How to Cite ChatGPT & AI Tools: APA, MLA, Harvard & More (2026)

Can you cite AI — and should you?

Generative AI tools such as ChatGPT, Claude, Gemini and Copilot have moved from novelty to everyday study aid, and referencing styles have raced to catch up. The headline is simple: if you have used an AI tool in producing a piece of work, you should be able to account for that use honestly — which usually means both a disclosure (a statement that you used AI, and how) and, where you draw on specific output, a citation.

But before any of the formatting below, one rule overrides everything else: check your institution’s policy and the specific assessment brief. Universities vary enormously. Some allow AI for brainstorming and proofreading but not for drafting; some require a declaration; some ban it entirely for certain assessments. Using AI where it is prohibited is academic misconduct no matter how carefully you cite it — the citation does not grant permission, it only documents use. When the brief is unclear, ask your tutor in writing before you rely on AI.

This guide assumes you are using AI legitimately and with permission — for example to explain a concept, to generate study questions, or as a starting point you then verify and rewrite — and need to reference it correctly.

Disclosure and academic integrity come first

Most institutions now distinguish between three things: using AI, disclosing that use, and citing specific output. Disclosure is usually a short statement — often at the end of the work or in a declaration — saying which tool you used and for what (for example, ‘ChatGPT was used to generate practice questions and to check grammar; all analysis and writing are my own’). Citation is the formal reference you add when you quote, paraphrase or build on a specific AI response.

Remember three practical points. First, AI output is often not recoverable: another reader cannot retrieve the exact response from your prompt, which is why styles treat it differently from a normal source. Second, AI can be confidently wrong and invents citations — so anything factual must be verified against a real, citable source, which you then reference instead. Third, AI text is not your own voice; submitting it as if it were is the integrity risk that matters, far more than a misplaced comma in the reference.

How each style cites ChatGPT and other AI tools

The major styles have published guidance, and while the details differ, they share a logic: name the tool, its maker, the version and the date, and make clear in the text what you prompted. The table shows the current recommended format in each style.

What an AI citation must include

Whatever the style, a complete AI citation captures five things:

• The tool’s name — ChatGPT, Claude, Gemini, Copilot.
• The developer — OpenAI, Anthropic, Google, Microsoft — which several styles treat as the author.
• The version and date — models change often, so ‘GPT-4o, March 2026 version’ matters for reproducibility.
• The URL of the tool.
• Your prompt — described or quoted in the text, so the reader understands what produced the output.

Because the conversation is not publicly retrievable, several styles also suggest including the full prompt and response in an appendix. That is good practice: it lets a marker see exactly how you used the tool, which supports your integrity declaration rather than undermining it.

Referring to AI in your text

In author–date styles (APA, Harvard), the in-text citation uses the developer as author and the year: (OpenAI, 2024). Describe the prompt in your sentence so the citation makes sense: When asked to summarise the causes of the 2008 crisis, ChatGPT produced a list of five factors (OpenAI, 2024), which I verified against… In numbered styles (Vancouver, IEEE) the AI tool simply takes the next number, like any other source. In MLA you begin the Works Cited entry with a shortened version of your prompt. The common thread is transparency: the reader should always be able to see that you used AI and what for.

The six most common mistakes when citing AI

1. Citing AI when it was banned. A citation does not authorise use — check the policy first.
2. Not disclosing use at all. Many institutions require a declaration in addition to any citation.
3. Treating AI output as a factual source. Verify claims against real sources and cite those instead.
4. Quoting invented references. AI fabricates citations; never copy a reference list AI produced without checking each entry exists.
5. Omitting the version and date. Models change; record which one you used.
6. Passing AI text off as your own. This is the serious integrity breach, regardless of formatting.

Why a citation is not a free pass

It is worth stating plainly, because students sometimes assume the opposite: citing ChatGPT does not make it acceptable to submit AI-written work as your own. A citation documents that you used a tool; it does not transform machine-generated prose into your own scholarship. Assessment exists to evidence your understanding, and markers — supported by detection tools and, more reliably, by viva-style questioning — are increasingly alert to work that a student cannot explain.

The defensible way to use AI is as a study aid you remain in control of: to explain a difficult concept, to generate practice questions, to suggest an essay structure you then write yourself, or to flag grammar issues you then fix. In each case the thinking, the analysis and the words that are assessed are yours; the AI is a tutor or a tool, disclosed and where relevant cited. Used that way, AI supports learning. Used to replace your own work, it puts your degree at risk — and no referencing format changes that.

When to reference the real source, not the AI

One of the most important habits is knowing when not to cite the AI at all. If you ask a tool for a fact, a statistic or a definition and it gives you one, the correct move is not to cite the AI — it is to find the original, authoritative source the fact comes from, verify it, and cite that. AI output is a starting point for finding evidence, not the evidence itself. A marker will trust ‘(World Health Organization, 2021)’ far more than ‘(OpenAI, 2024)’ for a public-health statistic, and rightly so.

You cite the AI tool itself only when the output is the thing you are discussing — for example if your assignment analyses how a model answers a prompt, or you used it to generate an example you then critique. For ordinary factual claims, treat AI as a knowledgeable friend who might be wrong: a lead to chase, never a source to quote. This single distinction prevents most AI referencing problems and stops you repeating the fabricated facts these tools sometimes produce.

Legitimate ways to use AI in your studies

Used within your institution’s rules, AI can genuinely support learning without crossing into misconduct. Defensible uses include asking it to explain a concept in simpler terms, to generate practice questions before an exam, to suggest a structure you then write yourself, to check grammar and clarity in your own draft, or to summarise your own notes for revision. In each case the analysis and the assessed writing remain yours.

The line is crossed when the AI does the thinking the assessment is meant to evidence: writing your argument, producing analysis you present as your own, or generating text you submit largely unchanged. If you would be uncomfortable explaining to your tutor exactly how you used the tool, that is a signal to stop. When you do use AI legitimately, disclose it, keep the prompts and outputs, and cite specific output in the style your course requires, using the formats above.

Format every source correctly, every time

Where the official AI guidance lives

Each style maintains its own up-to-date guidance: the APA Style blog, the MLA Style Center, Cite Them Right for Harvard, the Chicago Manual of Style Online, and the relevant medical and engineering bodies for Vancouver and IEEE. Because the rules are evolving quickly, check the official page for your style before submission — and, above all, check your own university’s AI and academic-integrity policy, which is the authority that actually governs your work.

Related referencing guides

• APA 7th edition referencing guide
• Harvard (Cite Them Right) referencing guide
• MLA 9th format & citation guide
• Citation styles compared: Harvard vs APA vs MLA vs more

Frequently asked questions

Do I have to cite ChatGPT?

If you use it and your institution permits AI, then yes — most now expect both a disclosure of how you used it and a citation where you draw on specific output. But first check your assessment brief, because some assessments forbid AI entirely, in which case you should not be using it at all.

Is using ChatGPT considered cheating?

It depends entirely on your institution’s policy and the task. Using it where it is permitted, with disclosure, is usually fine; using it where it is banned, or submitting AI-written work as your own, is academic misconduct. Citing it does not make prohibited use acceptable.

How do I cite ChatGPT in APA 7?

Treat it as software: OpenAI. (2024). ChatGPT (GPT-4o version) [Large language model]. https://chat.openai.com, with the in-text citation (OpenAI, 2024). Describe your prompt in the text and consider including the full exchange in an appendix.

Can I trust the references ChatGPT gives me?

No. AI tools frequently invent plausible-looking citations that do not exist. Never copy an AI-generated reference list without checking that each source is real; verify every fact against a genuine source and cite that instead.

Why do I need the version and date?

AI models are updated frequently, so the same prompt can produce different output over time. Recording the version (for example GPT-4o) and the date you used it supports reproducibility and transparency.

Does citing AI protect me from an academic-integrity case?

No. A citation documents that you used a tool; it does not authorise prohibited use or turn machine-generated text into your own work. The safest approach is to use AI only where permitted, disclose it, keep the thinking and writing your own, and verify everything.

Need your whole reference list checked or an assignment formatted in AI by a subject expert? Place an order or explore our proofreading & editing service — rated 4.4/5 across 871 verified Trustpilot and Sitejabber reviews.