How to get cited by ChatGPT: what I check first on a client site
Short answer
To get cited by ChatGPT, first make sure OpenAI's search crawler, OAI-SearchBot, can reach your pages. Then make each page answer one question clearly in its opening lines, back claims with sources and numbers, and get your business mentioned on sites people already trust. Measure the results through ChatGPT referral traffic, not screenshots.
To get cited by ChatGPT, your pages have to be reachable by OpenAI's search crawler, easy to quote, and backed by evidence that other sites agree with. Most of the work is ordinary, careful SEO; the difference is in the order you do it and in how you measure it.
I have built over 200 websites in 8 years, and when a business owner asks me why ChatGPT never mentions them, I run the same five checks in the same order. Here they are, with links to what OpenAI itself says about each one, because most guides on this topic quote each other instead of the source. If you want this done for you, it is part of my SEO, AEO and GEO work; the rest of my notes on AI search are in the SEO, AEO and GEO hub.
How does ChatGPT decide which websites to cite?
ChatGPT cites websites when it runs a web search to answer a question. OpenAI says ChatGPT search uses third party search providers plus content from partners, and that any public website can appear. Your page is only a candidate if it can be found, crawled by OAI-SearchBot and read. Only after that does quality decide.
When OpenAI announced ChatGPT search on 31 October 2024, it explained that answers draw on third party search providers and on content supplied directly by partners (OpenAI, Introducing ChatGPT search (opens in a new tab)). It does not name the providers. Its publisher and developer FAQ (opens in a new tab), updated in October 2026, says any public website can appear in ChatGPT search, as long as OAI-SearchBot is not blocked.
What OpenAI does not publish is a list of ranking factors. Everything you read about "the eight signals" or "the seven patterns" is someone's observation, not documentation. Some of it is useful. None of it is a rule. I treat it the way I treat Google ranking studies: a hint about what correlates, checked against what I see on real sites.
So I split the problem in two:
- Eligibility. Can ChatGPT find, crawl and read the page? This part is documented, and it is binary.
- Selection. When ChatGPT has several eligible pages, which does it quote? This part is not documented, so the best you can do is make your page the clearest, best supported answer.
Check 1: can OAI-SearchBot crawl your site?
OAI-SearchBot is the crawler OpenAI uses to show websites in ChatGPT search. OpenAI states that sites which opt out of OAI-SearchBot will not be shown in ChatGPT search answers. Blocking it, by accident or on purpose, is the most basic reason a site never gets cited, so this is always my first check on a client site.
OpenAI runs three crawlers, and they do different jobs. Mixing them up is the most common mistake I see in robots.txt files:
| Crawler | What OpenAI uses it for | What blocking it does |
|---|---|---|
| OAI-SearchBot | Showing websites in ChatGPT search | Your pages are not shown in ChatGPT search answers, apart from bare navigational links |
| GPTBot | Collecting content that may train OpenAI's models | Your content should not be used for training; it does not remove you from search |
| ChatGPT-User | Fetching a page when a user asks ChatGPT to visit it | robots.txt may not apply, and OpenAI says it is not used to decide what appears in search |
Source: OpenAI's crawler documentation (opens in a new tab). Each setting is independent, so you can allow OAI-SearchBot and still block GPTBot if you do not want your content used for training. OpenAI says robots.txt changes take about 24 hours to reach search.
What I actually check, in this order:
- Open
/robots.txtand look forOAI-SearchBotand for a blanketUser-agent: *withDisallow: /. Staging rules that were never removed are a classic. - Check the security layer. Firewalls and bot protection on some hosts block crawlers they do not recognize, which robots.txt will never show you. I check the firewall settings, not just the file.
- Look for
noindexon pages that should be found. OpenAI respects noindex, and it can only read the tag if the page is crawlable. - Make sure the important text is in the HTML, not loaded later by JavaScript. If a page is blank until a script runs, assume a crawler sees it blank.
On this site, robots.txt names OAI-SearchBot, GPTBot, ChatGPT-User and the other AI crawlers explicitly, so there is no doubt about what is allowed.
Check 2: is the page in the search indexes ChatGPT draws on?
ChatGPT search pulls candidate pages from third party search providers that OpenAI does not name. In practice, that means a page nobody can find in the major web search indexes is unlikely to reach ChatGPT at all. Before working on wording, I confirm the page is indexed in Google and Bing, and fix whatever stops it.
This is the step owners skip, because it feels like old SEO. It is old SEO, and it still decides eligibility. My routine is short:
- Submit an XML sitemap in Google Search Console and in Bing Webmaster Tools, and check that the page you care about shows as indexed in both.
- Search for the exact page title in Google and Bing. If it is missing from either, find out why before doing anything else.
- Give every page one canonical URL, a unique title and a description that says what the page answers.
- Link to the page from related pages on your own site, with anchor text that describes it. A page with no internal links is hard for any crawler to find.
Nobody outside OpenAI knows exactly which index feeds which answer, and anyone who says otherwise is guessing. Being properly indexed in both Google and Bing is the cheapest insurance against that uncertainty.
Check 3: does each page answer one question clearly in its first lines?
ChatGPT quotes passages, not whole pages. A page that states its answer plainly in the first two sentences, and opens every section with a short, self contained answer, gives ChatGPT something it can lift and attribute. A page that warms up for three paragraphs before saying anything gives it nothing to quote, however good the rest is.
This is the change most service business sites I review need, because they are written like brochures. The fixes are editorial, not technical:
- Answer first. The first two sentences of the page answer the question in its title. Background comes after.
- One question per section. Write headings as the questions people actually ask, and open each section with a 40 to 60 word answer that makes sense on its own.
- Name things fully. Write "Google Search Console", not "it". A quoted sentence loses the paragraph around it, so every sentence that might be quoted has to stand alone.
- Use tables and numbered steps where the answer is a comparison or a process. They are easier to read for people and easier to extract for machines.
Every note on this site follows that structure, including this one, so you can see what it looks like in practice.
Check 4: is every claim backed by numbers and sources?
Pages that add statistics, quotations and citations to sources are more likely to be used in AI generated answers. The best known study on this found that those three changes raised a page's visibility in generative engine answers, while stuffing in keywords made it worse. Evidence is what separates a quotable page from a generic one.
The study is the GEO paper from Princeton and collaborators (opens in a new tab), published at the KDD conference in 2024. In its main results table, adding quotations scored about 41 percent above the unoptimized baseline, adding statistics and citing sources also scored well above it, and keyword stuffing scored below it. It was run before ChatGPT search existed and is now two years old, so I read it as direction rather than a formula.
What that means on a real service business site:
- Replace "we are trusted by many clients" with the specific, checkable facts you can publish: years in business, projects completed, the markets you serve.
- Link the claim to its source. If you quote a regulation, a price or a statistic, link the official page.
- Publish something only you can say: your own process, your own checklist, a cost breakdown from your own quotes. That is what other sites end up citing, and what makes you worth quoting.
- Never invent a number. One made up statistic undoes the trust that every honest one built.
Check 5: do other trusted sites mention your business?
ChatGPT tends to cite and recommend businesses that other sources already talk about, not only businesses that talk about themselves. Mentions in industry directories, local press, partner sites and well known review platforms give it independent evidence. A site that only describes itself is asking ChatGPT to take its word for it.
I cannot point you to an OpenAI document that says this, because there is none. It is a pattern I see when I test prompts for clients: the businesses that get named tend to be the ones that appear in several places with the same name, the same description and the same contact details. Treat it as experience, not a documented rule.
The practical version is unglamorous:
- Make sure your business name, what you do and where you work are described the same way on your site, your LinkedIn page and every directory you are listed in.
- Ask clients and partners who already link to you to describe what you do in their own words, not only to add your logo.
- Contribute something useful to the places your buyers read: an industry association article, a guest note with real numbers, an answer on a forum where you actually know the topic.
How do you know if you get cited by ChatGPT?
Measure ChatGPT citations in two ways: real referral traffic and a fixed monthly prompt test. ChatGPT adds utm_source=chatgpt.com to the links it sends, so Google Analytics can show those visits and what they did next. A repeated set of prompts then shows whether your business is named, and how often, as answers change between runs.
The traffic side takes ten minutes to set up. In Google Analytics 4, open the traffic acquisition report, filter session source to chatgpt.com, and save it. OpenAI confirms the parameter in the publisher FAQ linked above. Add a conversion, such as an enquiry form submission, so you see whether those visits turn into business, not just whether they arrive.
The prompt side needs discipline more than tools:
- Write 15 to 20 questions your buyers actually ask, in their words. Include the ones that name your service and your city or market.
- Run each question in a fresh ChatGPT session with search on, three times, once a month. Answers vary between runs, so one screenshot proves nothing.
- Record in a spreadsheet whether you are named, whether you are linked, and who is named instead.
- Compare month to month. The trend matters; any single answer does not.
This is the same principle as my monthly reports for retainer clients: a short page that shows visits and enquiries, not a 40 page dashboard export.
What I do not spend time on
Some popular advice does not hold up when you check it against OpenAI's own documentation, so I leave it out. There is no documented paid route into ChatGPT search answers, no published list of ranking factors, and no file you can add that guarantees a citation. Time spent on those is time taken from the five checks above.
- Guarantees. Nobody can promise that ChatGPT will cite you, and I do not. I can promise the site will be eligible, clear and well evidenced.
- Proprietary scores. An "AI visibility score" with no published method tells you less than your own prompt spreadsheet.
- Rewriting for robots. Pages written for a crawler read badly to people, and people are the ones who send enquiries.
If you want me to run these five checks on your site, send me your website address. I read every enquiry personally and reply within 24 hours.
Frequently asked questions
01 Does blocking GPTBot stop ChatGPT from citing my site?
No. GPTBot collects content that may be used to train OpenAI's models. ChatGPT search uses a different crawler, OAI-SearchBot, and OpenAI treats the two settings independently. You can block GPTBot to keep your content out of training and still allow OAI-SearchBot, so your pages can appear and be cited in ChatGPT search answers. Check both lines in your robots.txt file.
02 Can I pay to be cited or featured in ChatGPT answers?
OpenAI's documentation for publishers describes no paid way to appear in ChatGPT search answers. Inclusion depends on your pages being crawlable by OAI-SearchBot and on ChatGPT choosing them as sources. Be wary of anyone selling guaranteed ChatGPT placement; the most they can honestly offer is to make your site eligible, clearer and better supported by evidence.
03 Do I need to rank on Google to be cited by ChatGPT?
Not necessarily on page one, but your page needs to be found in the search indexes ChatGPT draws on. OpenAI says ChatGPT search uses third party search providers without naming them. In practice, I make sure a page is indexed in both Google and Bing, because a page missing from the major indexes is unlikely to be picked up at all.
04 How long does it take to get cited by ChatGPT?
There is no published timeline. OpenAI says robots.txt changes reach ChatGPT search in about 24 hours, so fixing a blocked crawler works quickly. Being chosen as a source depends on indexing, the quality of the page and mentions elsewhere, and nobody can predict how long that takes. Track it monthly with referral traffic and a fixed set of prompts.