You used ChatGPT in your research. Now the references section looms. How to cite ChatGPT in academic papers? The major style guides have issued guidance, your institution has a policy (check it), and there’s a deeper question underneath: should you be citing it at all? Let’s cover the mechanics and the judgment.
How to cite ChatGPT in academic papers
First: check your institution’s policy (non-negotiable)
Before formatting anything: your university, department, or journal has an AI policy, and it overrides everything below. Some require disclosure statements; some restrict AI use to specific tasks; a few still prohibit it in assessed work. Citing correctly but violating policy is still violating policy. Find the policy, read it, follow it. If it’s ambiguous, ask your advisor — the conversation takes five minutes and prevents real trouble.
APA Style (7th edition + AI guidance)
APA treats AI-generated text as personal communication-like non-retrievable… with a twist: since readers can’t retrieve your exact chat, APA recommends citing in-text with the prompt included and referencing the model version. Format:
In-text: (OpenAI, 2026) with the full prompt described in your text: “When prompted with ‘[your prompt],’ ChatGPT responded…”
Reference list: OpenAI. (2026). ChatGPT (GPT-5 version) [Large language model]. https://chat.openai.com/
APA also recommends including the full AI output in an appendix for verifiability. Check the APA Style website for the latest — this guidance evolves.
MLA Style (9th edition + AI guidance)
MLA‘s approach: cite as you would any generated source, with the prompt as the “title.” Format:
Works Cited: “Explain the causes of the French Revolution in 200 words.” prompt. ChatGPT, GPT-5 version, OpenAI, 14 Oct. 2026, chat.openai.com.
In-text, use a shortened version of the prompt as the parenthetical. MLA’s core logic: the prompt is what makes your AI output unique and (quasi-)retrievable.
Chicago Style
Chicago generally recommends citing AI-generated content in notes (not bibliography), since it’s non-retrievable:
Footnote: Text generated by ChatGPT, OpenAI, GPT-5, October 14, 2026, in response to the prompt “[your prompt].”
Chicago’s Manual hasn’t fully standardized this; follow your publisher’s specific guidance where it exists.
The deeper question: should you cite it at all?
Mechanics aside, think about what you’re citing AI for:
- AI as a source of facts: Don’t. AI isn’t a primary source — track down the actual source and cite that. Citing ChatGPT for a factual claim is like citing “someone told me.” (And given why ChatGPT makes up facts, the facts need verification anyway.)
- AI as a tool (editing, brainstorming, coding help): Usually disclosed in a methods/acknowledgments note rather than cited: “Drafts were edited with assistance from ChatGPT (OpenAI, 2026).” Most policies prefer this over formal citation for tool-use.
- AI output as an object of study: If you’re analyzing AI responses themselves (linguistics, HCI, AI safety research), cite formally — the output is your data.
- When in doubt: disclose. Over-disclosure is never punished; under-disclosure increasingly is.
Disclosure statements (what they look like)
Many journals now want something like: “The authors used ChatGPT (OpenAI, GPT-5, 2026) for language editing and literature search assistance. All content was reviewed and verified by the authors, who take full responsibility for the work.” Adapt to your policy’s requirements. Place it in methods, acknowledgments, or a dedicated AI-use statement per the venue’s instructions.
Common mistakes
- Citing AI for facts instead of finding sources: the #1 error. AI points; sources prove.
- Omitting the prompt: without it, the citation is meaningless — the same model gives different outputs to different prompts.
- Omitting the version/date: models change; “ChatGPT” without version/date is unverifiable.
- Hiding AI use: the integrity risk dwarfs any benefit. Disclosure is the norm now, not the exception.
- Trusting AI-generated citations: verify every reference independently — fabricated citations in AI-assisted papers are an academic misconduct fast-track.
Bottom line
Sample disclosure statements (copy-adapt)
Here are templates for common situations — adapt to your institution’s requirements: For language editing: “The authors used ChatGPT (OpenAI, GPT-5, October 2026) for grammar and language polishing. All scientific content, analysis, and conclusions are the authors’ own.” For coding assistance: “Analysis code was developed with assistance from ChatGPT (OpenAI, 2026); all code was reviewed, tested, and validated by the authors.” For literature search: “ChatGPT (OpenAI, 2026) assisted in identifying relevant literature; all cited works were independently verified by the authors.” For brainstorming: “Initial project ideas were explored with ChatGPT (OpenAI, 2026); the research design and execution are the authors’ own.” The pattern: name the tool, specify the version/date, delimit exactly what it did, and assert your responsibility for the rest. Transparency with boundaries — that’s what editors and committees want to see.
Learn your style guide’s format, follow your institution’s policy, disclose tool use honestly, and never cite AI as a factual authority. AI for literature reviews covers the research workflow where these questions arise most, and Perplexity vs ChatGPT citations helps when you need AI that actually sources claims. The academic norm has settled: AI assistance is fine, AI authorship isn’t, and transparency is mandatory. Cite the tool, own the work.
