Can you cite ChatGPT or AI as a source?

Why you shouldn’t cite AI as a source of facts

A citation exists so a reader can check your claim against the evidence. A large language model isn’t evidence — it predicts plausible-sounding text, which means it can state something false with total confidence and even produce realistic-looking citations, page numbers, and quotes that were never written. Citing the AI just moves your claim one step further from the evidence, in the wrong direction. (For the general version of this test, see how to tell if a source is reliable.)

Disclosing the tool vs. citing the fact (the part people get wrong)

There are two different things, and conflating them is the mistake:

Disclosing AI use — telling the reader you used ChatGPT to brainstorm, draft, or summarize. APA and MLA both have formats for this, and many schools and journals now require it. This is legitimate and honest.
Citing AI as a source of a fact — using “ChatGPT said so” as your evidence for a claim. This is what you should never do, because the AI is not a reliable witness to the facts inside its own answer.

Disclose the tool if required; verify and cite the primary source for every fact.

How to verify an AI claim before you use it

Isolate the claim. Pull out the specific assertion — the name, number, date, or quote — from the AI’s explanation around it.
Find the primary source. Search for the original the claim should rest on. Don’t accept a citation the AI gave you without opening it — invented citations are common.
Check the source actually says it. Confirm the same number and the same context. “Right document, wrong number” is the most frequent AI error.
Cite the primary source, not the AI. Reference what you verified. Disclose AI use separately if your assignment requires it.

What about AI search engines like Perplexity?

AI search tools that show citations are a step better — but the citation is a starting point, not a guarantee. The model can still summarize a source incorrectly, so follow each citation to the primary source and confirm it says what the answer claims before you rely on it.

The fast way to check AI/ML claims

For claims about AI and machine-learning research specifically, SourceScore’s VERITAS verification API does this verification step for you: submit a claim and it returns a confidence score plus the primary source behind it, drawn from a hand-verified catalog. See the methodology for how each claim is sourced and signed, or browse the reliability scores for 130+ sources you might cite.

Frequently asked questions

Can you cite ChatGPT in an essay or research paper?

Not as a source of facts. ChatGPT can produce confident, false statements and even fabricate citations, so it fails the basic test of a reliable source. You may need to DISCLOSE that you used it (many schools and journals require this) as a tool or method — but for any factual claim, find the primary source it should rest on and cite that instead.

How do you cite ChatGPT in APA or MLA?

Both have formats for disclosing AI tool use. APA 7th treats a ChatGPT response like software output (e.g., OpenAI. (2026). ChatGPT [Large language model]. https://chat.openai.com) and recommends describing your prompt and how you used it. MLA cites it as a source you consulted, noting the prompt and date. But this documents that you USED the tool — it does not make the AI a trustworthy source of the facts inside its answer. Verify those separately.

Is using ChatGPT plagiarism?

Using it isn't automatically plagiarism, but presenting AI-generated text as your own original work — without disclosure where it's required — can be. Policies vary by institution, so check yours. The safer pattern: use AI to draft or explore, write in your own words, disclose where required, and cite the primary sources behind every factual claim.

Can ChatGPT be wrong or make up sources?

Yes, frequently. Large language models predict plausible text, not verified truth, so they can state false facts with full confidence and invent realistic-looking citations, page numbers, and quotes that don't exist (often called 'hallucination'). Always treat an AI answer as a lead to verify, never as a finished fact.

How do you check if what ChatGPT says is true?

Isolate the specific claim, find the primary source it should rest on, and confirm that source actually says it — same number, same context. Don't trust a citation the AI provides without opening it. SourceScore's VERITAS API automates this verification step for AI/ML claims, returning a confidence score and the primary source for each.