Much of the advice on ChatGPT search repeats one rule: if a page is not in Bing, ChatGPT will never find it. OpenAI has not written that rule anywhere.
What OpenAI does document is narrower and more useful. ChatGPT search sometimes works with outside search providers, Microsoft among them, and OpenAI also keeps its own store of web content. It names two conditions a site has to meet to be eligible, only one of which most guides mention.
This guide sets OpenAI's own words beside the advice that gets repeated, then turns what is documented into checks a team can run.
Key takeaways
- OpenAI's help center says ChatGPT search "sometimes partners with other search providers." It lists Microsoft and Shopify through their privacy policies. It does not say Bing indexing is required.
- OpenAI states that eligibility needs two things: allowing its search crawler, OAI-SearchBot, and confirming that the website host or content delivery network (CDN) "allows traffic from OpenAI's published searchbot IP addresses." The second check is rarely covered.
- When ChatGPT search turns to a provider, it typically searches a rewritten version of the prompt, not the question as typed. It may send follow-up queries after reading results, share approximate location with providers, and use saved memories in the rewrite.
- OpenAI says ChatGPT ranks results on "multiple factors" and that "Placement is not guaranteed." No public page says how much Bing contributes or how often a partner is used.
Does ChatGPT search use Bing?
Sometimes, by OpenAI's own account, and Bing is not the only source. The table below sets each documented statement beside the version most often repeated.
| Topic | What OpenAI's pages say | What is widely repeated |
|---|---|---|
| Outside search engines | It "sometimes partners with other search providers" (help center) | Bing is the engine behind every ChatGPT search |
| Which providers | The help center links the Microsoft and Shopify privacy policies; the 2024 launch post cites "third-party search providers, as well as content provided directly by our partners" | Bing alone |
| An OpenAI index | Exists: its offline web search page refers to "OpenAI's indexed and cached web content" | Rarely mentioned |
| Eligibility | OAI-SearchBot allowed, plus a host or CDN that lets OpenAI's search crawler addresses through | Indexing in Bing |
| Order of results | "Multiple factors intended to help users find relevant, reliable information," with "Placement is not guaranteed" | Bing rank decides which sources get cited |
The repeated version is not baseless. Microsoft is a documented partner, so a page Bing cannot see may lose one route into an answer. But OpenAI describes several routes and does not rank them, and no OpenAI page makes Bing indexing a condition of appearing.
One more caution: at least one published guide quotes the help center as saying ChatGPT uses providers "including Microsoft Bing." No archived copy of the page from April 2025 to October 2026 carries that wording. Until mid-2026 the page did list Bing by name, as a provider whose privacy policy applies; by September 2026 the same link read "Microsoft privacy statement."
What does a site need to be eligible for ChatGPT search?
OpenAI's help center gives two conditions in one sentence: "allow OAI-Searchbot to crawl the site and confirm that the website host or content delivery network allows traffic from OpenAI's published searchbot IP addresses."
The first is well known. The AEO optimization checklist covers which OpenAI crawler does what, and why blocking GPTBot for training does not remove a site from ChatGPT search.
The second is rarely covered in published advice. OpenAI publishes the list of server addresses its search crawler uses. Bot protection at a CDN or firewall can turn that traffic away even when robots.txt allows it, and the site owner sees nothing wrong.
OpenAI's own list of reasons a page goes missing points the same way. Its offline web search page names these:
- The site blocks crawling through robots.txt or similar controls.
- The site uses "CDN or bot-blocking behavior."
- The page requires login or personalization.
- The content depends heavily on scripts or dynamic loading.
- The page is "new, rarely accessed, or low-signal."
- The site structure makes deeper pages harder to discover.
Two smaller facts sit on OpenAI's crawler page. A site that blocks OAI-SearchBot will not appear in ChatGPT search answers, but "can still appear as navigational links." And after a robots.txt change, "it can take ~24 hours" for search to adjust.
Why does ChatGPT cite different sources for the same question?
Partly because the search behind the answer is usually not the question as asked. OpenAI's help center says ChatGPT search "typically rewrites" the prompt "into one or more targeted queries" before sending them to providers. After reading the first results, it "may send additional, more specific queries."
The rewritten queries can also differ by person. ChatGPT collects general location from a user's IP address and may share it with search providers. "If memory is enabled, ChatGPT may use relevant saved memories when rewriting a search query."
Two buyers typing the same question can therefore trigger different searches and see different sources. A single screenshot of one answer says little about where a brand stands.
Checking a brand's citations takes repeated runs under fixed conditions, read as a range. The AEO tracking guide covers how much those numbers move on their own and how to read them.
What can a team check this week?
Each item below comes from an OpenAI page, except the last.
- OAI-SearchBot is allowed in robots.txt, and server logs show it arriving.
- The CDN and firewall let OpenAI's search crawler through. Bot-protection rules match OpenAI's published address list, and logs show no blocked requests from it.
- Key pages load their main text without heavy scripts and sit no more than a few clicks from the homepage.
- Pages that matter are not behind a login or a personalization wall.
- Bing Webmaster Tools, a sitemap, and IndexNow (a way to tell Bing a page has changed) are set up. This is a low-cost hedge on a documented Microsoft partnership, not a documented requirement.
What gets a page chosen once it is eligible is a separate question. The guide to getting mentioned in ChatGPT answers covers the content and coverage side. The comparison of Gemini, AI Overviews, and AI Mode covers how Google's engines choose sources.
What does OpenAI not say about ChatGPT search?
The limits matter as much as the rules. OpenAI does not say how often ChatGPT uses a search partner, how much Bing contributes, or which factors its ranking weighs.
The offline web search page describes a workspace feature; it is cited here only as OpenAI's statement that its own index exists, not as how consumer ChatGPT searches by default. The help center blocks automated reading, so the quotes here were checked against archived copies of the pages from August and October 2026.
Zumi, an AI Search Intelligence Platform, tracks which brands ChatGPT mentions and which sources it cites, run after run, so an access fix can be read against a stable baseline. Book a demo to see it on a brand's own questions.
Sources
- OpenAI Help Center, "Searching the web with ChatGPT", read from the 12 October 2026 archived copy.
- OpenAI Help Center, "Offline web search for ChatGPT workspaces", read from the 15 August 2026 archived copy.
- OpenAI, "Overview of OpenAI crawlers", checked 13 October 2026.
- OpenAI, "Introducing ChatGPT search", 31 October 2024, read from the 2 October 2026 archived copy.