the tool as an online software source.
In-text:
Generative models can summarise large clinical datasets quickly (1).
Reference list:
1. OpenAI. ChatGPT (GPT-4o) [large language model]. San Francisco: OpenAI; 2026 [cited 2026 Jun 14]. Available from: https://chat.openai.com/
How to Cite ChatGPT in IEEE Style
IEEE — the engineering and computer-science standard — also uses bracketed numbers in the text linked to a numbered list. Cite ChatGPT as software, giving the author, the tool with version, a descriptor, the date and the URL.
In-text:
The generated pseudocode followed a standard divide-and-conquer pattern [1].
Reference list:
[1] OpenAI, “ChatGPT (GPT-4o),” Large language model, OpenAI, 2026. [Online]. Available: https://chat.openai.com/
Side-by-Side: ChatGPT Citation by Style
Use this comparison to choose the right pattern at a glance. The reference-list shape and the in-text mechanism differ, but the four core elements (author, date, descriptor, location) stay constant across every style.
| Style | In-text form | Reference-list shape | Treats output as |
|---|---|---|---|
| APA 7 | (OpenAI, 2026) | OpenAI. (2026). ChatGPT (version) [Large language model]. URL | Software / algorithm output |
| MLA 9 | (“Prompt first words”) | “Prompt” prompt. ChatGPT, version, OpenAI, date, URL. | Sourced container |
| Harvard | (OpenAI, 2026) | OpenAI (2026) ChatGPT (version) [LLM]. Available at: URL (Accessed: date). | Online software |
| Chicago | Note number / in-text mention | Text generated by ChatGPT, OpenAI, date, URL, prompt. | Non-recoverable personal-type source |
| Vancouver | (1) numbered | 1. OpenAI. ChatGPT (version) [LLM]. 2026 [cited date]. URL | Online software |
| IEEE | [1] bracketed | [1] OpenAI, “ChatGPT (version),” LLM, 2026. [Online]. Available: URL | Online software |
Which Style Should You Use?
You rarely get to choose — your discipline picks the style for you, and your module handbook is the final word. As a quick orientation: APA dominates psychology, education, nursing and the social sciences; MLA is standard in English literature, languages and the humanities; Harvard variants are common across UK business, social science and many undergraduate programmes; Chicago appears in history and some humanities; while Vancouver and IEEE are the numbered systems used in medicine and in engineering and computing respectively. If your brief names a style, follow it to the letter; if it does not, ask your tutor rather than guessing, because mixing systems is itself a marking penalty. Whichever you land on, the four core elements stay the same — only the order, punctuation and in-text mechanism change.
How to Document the Prompt and Response
Because ChatGPT output cannot be retrieved by your reader, most style guides and universities ask you to preserve a record. This is the single most useful habit for staying integrity-safe. Add an appendix that captures, for each AI interaction:
- The exact prompt(s) you typed, word for word
- The full response ChatGPT returned
- The model version (e.g. GPT-4o) and the date and time
- A one-line note on how you used the output — brainstormed, paraphrased, quoted, or rejected
Then point to the appendix in your method or declaration, for example: “The full prompts and responses are reproduced in Appendix B.” This turns an unverifiable source into something a marker can actually inspect, and it demonstrates exactly where your own analysis took over. For the underlying principles that apply to every source type, see our wider guide on how to cite sources properly.
“When you use ChatGPT or other AI tools in your research and writing, it is important to cite them appropriately. … We recommend that students attribute the use of large language models in their work.” — OpenAI / APA Style guidance on citing AI tools
Common ChatGPT Citation Mistakes to Avoid
Even careful students slip on the same few points. Watch for these errors, all of which can weaken your marks or trigger an integrity query:
- Listing “ChatGPT” as the author in the reference list instead of OpenAI (the maker).
- Omitting the model version — GPT-3.5 and GPT-4o can produce very different answers.
- Treating ChatGPT as a normal retrievable webpage with a stable URL to the exact answer (the answer is not retrievable).
- Quoting a “source” ChatGPT gave you without checking it exists — fabricated citations are a classic AI failure.
- Forgetting the institutional declaration or methods note, even when the in-text citation is perfect.
- Using AI output as evidence for a factual claim instead of citing the underlying primary source.
ChatGPT can sound authoritative even when it is wrong, so its fluent, human-like responses need the same scepticism you would apply to any unverified claim. If you are weighing up where AI fits in your workflow at all, our comparison of how AI drafting stacks up against the work of professional writers is a useful reality check.
Staying Within Your University’s AI Policy
Citation rules and permission rules are two different things. Citing ChatGPT correctly does not make its use automatically allowed — your department decides whether, and how, generative AI may be used on a given assessment. Policies fall into roughly three camps:
| Policy stance | What it means | Your action |
|---|---|---|
| Prohibited | No AI use permitted for the assessment | Do not use ChatGPT for content; declare nothing because there is nothing to declare |
| Permitted with attribution | AI allowed if disclosed and cited | Cite every use as shown above and add an AI-use declaration |
| Permitted for limited tasks | Allowed for, e.g., brainstorming or proofreading only | Stay inside the stated scope and acknowledge it |
Before submission, confirm your work reads as your own and reflects your understanding. Markers and integrity officers increasingly run submissions through an AI detector to flag undisclosed machine-written passages, so transparent citation is not just good etiquette — it is your protection. If you have used AI heavily for drafting and want a human to make the final work genuinely yours in voice and rigour, our comprehensive guide to proper AI attribution and our editorial team can help you do it the honest way.
Check Your Work Before You Submit
Run your draft through our free AI detector to see how much reads as machine-written — then cite and edit with confidence.
Citing ChatGPT well comes down to three habits: name OpenAI and the model version, follow your style’s pattern for the in-text and reference-list entry, and keep an appendix record of the prompt and response. Do those three things, declare your use where your institution requires it, and you can use AI tools transparently without putting your integrity — or your grade — at risk. If you want a final human polish, our Proofreading Services team can check your referencing and tighten your writing before submission.
