Guide
Published · By Shoaib Mustaque· 6-minute read
ChatGPT search retrieves candidate pages through third-party search providers; Microsoft has said Bing is ChatGPT's default search experience. A page Bing has not indexed is therefore unlikely to be cited. From the candidates, the model keeps passages that answer the question directly and links to them. OpenAI does not publish a ranking formula, so the rest is inference from evidence.
On the record
Three statements are public. OpenAI's launch post for ChatGPT search, 31 October 2024, says it “leverages third-party search providers, as well as content provided directly by our partners”. Microsoft's Build announcement, 23 May 2023, says “Bing is coming to ChatGPT as the default search experience”. OpenAI's crawler documentation names OAI-SearchBot as the bot that surfaces sites in ChatGPT search.
What is not public: how much of ChatGPT's retrieval comes from Bing versus other providers or OpenAI's own crawling, how candidates are ranked, and how the model decides which of them to cite. OpenAI's help centre says ChatGPT sometimes rewrites a question into one or more targeted queries before sending them to a search provider; it does not say which provider or how results are weighed.
Microsoft's statement is from May 2023 and described ChatGPT's browsing feature, 17 months before ChatGPT search launched. OpenAI has not named Bing in its own documentation since. Our tracking agrees with the older statement: pages absent from Bing are absent from ChatGPT. Everything below separates what the documentation supports from what we infer from running prompts every month.
Step one
Being indexed by Bing is the prerequisite for ChatGPT search. If Bing has not crawled a page, it cannot be returned as a candidate, and nothing else about the page matters. Most sites have never opened Bing Webmaster Tools. Verifying the site, submitting the sitemap and enabling IndexNow takes an afternoon and is where every engagement starts.
Import from Google Search Console if you have it, or verify with a DNS record. Submit the XML sitemap.
IndexNow is an open protocol, supported by Bing, Yandex, Naver, Seznam and Yep, that pings participating engines the moment a URL changes. Without it, Bing may take days or weeks to notice a new page.
Both must be allowed. A rule that blocks all bots except Googlebot, common on sites built for Google alone, removes you from ChatGPT search.
Vercel, Cloudflare and similar platforms offer one-click AI bot blocking. Make sure it is off for the search crawlers you want.
Search site:yourdomain.com on Bing. The count should be close to Google's. Pages missing from Bing are missing from ChatGPT.
Step two
OpenAI documents three user agents, and each is controlled independently in robots.txt. OAI-SearchBot decides whether your pages can appear in ChatGPT search. ChatGPT-User fetches a page when a user's question needs it. GPTBot collects content for model training. Blocking GPTBot does not remove you from search, and allowing it does not add you.
| User agent | What OpenAI says it does | If you block it |
|---|---|---|
| OAI-SearchBot | “Used to surface websites in search results in ChatGPT's search features” | Your pages stop appearing in ChatGPT search |
| ChatGPT-User | Visits a page when a user asks something that needs it; “not used for crawling the web in an automatic fashion” | ChatGPT cannot read the page live for that user; does not affect search inclusion |
| GPTBot | “Used to make our generative AI foundation models more useful and safe”: training data | No effect on search; your content is excluded from future training |
A sensible default for a brand that wants to be cited: allow OAI-SearchBot and ChatGPT-User, decide about GPTBot on its own merits. The same split exists at Anthropic (Claude-SearchBot and Claude-User for search and fetching, ClaudeBot for training) and at Perplexity (PerplexityBot for search, Perplexity-User for fetching).
Step three
Once a page is a candidate, the model keeps the passages that answer the question with the least work. The best-known controlled study, the 2023 GEO paper, found that adding citations, quotations and statistics raised a source's share of the answer by 30–40 percent, while keyword stuffing did nothing. Our monthly screenshots point the same way.
Timeline
Expect the first ChatGPT citations two to eight weeks after a page is indexed in Bing, for long-tail prompts with few established sources. Competitive prompts such as “best X in Bangalore” take six to twelve months, because ChatGPT already has sources it returns for them. Perplexity usually moves first, within one to two weeks, and is the early signal.
| When | What is usually true |
|---|---|
| Days 1–14 | Site verified in Bing Webmaster Tools; IndexNow live; crawler access fixed; Bing picks up the sitemap |
| Weeks 2–6 | Fixed pages re-crawled by Bing; first Perplexity citations for long-tail prompts |
| Weeks 2–8 after indexing | First ChatGPT citations for long-tail prompts |
| Months 2–4 | Consistent ChatGPT citations for low-competition prompts; brand description corrected |
| Months 6–12 | Citations for competitive category prompts, as third-party mentions accumulate |
We cannot guarantee any of these. ChatGPT's answers vary between sessions, and OpenAI changes retrieval without notice. What we can do is run the same prompts every month and show you the screenshots.
Checklist
Six checks cover most of what decides whether ChatGPT can cite you. Each takes minutes and needs no tool beyond a browser. If any fails, fix it before writing a word of new content: the fastest ChatGPT wins we have seen came from indexing and rendering fixes on pages that already existed.
Is the count close to Google's? Are your key pages there?
Are Bingbot, OAI-SearchBot and ChatGPT-User allowed? Is the platform's AI-bot blocker off?
Is the answer in the HTML without JavaScript? Is the title specific to the page?
Does it describe what you do, where you are and who runs it correctly? Where did the wrong facts come from?
Who is cited? Which third-party pages? That is your mentions list.
Does it name the same entity as your Google Business Profile and LinkedIn page? Does it validate?
The free AI visibility audit runs these checks and 50 prompts for you, with screenshots, in five working days.
Sources
Every factual claim above about how an engine works, or about what a company publishes, links to the source it came from. Each URL was opened and checked on the date shown at the top of this guide. If a source later changes or disappears, the claim should be re-checked, not assumed.
Keep reading
The free audit runs your category's prompts on ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews and shows who each engine cites today. Five working days, screenshots included.