Watercolour illustration: EasyFound's blue mascot, Arrow, stands on a wooden desk by a shuttered shophouse window, lifting one glowing sheet from a stack of pages between an old typewriter and a fountain pen, while a paper crane waits on the sill.
AIResearchGuide

Can an AI-written article get cited by ChatGPT?

EasyFound · · 7 min read

TL;DR

Yes, an article drafted with AI can be cited by ChatGPT. OpenAI's guidance for websites doesn't mention how a page was written, and in our test ChatGPT cited one of our AI-drafted, human-checked articles once. It has also cited other pages that say they were drafted with AI. That shows it can happen, not how often, so check every fact and add what only you know.

Yes, an article drafted with AI can get cited by ChatGPT. OpenAI's guidance for websites says nothing about how a page was written, and when we tested our own AI-drafted, human-checked articles, ChatGPT cited one of them once, as its only source. That shows it can happen, not how often.

We draft EasyFound's articles with AI. A person picks the question, runs the tests and checks every figure before anything goes live. So we asked ChatGPT the questions our own articles answer, and looked for other cited pages that say they were written with AI.

Does ChatGPT check whether an article was written by AI?

ChatGPT isn't known to check whether an article was written by AI: OpenAI's guidance for getting into its answers doesn't mention AI writing at all. Its help page says ChatGPT "ranks search results using multiple factors intended to help users find relevant, reliable information. Placement is not guaranteed."1 Its publisher FAQ asks only that you "make sure you aren't blocking OAI-SearchBot", its search crawler2.

OpenAI's own attempt at spotting AI writing didn't hold up. It withdrew its AI-text classifier in July 2023 "due to its low rate of accuracy"3. In OpenAI's tests it wrongly labelled human writing as AI-written 9% of the time. So we didn't use a detector: we tested pages whose authorship we know.

Did ChatGPT cite our own AI-drafted articles?

ChatGPT cited one of our AI-drafted articles once in our test, as the only source for its answer. We asked it the title question of each of the 15 articles OpenAI's search crawler had already fetched, twice each, on 4 October 2026 (all 30 runs, CSV).

What ChatGPT did with 30 answers to the title questions of EasyFound's AI-drafted articles (OpenAI API with web search, Malaysia, 4 October 2026)
What ChatGPT did Answers
Answered from memory, no web search 19
Searched and cited only OpenAI's own pages 6
Searched and cited Google's help pages or customs bodies 4
Searched and cited our article 1

The cited article answers "What sources does ChatGPT cite most?". ChatGPT drew on its paragraph about US studies ranking Wikipedia and Reddit at the top, not on our headline Malaysian finding. Asked the same question a second time, it cited OpenAI's pages instead. In the other searched answers, our pages weren't even in its search results.

One citation in 30 answers shows that an AI-drafted article can be cited. It isn't a citation rate.

Our reading: most runs failed for reasons that need no AI explanation. ChatGPT often didn't search, and for questions about itself it went to OpenAI's pages. But with no human-written version to compare, 30 answers can't rule authorship out.

Does ChatGPT cite other pages that say they were written with AI?

Yes. In our saved ChatGPT answers, at least three cited pages said they were drafted or generated with AI. We fetched all 4,073 pages ChatGPT cited in our tests from late August to early October 2026 and searched the text for AI-writing notes. 2,660 of them opened as readable pages (classified matches, CSV).

  • A renovation cost guide on a directory site, which says it was drafted with AI from its listings' cost data and checked against local contractor rates. ChatGPT quoted its price range when asked to suggest a renovation contractor.
  • An AI search tool's study of which sites ChatGPT cites, drafted with AI from the firm's own tracking data and reviewed by an editor.
  • A video app's question-and-answer page, under a site-wide note that it contains AI-generated content.

Three is a minimum, not a share. Most AI-assisted pages carry no note, and a note was the only way we could tell.

Our reading: in the two citations we traced, our article and the cost guide, ChatGPT lifted one passage that answered part of the question (a paragraph on US studies, a price range), not the page as a whole. Two cases can't show that the passage is what won the citation.

What do Google and researchers say about AI-written content in AI answers?

Google says the method matters less than the result. Its Search team wrote: "Using AI doesn't give content any special gains. It's just content."4 Its current guidance adds: "It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing."5

The line Google draws is volume without value: its spam policy targets pages mass-produced to manipulate rankings6, as our AI SEO explainer covers. That's Google Search. OpenAI's two pages above publish nothing similar.

Two lab studies found AI systems favouring AI-written text:

  • Search models. Neural retrieval models ranked AI-rewritten versions of benchmark documents above the human-written originals. The authors call this "source bias"7.
  • AI assistants. Models including GPT-4 picked products and papers described by an AI more often than human judges did8.

Neither tested ChatGPT's web search. Our reading: a possible lab-scale edge for AI-style text, not a reason to publish unchecked drafts.

What can't this test tell me about AI-written articles and ChatGPT?

  • Whether a human-written version would do better or worse. We had no matched human-written articles to compare.
  • Anything beyond one small, new site. It's 30 answers through OpenAI's API, which isn't the ChatGPT app.
  • How common AI-written pages are among citations. The scan only finds pages that say so.

How do I use AI to write an article ChatGPT will cite?

Use AI for the draft and put your effort into the kind of passage ChatGPT quoted in our test: facts only you have, checked, under a clear question. Writing for AI answers this way is often called GEO, generative engine optimisation.

  1. Request EasyFound's free check. See what ChatGPT says about your business now, then go through it with us on a free call.
  2. Give the AI your own numbers to work from, such as price bands, areas covered and typical job times. The renovation cost guide above was quoted for a price range built from its own listings.
  3. Fact-check every figure, name and date yourself. Google calls this review "critical", and a wrong price quoted by ChatGPT is your problem, not the AI's.
  4. Answer one customer question per section, in the first sentence, as in how to write a page ChatGPT will quote.
  5. Publish a few useful pages, not dozens of near-copies. Mass-produced pages are what Google's spam policy targets.
  6. Make sure ChatGPT can reach the page. Our citation guide covers OpenAI's search crawler.

Checking what AI tools say and fixing the pages behind it is the job EasyFound does, as an AI SEO service for Malaysian businesses: six AI tools, in English, Malay and Chinese. Start with the free check below.

Which of your pages does ChatGPT use?

However your pages were written, see what ChatGPT says when customers ask about your business, then go through it with us on a free call.

Frequently asked questions

Will ChatGPT cite my article if I wrote it with AI?

It can. OpenAI's help pages for websites don't mention how a page was written: they ask that its search crawler, OAI-SearchBot, isn't blocked and say ChatGPT ranks results for relevant, reliable information. In EasyFound's October 2026 test through OpenAI's API, ChatGPT cited one of our own AI-drafted, human-checked articles once, as the only source for its answer. That shows it can happen, not how often: in most runs ChatGPT didn't search or cited other pages.

Can ChatGPT tell if an article was written by AI?

OpenAI doesn't say ChatGPT checks, and reliable detection is hard. OpenAI withdrew its own AI-text classifier in July 2023 'due to its low rate of accuracy'. In OpenAI's tests it caught only 26% of AI-written text. It also wrongly labelled human writing as AI-written 9% of the time, so don't rely on an AI detector to judge your own article either.

Does Google's guidance on AI content apply to ChatGPT?

Not directly. Google's guidance covers Google Search, where it says 'Using AI doesn't give content any special gains. It's just content.' OpenAI's help pages for websites publish no similar rule about AI writing. Google's advice to fact-check and review all AI-generated content before publishing is still worth following for any page ChatGPT might quote.

Should I say my article was written with AI?

Where readers would wonder, yes. Google says AI or automation disclosures 'are useful for content where someone might think "How was this created?"'. Some pages ChatGPT has cited carry a short note saying they were drafted with AI and checked by a named person. EasyFound says on this page that we draft with AI and a person checks every figure.

Sources

Our data:

  • EasyFound, Does ChatGPT cite our AI-drafted articles?, 4 October 2026. The title questions of the 15 EasyFound articles that OpenAI's search crawler, OAI-SearchBot, had fetched by 3 October 2026 (from our server logs), each asked twice through OpenAI's API (model gpt-5.4-mini, web search available, location Malaysia), not the ChatGPT app. The articles were drafted with AI and checked and edited by a person. Download the runs (CSV).
  • EasyFound, AI-writing notes on pages ChatGPT cited, 4 October 2026. The 4,073 different pages cited in EasyFound's saved ChatGPT answers from 25 August to 4 October 2026, fetched once without running scripts. 2,660 returned readable web pages; the rest were PDFs, blocked or gone. We searched the text for English, Malay and Chinese phrases such as "drafted with AI" and read every match by hand. Business and agency addresses are replaced with type labels. Download the classified matches (CSV).

Other sources:

  1. OpenAI, Searching the web with ChatGPT, Help Center, read 4 October 2026 (link)
  2. OpenAI, Publishers and Developers - FAQ, Help Center, read 4 October 2026 (link)
  3. OpenAI, New AI classifier for indicating AI-written text, 31 January 2023, withdrawn 20 July 2023 (link)
  4. Google Search Central (Danny Sullivan and Chris Nelson), Google Search's guidance about AI-generated content, February 2023 (link)
  5. Google Search Central, Google Search's guidance on using generative AI content on your website, updated 1 October 2026 (link)
  6. Google Search Central, Spam policies for Google web search (link)
  7. Dai et al., Neural Retrievers are Biased Towards LLM-Generated Content, KDD 2024; texts rewritten by Llama 2 and ChatGPT from public benchmark corpora (link)
  8. Laurito et al., AI–AI bias: Large language models favor communications generated by large language models, PNAS 122(31), 2025 (link)

The results are shared under CC BY 4.0: you're free to reuse them, with credit to EasyFound.

Checked 4 October 2026. ChatGPT's search and the pages it cites change often; treat these as dated findings.

Research by Janice Pang, EasyFound.