Bingbot

Bingbot is Microsoft's web crawler for Bing: the client that discovers and fetches pages to build the Bing search index. Its relevance to AI answers is structural — Bing's Webmaster Guidelines describe that same index serving "Bing search experiences, Copilot, and grounding API results."

Microsoft is the only major engine that names an AI assistant in its webmaster documentation, which makes bingbot the AI-adjacent crawler with the least guesswork attached to it.

What does bingbot do, and what does its index feed?

It crawls and indexes for Bing, and Bing's guidelines put three surfaces on that one index. Their first numbered section states that Bing and Copilot experiences "rely on the same core crawling, indexing, and ranking foundation as traditional search" — meaning there is no separate Copilot crawler to allow or block, and no separate index to be admitted to.

Bing Webmaster Guidelines and IndexNow documentation, read 2026-08-21.

The practical read: bingbot access is the entry condition for all three surfaces at once, and the directives that gate them are the same directives that gate ordinary Bing results. Which directive does what to Copilot specifically — including the two legacy ones most templates emit by accident — belongs to optimizing for Microsoft Copilot.

One shared pipeline also changes how you diagnose a problem. When a page is missing from a Copilot answer, the first questions are the ordinary indexing ones — is the URL crawlable, indexed, and un-gagged by a meta directive — rather than anything AI-specific. There is no separate Copilot submission queue to check and no Copilot-only markup to add, because Microsoft documents neither.

How does bingbot find out that a page changed?

By crawling on its own schedule, or by being told. The told path is IndexNow: host a UTF-8 key file (8 to 128 characters, limited to letters, numbers and dashes) at your site root, then send a URL — or a batch of up to 10,000 URLs in one POST — to a participating engine's endpoint. Engines that adopt the protocol agree that submitted URLs are automatically shared with all other participating search engines.

What submission actually buys is worth quoting rather than paraphrasing, because vendors oversell it: per Microsoft's own IndexNow page, using it "does not guarantee that web pages will be crawled or indexed by search engines," and submitted URLs still count against a site's crawl quota. It is a notification protocol, not an admission ticket. The key file, endpoints and response codes are walked through in our IndexNow setup guide; IndexNow shipped on day 1 of our insurance lead-gen build [our data], and we have published no before-and-after crawl-latency measurement of our own, so we credit it with submission and nothing more.

Is bingbot the crawler behind ChatGPT's search results?

Nothing published says so, and we will not infer it. The claim circulates because assistants and search indexes have licensed each other's data before, but no document from Microsoft or OpenAI on our source list describes a Bing index dependency for ChatGPT.

What is documented is that OpenAI runs its own search crawler. OpenAI states that sites opted out of OAI-SearchBot "will not be shown in ChatGPT search answers" — a direct, first-party control that does not route through Bing. So the honest operating rule is to treat the two as separate systems with separate crawlers, and to allow both if you want either. The ChatGPT-side specifics are in how to get cited by ChatGPT.

What would settle the question is a sentence from Microsoft or OpenAI describing the arrangement, and neither has published one. Until that exists, an assistant's index sourcing is a commercial arrangement between two companies, not a property of your site — and the only thing a publisher can act on is the control each company documents for itself.

Should you allow bingbot?

For most sites, yes — and the cost-benefit is different from the AI-crawler question. Bingbot builds a search index rather than a training corpus, so allowing it trades bandwidth for eligibility rather than content rights, and blocking it removes Bing search, Copilot answers and grounding API results in one move.

The against-interest half: Bing's share of search sends many sites very little traffic, and if your case for allowing bingbot rests on visits alone, the numbers may not impress you. We allow it across all 3 of our production builds [our data], and we have no Copilot citation observation to report from those builds — none, in either direction — so we are recommending eligibility, not a measured result. What a crawler's fetch does and does not prove is defined in web crawler; the wider question of what any of this eligibility is worth runs through our generative engine optimization guide.