ChatGPT can show and cite your website only if OpenAI's search crawler, OAI-SearchBot, is allowed to fetch it. That means robots.txt has to allow it, and so does your host or CDN. OpenAI's help pages say so, and they add that ChatGPT ranks results on factors OpenAI does not list. Once the crawler gets in, each page has to say plainly what you do.
I start with access, since that is the part OpenAI documents in detail. It says very little about the rest. Below I separate what OpenAI states from what we have seen ourselves and from what nobody outside OpenAI knows. Each OpenAI statement comes with the page and the date I read it.
When does ChatGPT search the web?
ChatGPT may search the web on its own when a question would benefit from current information, according to OpenAI’s help page on web search. A user can also type a slash in the message box and choose Search. When search is off, OpenAI says the answer comes from what the model learned in training. Its page on accuracy adds that the model knows nothing after its training cutoff unless a tool such as search is used.
So a company reaches a ChatGPT answer by two routes. Live search is the route you can open or close this week. Training data is fixed at the model's cutoff, so a correction you publish today reaches this model's answers only through search.
Which OpenAI crawler has to reach your pages?
OpenAI runs three agents that matter to a website, and its crawler overview treats each one separately. OAI-SearchBot is the one that decides search visibility.
| Agent | What OpenAI says it does | What blocking it does |
|---|---|---|
| OAI-SearchBot | It surfaces websites in the search results of ChatGPT. | The site is not shown in ChatGPT search answers. It can still appear as a navigational link. |
| GPTBot | It crawls content that may be used to train OpenAI’s foundation models. | It tells OpenAI not to use the content for training. OpenAI says this setting is independent of search. |
| ChatGPT-User | It visits a page when a user’s request needs it. It does not crawl on its own. | OpenAI says robots.txt rules may not apply to these user actions. ChatGPT-User does not decide what appears in search. |
Source: Overview of OpenAI crawlers, OpenAI, read September 26, 2026. The page shows no date of its own.
After you change robots.txt, OpenAI says its search systems can take about 24 hours to adjust. It is the only timing OpenAI gives.
You do not have to name the crawler. Under RFC 9309, the standard behind robots.txt, a crawler with no group of its own follows the group marked with an asterisk. We measured how often sites name AI crawlers at all. In our study of 303 non ecommerce websites, 97.7 percent of the 299 reachable homepages served a robots.txt and 28.4 percent named at least one AI crawler in it (July 2026). A name can mean an allow or a block, so the name alone tells you little. On most of the other sites, OAI-SearchBot falls under the asterisk group. Read that group first.
Why robots.txt alone does not open the door
I came to this work from cybersecurity assessments. I read robots.txt the way I read a written policy in an audit. It states what the owner wants. Your firewall, CDN and hosting platform decide what gets through. RFC 9309 says it in one line: “These rules are not a form of access authorization.”
That is why OpenAI asks for two separate things. Robots.txt must allow OAI-SearchBot. The host or CDN must also let traffic from OpenAI’s published search IP addresses through. In audit terms, the control can override the policy. A bot protection setting that challenges unfamiliar crawlers can turn OAI-SearchBot away while robots.txt invites it in.
When you open the firewall, open it by address. Any client can put OAI-SearchBot in the user agent of a request. Only OpenAI sends from the ranges it publishes at openai.com/searchbot.json. An allow rule tied to those ranges lets the real crawler in and nobody else who borrows the name.
We made the same choice on our own site. Our robots.txt, read on September 26, 2026, names OAI-SearchBot and GPTBot in separate groups and allows both. Each group closes the admin, cart, checkout, orders and account paths. Your choice on GPTBot can differ from ours, since OpenAI treats training and search as separate settings.
How does ChatGPT turn a question into a search?
ChatGPT does not pass the user’s sentence to a search engine as it was typed. OpenAI’s help page says ChatGPT search sometimes partners with other search providers. When it does, it typically rewrites the question into one or more targeted queries. In OpenAI’s own example, a researcher asks about drugs that target CCR8. ChatGPT might first send “CCR8 immunotherapy drug development 2025” and then “CHS-114 conference 2025.”
The rewrite can also carry context about the person asking. OpenAI says ChatGPT may use an approximate location based on the IP address. When memory is on, it may also use relevant saved memories. So two people who type the same question can trigger different searches.
I read OpenAI’s example this way: the query that reaches a search provider is shorter and more specific than the question. A page that states names, model numbers, places and dates in plain text gives that query something to match. OpenAI does not publish this as a rule, so treat it as my inference from its example.
OpenAI does not name one search provider. The help page links the privacy statements of Microsoft and Shopify for how providers handle queries. The launch post of October 31, 2024 said ChatGPT search uses third party search providers and content from partners. On ranking, OpenAI says only that it uses multiple factors to find relevant, reliable information. It also says it does not guarantee placement.
What is the difference between a citation and the Sources list?
A citation is a link inside the answer. The Sources button shows the cited sources together with other relevant links. So a page can sit in Sources without supporting any sentence of the answer. OpenAI also warns that search results and citations can be incomplete, outdated or incorrect.
When you record answers, keep citations and Sources appearances in separate columns, since a combined count overstates how visible a site is.
How do ChatGPT visits show up in analytics?
OpenAI’s publishers FAQ says ChatGPT adds utm_source=chatgpt.com to referral URLs on its own. In GA4 those visits appear under the session source chatgpt.com. That is where we read our own ChatGPT visits.
Treat the number as a floor. Some people copy your address out of an answer and type it into a browser. Those visits arrive without the tag, and GA4 cannot credit ChatGPT for them.
What OpenAI does not document
I read the OpenAI pages listed below on September 26, 2026. Here is what none of them covers.
- The ranking factors are not listed, beyond relevance and reliability.
- Nothing says how often OAI-SearchBot returns to a page, or how long an older version can stay in answers.
- The weight of each outside search provider is not given, and neither is the provider behind a given query.
- There is no rule for when ChatGPT cites a page and when it only lists it among the other links.
- Neither llms.txt nor schema markup appears on these pages as a factor in ordinary ChatGPT search.
Google is clearer about its own product. Its guide to generative AI features (last updated July 10, 2026) says Google Search ignores llms.txt files. OpenAI says nothing either way. Any list of ChatGPT ranking factors you read elsewhere is an observation or a guess, and the careful ones say which of the two it is.
Where do shopping, recommendations and ads fit?
Products follow a separate path. When a question suggests shopping intent, OpenAI says ChatGPT can show product options with images, details and links to sites that sell them. OpenAI adds that product results are selected independently and are not ads. How ChatGPT shopping works covers product feeds, the Shopify and Etsy integrations and merchant rankings.
How ChatGPT recommends products explains where the names in a recommendation come from. Whether ChatGPT is paid to recommend products covers the ads OpenAI began testing in the United States on February 9, 2026. They appear below the answer with a sponsored label.
What our scan checks, and what it cannot
The free scan checks the robots.txt rule for each page it reads against eight AI crawler names, each on its own. They are GPTBot, OAI-SearchBot and ChatGPT-User from OpenAI, ClaudeBot, Claude-SearchBot and Claude-User from Anthropic, and PerplexityBot and Perplexity-User from Perplexity.
The scan cannot prove that your firewall admits OpenAI’s real crawler. Our requests do not come from OpenAI’s addresses. For that check, look in your CDN’s bot settings or your server logs. You want OAI-SearchBot requests from the ranges OpenAI publishes, answered with a normal page.
We run the same scan on visibilitymesh.com every month, and you can see our result on Our Own Score. For the rest of AI visibility beyond ChatGPT, read AI visibility: what it is, what the work covers and how to measure it. For the rules themselves, read robots.txt for AI crawlers.
Questions people ask
Can I make ChatGPT show my company?
No one can make ChatGPT show a company. A service that guarantees a mention is promising something it does not control. You can influence what ChatGPT is able to find and read. OAI-SearchBot has to be able to fetch your pages, the pages have to say plainly what you do, and other sites should describe you the same way. Ask the same questions every few weeks and write down the date of each check. Then you can see whether anything changed.
How does ChatGPT search find websites?
ChatGPT finds websites through outside search providers and through OpenAI's own crawler, OAI-SearchBot. OpenAI's help center says ChatGPT searches when a question needs current information or when a user chooses Search. It sometimes rewrites the question into targeted queries for other search providers. Sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers, according to the crawler page, though they can still appear as navigational links. OpenAI ranks the results on several factors it does not publish.
Can I submit my website to ChatGPT?
OpenAI does not document any form for submitting a website to ChatGPT search. Its help center says any public website can appear. It asks site owners to allow OAI-SearchBot in robots.txt and to let traffic from OpenAI's published IP addresses through the host or CDN. After a robots.txt change, OpenAI says its search systems take about 24 hours to adjust. Merchants have one extra route. They can apply to share a product feed on OpenAI's merchant page, and stores that sell through Shopify or Etsy are already integrated.
Which search engine does ChatGPT use?
OpenAI does not name one engine. Its help center says ChatGPT search sometimes partners with other search providers and sends them rewritten queries. The same page links the privacy statements of Microsoft and Shopify for how those providers handle queries. OpenAI's launch post of October 31, 2024 said ChatGPT search uses third party search providers and content from partners. OpenAI also runs its own crawler, OAI-SearchBot. I would keep a site indexable in Bing as a precaution, but OpenAI does not say how much weight any provider carries.
How does ChatGPT decide which sources to cite?
OpenAI says ChatGPT ranks search results on multiple factors meant to help users find relevant, reliable information. It does not publish those factors. It does describe the process around them. ChatGPT rewrites the question into targeted searches and may use saved memories and an approximate location while it does so. Each citation links to its source, and only sites that allow OAI-SearchBot can be shown in search answers. Anything more specific, such as a preferred page format or word count, is someone's observation or someone's guess.
Why does ChatGPT list sources that are not cited in its answer?
The Sources list holds more than the citations. OpenAI's help center says responses that use web search may include citations. It also says that selecting Sources shows the cited sources and other relevant links. So a page can appear in that list without supporting any sentence in the answer. OpenAI does not say how it decides between the two. When you measure, count a citation only when the answer itself links to your page. Log Sources appearances in a separate column.
How long does ChatGPT search take to reflect changes to my website?
OpenAI documents one timing. After a robots.txt change, its search systems can take about 24 hours to adjust. OpenAI does not publish how often OAI-SearchBot comes back to a page or how quickly new content enters ChatGPT search. It also does not say how long an old version can keep appearing in answers. A model's own knowledge stops at its training cutoff unless a tool such as search is used, according to OpenAI. Ask the same questions a few days and a few weeks after a change, and keep the dates.
Does SEO help my website appear in ChatGPT?
SEO helps with part of it, and OpenAI does not say how much. ChatGPT search uses outside search providers alongside its own crawler. A site that search engines can crawl and index gives those providers something to return. Plain pages that state what you offer also help every reader, human or machine. OpenAI ranks results with its own factors, so a top Google position does not promise a ChatGPT citation. Check your search indexing, and check that OAI-SearchBot gets through both robots.txt and your firewall.
Sources
I opened every source below on September 26, 2026.
- Overview of OpenAI crawlers, OpenAI developer documentation.
- Searching the web with ChatGPT, OpenAI Help Center.
- Publishers and Developers FAQ, OpenAI Help Center.
- Does ChatGPT tell the truth?, OpenAI Help Center.
- Introducing ChatGPT search, OpenAI, October 31, 2024.
- Shopping with ChatGPT Search, OpenAI Help Center.
- Ads in ChatGPT, OpenAI Help Center.
- RFC 9309, Robots Exclusion Protocol, IETF, September 2022.
- Optimizing your website for generative AI features on Google Search, Google Search Central, last updated July 10, 2026.
- AI visibility study of 303 non ecommerce websites, Visibility Mesh, measured July 12 and 13, 2026.
- The robots.txt file of visibilitymesh.com, read on September 26, 2026.
See whether the crawlers of OpenAI, Anthropic and Perplexity are allowed into your pages. The free scan reads up to five of your key pages, and you do not need a card or a call. Run the free scan.