To reference ChatGPT, first check your unit’s policy on whether AI use is permitted and must be disclosed, then cite it as software. In APA 7, the reference is: OpenAI. (Year). ChatGPT (version) [Large language model]. URL, with an in-text citation of (OpenAI, Year). In Harvard, use an author-date entry: OpenAI Year, ChatGPT (version), large language model, accessed Day Month Year, URL.
Generative AI tools like ChatGPT are now part of how many students draft, brainstorm and revise. But citing them correctly is trickier than citing a journal article, because a chatbot has no stable author, no page numbers, and an output nobody else can reproduce. Getting it wrong can read as an integrity breach rather than honest disclosure. This guide explains how to reference ChatGPT and other AI tools in both APA 7 and Harvard, the Australian academic-integrity context you must respect, and the common mistakes that get students into trouble. If you are still weighing up whether AI belongs in your work at all, our overview of whether you can use AI for university assignments is a useful companion read. If you would rather hand the writing to a human, a professional writing service can deliver original, AI-free work checked through Turnitin.
Key takeaways
- Policy first: Always check your unit guide and your university’s academic-integrity rules before using or citing AI; some units ban it outright.
- APA 7 treats ChatGPT as software: OpenAI as author, year, ChatGPT plus version, “[Large language model]” descriptor, and the URL.
- Harvard uses author-date: OpenAI, year, model and version, an accessed date, and the URL, with an in-text (OpenAI Year).
- Disclose your method: Describe how you used AI in a methods note or appendix, not just in the reference list.
- Never treat AI as a source of truth: ChatGPT can fabricate facts and invent references, so verify everything against real, citable sources.
Step one: check your unit’s policy before you cite anything
Before you worry about formatting, confirm that using ChatGPT is allowed in the first place. In Australia, academic integrity is overseen by TEQSA (the Tertiary Education Quality and Standards Agency), which has published specific guidance on generative AI and assessment. TEQSA’s position is that universities must protect the integrity of their assessments, and each institution translates that into its own rules. That means the answer to “can I use AI?” is set by your university and, more precisely, by your individual unit or course.
Policies vary widely. Some units encourage AI for brainstorming but forbid it for the final text. Others permit limited use only if you disclose it. A growing number ban it entirely for certain assessments. For example, a Monash student in a coding unit might be allowed to use ChatGPT to explain an error message, while the same student’s law unit prohibits any AI involvement in a written submission. Using AI where it is banned, even with a perfect citation, is still a breach. If you are unsure how examiners even spot AI involvement, our guide on whether AI can be detected in an assignment explains what the tools actually flag.
So your real first step is to read three things: the unit guide, the assessment instructions, and your university’s academic-integrity policy. If anything is ambiguous, email your unit coordinator and keep the reply. Honest disclosure protects you; silence does not. It also helps to know how markers verify suspicions, which our walkthrough on how to check if an assignment is AI-written sets out in plain terms.
How to reference ChatGPT in APA 7
APA 7 does not have a bespoke “chatbot” category, so the American Psychological Association advises treating ChatGPT output like software. The company that made the tool, OpenAI, is the author. The year is the version year you used. The title is the tool name plus the version, followed by a bracketed descriptor and the URL.
APA 7 reference list format
The structure is: OpenAI. (Year). ChatGPT (Month Day version) [Large language model]. URL
A worked example looks like this: OpenAI. (2025). ChatGPT (June 25 version) [Large language model]. https://chat.openai.com. The “[Large language model]” descriptor tells your reader what kind of source this is, and the version detail signals that the model can change over time. If your law or criminology unit uses a footnote style instead, our AGLC4 referencing guide shows how to slot non-standard sources into that system.
APA 7 in-text citation
In the body of your essay, cite it as (OpenAI, 2025). If you quote or paraphrase a specific exchange, mention the prompt in your text rather than inside the citation, because there are no page numbers. For example: “When prompted to summarise the causes of federation, ChatGPT produced a three-point overview (OpenAI, 2025).” APA also suggests placing the full prompt and the AI’s response in an appendix if the exchange is central to your argument, since the output is not retrievable by your marker. Our full APA 7 referencing guide walks through software and unusual-source citations in more detail.
How to reference ChatGPT in Harvard
Harvard is a style family rather than a single fixed standard, so the exact punctuation can differ between universities (UTS Harvard, Swinburne Harvard, and so on). The shared logic is author-date: identify the author, the year, the title, and where the source can be found, plus an accessed date because online content changes.
Harvard reference list format
A widely accepted Harvard structure is: OpenAI Year, ChatGPT (version), large language model, accessed Day Month Year, URL.
For example: OpenAI 2025, ChatGPT (June 25 version), large language model, accessed 25 June 2025, https://chat.openai.com. The accessed date matters more here than in many other Harvard entries, because a later version of the model could return a different answer to the same prompt. Always confirm your school’s exact Harvard punctuation in its library guide; our Harvard referencing guide covers the common Australian variants.
Harvard in-text citation
In Harvard, cite in text as (OpenAI 2025). As with APA, describe the prompt in your prose so your reader understands what produced the output. If your unit uses a numbered or footnote style instead, follow that convention, but the same disclosure principle applies.
Side-by-side examples
The table below sets the two styles next to each other so you can copy the right pattern for your unit. Replace the version and date with the ones that match your own session. Once your references are sorted, it is worth checking how the AI text itself reads, since our explainer on whether a Turnitin similarity score is acceptable shows where examiners draw the line.
| Element | APA 7 | Harvard (author-date) |
|---|---|---|
| In-text | (OpenAI, 2025) | (OpenAI 2025) |
| Reference list | OpenAI. (2025). ChatGPT (June 25 version) [Large language model]. https://chat.openai.com | OpenAI 2025, ChatGPT (June 25 version), large language model, accessed 25 June 2025, https://chat.openai.com |
| Author | OpenAI | OpenAI |
| Descriptor | [Large language model] | large language model |
| Accessed date | Not required | Required |
| Where to note the prompt | In text or appendix | In text or appendix |
Worried your AI-assisted draft will not pass an integrity check? Our Australian writers produce original, AI-free, Turnitin-checked work written to your rubric, with every source properly referenced.
How to describe your AI use in a methods note or appendix
A reference list entry tells your marker that AI exists in your work, but it does not explain how you used it. Most Australian universities that permit AI now expect a short disclosure statement as well. This is where you stay transparent and demonstrate that your thinking, not the chatbot’s, drove the work. The note usually sits in your methods section or an appendix, so it helps to be confident about university essay structure before you decide where it belongs.
Keep the note specific. State which tool and version you used, what you used it for, and what you did not use it for. A strong methods note might read: “ChatGPT (June 25 version) was used to generate initial brainstorming questions and to suggest synonyms for clarity. All arguments, sources and final wording are my own, and every factual claim was verified against peer-reviewed literature.” If your prompts and the AI’s replies are extensive, place them in an appendix and refer to it in your text. This kind of transparent, source-led approach matters most in longer pieces, which is why our guide on how to write a literature review stresses verifying every claim against the original scholarship.
Disclose exactly what the tool did and what it did not do; a precise methods note is your best defence in an integrity review.
BAO academic support team
What NOT to do when citing AI
Citing ChatGPT correctly is as much about avoiding errors as following a format. The most damaging mistakes treat a language model as if it were a scholarly authority. This matters even more if you lean on a tool marketed as one of the ChatGPT essay writers, since a polished draft can hide unverified claims. Keep these rules in mind.
- Do not cite AI as a peer-reviewed source. ChatGPT is not a journal, a textbook or an expert. Cite it as a tool you used, never as evidence for an academic claim.
- Do not trust its facts. Verify every statistic, date and quotation against a real, citable source before you include it.
- Do not trust its references. AI can hallucinate convincing but non-existent citations, complete with fake DOIs. Markers check these, so confirm each reference actually exists.
- Do not hide your use. If your unit allows AI but requires disclosure, omitting the note is itself a breach.
- Do not paste AI text as your own. Even with a citation, submitting large blocks of unedited AI output usually breaks the rules and reads as your own work, which it is not.
For example, a UNSW student who once submitted an essay built on five “sources” suggested by ChatGPT found that three of them did not exist when the marker searched the library catalogue. The penalty was not for using AI; it was for failing to verify. If you are anxious about how detection tools read AI-assisted writing, our explainer on whether Turnitin detects AI sets out what these systems can and cannot see.
A safe, honest workflow for AI-assisted writing
If your unit permits AI, you can use it without risking your grade by treating it as an assistant, not an author. Brainstorm with it, ask it to explain a concept in simpler terms, or use it to check your own paragraph for clarity. Then do the scholarly work yourself: read the real sources, build the argument, and write the prose in your own voice. If you want a clearer sense of where automated drafting helps and where it puts your grade at risk, our piece on whether how to write an assignment without AI detection is a frank look at the trade-offs.
When you are ready, run a final pass. Confirm every fact against a citable source, rebuild any reference the AI suggested from the original, add your APA or Harvard entry for ChatGPT, and write your methods note. That sequence keeps your work both defensible and genuinely yours. If the workload is too much, handing the writing to a human is a legitimate alternative, and you can read about the difference in our comparison of an AI-written assignment versus a human-writer assignment.
Frequently asked questions
How do I reference ChatGPT in APA 7?
To reference ChatGPT in APA 7, treat it as software with OpenAI as the author. The reference list entry is OpenAI. (Year). ChatGPT (version) [Large language model]. URL, and the in-text citation is (OpenAI, Year). Describe your prompt in the text and place long exchanges in an appendix, since AI output has no page numbers and cannot be retrieved by your marker.
How do I reference ChatGPT in Harvard style?
To reference ChatGPT in Harvard style, use an author-date entry: OpenAI Year, ChatGPT (version), large language model, accessed Day Month Year, URL. The in-text citation is (OpenAI Year). Because Harvard varies between Australian universities, confirm the exact punctuation in your library guide before you submit.
Is it allowed to use ChatGPT in Australian universities?
Whether it is allowed to use ChatGPT in Australian universities depends entirely on your specific unit and your institution’s academic-integrity policy, which sits under TEQSA’s national framework. Some units permit AI with disclosure, others ban it for certain assessments. Always read your unit guide and assessment instructions, and ask your coordinator if anything is unclear.
Do I need to disclose that I used AI?
You usually need to disclose that you used AI whenever your unit permits it, and failing to disclose is itself an integrity breach in most policies. Add a short methods note stating which tool and version you used, what you used it for, and what remained entirely your own work. Keep any approval emails from your coordinator as a record.
Can ChatGPT give me real references?
ChatGPT cannot reliably give you real references, because it can hallucinate plausible but non-existent citations, including fake authors and DOIs. Never paste an AI-suggested reference without checking that the source genuinely exists in a library database or on the publisher’s site. Rebuild every confirmed reference from the original source rather than the chatbot’s version.
Should I cite ChatGPT as a source of information?
You should not cite ChatGPT as a source of information or evidence, because it is a tool, not a peer-reviewed authority. Cite it only to acknowledge that you used it, and support every factual claim with a real, citable academic source. Treating AI output as proof of a point is one of the most common and most penalised mistakes.
What is the difference between citing ChatGPT in APA and Harvard?
The difference between citing ChatGPT in APA and Harvard is mainly punctuation and the accessed date. APA 7 uses OpenAI. (Year). ChatGPT (version) [Large language model]. URL with no accessed date, while Harvard uses OpenAI Year, ChatGPT (version), large language model, accessed Day Month Year, URL. Both name OpenAI as the author and both expect you to describe your prompt in the text.
Where can I get help writing an essay without using AI?
You can get help writing an essay without using AI from a human writing service, where Australian Masters- and PhD-qualified writers produce original work to your rubric. Every assignment is written from scratch, checked through Turnitin, and fully referenced in your required style. Just send your brief for a free quote.