Skip to content

AI visibility measurement guide · By Yes AI

How does ChatGPT search find and cite websites?

Published 6 min read

ChatGPT answers some questions from what its model already knows and searches the web for others. When it searches, it can only cite pages that its search sources can reach. This guide sets out what OpenAI has published about that process and what an Australian business can check on its own site.

Quick answer

OpenAI says ChatGPT decides whether to search the web based on the question, and users can also start a search themselves. ChatGPT search uses third-party search providers and content from OpenAI's partners. OpenAI's OAI-SearchBot crawler surfaces websites in ChatGPT search, and sites that opt out of it are not shown in search answers except as navigational links. GPTBot is a separate crawler used for model training.

Business owner at a desk in a Perth workshop office reviewing website settings on a laptop

When does ChatGPT search the web?

OpenAI's announcement of ChatGPT search, published in October 2024 and updated in February 2025, says ChatGPT will choose to search the web based on what you ask, or the user can choose to search by clicking the web search icon. Answers that use search include links to sources, and a Sources button opens the list of references. Answers that do not use search come from the model's training data, which has a cut-off date. A business that opened, moved or changed its services after that date may be missing or out of date in those answers until the model is updated.

The product has changed since that announcement, so treat the post as a description of how search was designed and check the current interface when you test.

Which OpenAI crawlers visit your site?

OpenAI's crawler documentation lists four user agents, each with its own purpose and its own published list of IP addresses. Two of them are controlled through robots.txt tags that OpenAI describes as independent of each other. A site can allow OAI-SearchBot so it appears in search results and disallow GPTBot so its content is not used to train OpenAI's foundation models. If both are allowed, OpenAI says it may use one crawl for both purposes. The other two agents act only in specific situations: ChatGPT-User when a person's request leads ChatGPT to open a page, and OAI-AdsBot when a page has been submitted as an ad.

  • OAI-SearchBot surfaces websites in ChatGPT search. IP list: openai.com/searchbot.json.
  • GPTBot crawls content that may be used to train foundation models. IP list: openai.com/gptbot.json.
  • ChatGPT-User visits pages for user actions, and OpenAI says robots.txt rules may not apply. IP list: openai.com/chatgpt-user.json.
  • OAI-AdsBot checks landing pages submitted as ChatGPT ads. IP list: openai.com/adsbot.json.

What happens if you block OAI-SearchBot or GPTBot?

OpenAI states that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though they can still appear as navigational links. It recommends allowing OAI-SearchBot in robots.txt, and allowing requests from its published IP ranges, if you want your site to appear. Disallowing GPTBot signals that your content should not be used to train OpenAI's generative AI foundation models. OpenAI also says ChatGPT-User is not used to decide whether content may appear in search. After a robots.txt change, OpenAI says its systems can take about 24 hours to adjust.

A common mistake is one rule written to keep out AI training that also catches OAI-SearchBot, such as a blanket Disallow in the group for every user agent. Check the group that applies to each agent separately.

Where do ChatGPT's search results come from?

OpenAI's announcement says ChatGPT search uses third-party search providers as well as content provided directly by its partners. The page does not name the providers. Microsoft states that its grounding service powers nearly every major AI assistant on the market; read that as Microsoft's description of its own business. For a site owner, the practical point is that being in ordinary search indexes still matters for AI search. Verify the site in Bing Webmaster Tools as well as Google Search Console, and confirm that your key pages are indexed in both.

What makes a page usable in a ChatGPT answer?

OpenAI has not published ranking factors for ChatGPT search, so be wary of anyone who claims to know them. What you control is whether the page is reachable and whether the facts a buyer needs are written on it. Put your service area, licences, the work you take on and decline, and how to book in plain text on the relevant page, so the page makes sense when read on its own. Keep the same details consistent across your website, Google Business Profile and directory listings, because an answer drawing on several sources may repeat whichever version it finds.

How do you test what ChatGPT says about you?

Write a short list of questions a buyer would ask before choosing a supplier in your category and area, without naming your business. Run each one, save the full answer and the cited links, and note the date, whether you were logged in and the location setting. Answers are generated each time and can differ between runs of the same question, so one answer is a sample. Repeat the same set on a schedule and compare like with like. The cited links are the most useful part: they show which pages the search drew on, including competitors' pages and directories.

OpenAI agents and what blocking each one changes

AgentPurpose (per OpenAI)robots.txt controlEffect of blocking
OAI-SearchBotSurfaces websites in ChatGPT searchYes, by its own tagSite not shown in ChatGPT search answers, apart from navigational links
GPTBotCrawls content that may be used to train foundation modelsYes, by its own tagSignals content should not be used for training; the OAI-SearchBot setting is independent
ChatGPT-UserVisits pages for user actions in ChatGPT and custom GPTsRules may not apply, as visits are user-initiatedNeeds a firewall rule; not used to decide search inclusion
OAI-AdsBotChecks landing pages submitted as ChatGPT adsNot describedOnly visits pages submitted as ads

Practical checklist

  • The robots.txt group that applies to OAI-SearchBot allows the pages you want cited.
  • The GPTBot decision is recorded separately, with the reason.
  • The firewall or CDN allows requests from the addresses in openai.com/searchbot.json.
  • Key facts appear as text in the page HTML, not only in images or PDFs.
  • Key pages are indexed in both Google Search Console and Bing Webmaster Tools.
  • Test questions, answers, cited links, dates and account state are saved together.

Common questions

Can I submit my website to ChatGPT?

OpenAI's announcement says any website or publisher can choose to appear in ChatGPT search, and its crawler documentation describes how: allow OAI-SearchBot in robots.txt and allow requests from its published IP ranges. Neither document describes a submission form or a paid listing. Being allowed means the site can be used; whether it is cited for a particular question depends on the search at that time.

Will blocking GPTBot stop ChatGPT recommending us?

Blocking GPTBot signals that your content should not be used to train OpenAI's foundation models. OpenAI describes the GPTBot and OAI-SearchBot settings as independent, so a GPTBot block leaves your ChatGPT search eligibility unchanged. Models trained before the change are not retrained by it, so anything they learned earlier can still appear in answers drawn from training data.

Does ChatGPT-User obey robots.txt?

OpenAI says ChatGPT-User is used for certain user actions, such as when a person's question leads ChatGPT to visit a page, and that because these actions are user-initiated, robots.txt rules may not apply. It also says ChatGPT-User is not used to decide whether content may appear in search. Stopping these visits has to be done at the firewall, using OpenAI's published IP list.

How long does a robots.txt change take to affect ChatGPT search?

OpenAI says it can take about 24 hours from a robots.txt update for its systems to adjust for search results. That covers the crawler permission only. It says nothing about when, or whether, a page will be cited, which depends on the questions people ask and the search results available at the time.

Why does ChatGPT name a competitor and not us?

OpenAI has not published how ChatGPT chooses which businesses to name. Start with the cited links in the answer, which show the pages the search used. Compare those pages with yours for the facts the question asked about, such as service area or the specific job. Repeat the question several times before drawing a conclusion, because answers vary between runs.

Sources and further reading

Continue planning

Plan an evidence-led AI visibility review

Describe your business category, service area and the questions buyers ask. We can discuss a repeatable visibility baseline and which public content needs review. Keep customer records and confidential information out of this enquiry.