How does Microsoft Copilot choose citations, and why does Bing decide?

Last updated October 14, 2026. Written by the Third Coast AI team.

Microsoft Copilot chooses its citations by turning a person's prompt into a short search query of a few words, sending that query to the Bing search service, and composing an answer from the pages Bing returns, with the exact search queries it used shown alongside the citations, so a company is cited in Copilot when its pages are indexed and ranking well in Bing for the short queries Copilot writes, and the fastest way to influence that is through Bing Webmaster Tools, which now includes a free AI Performance report showing how often your pages are cited in Copilot and in Bing's AI-generated answers. Microsoft's own documentation describes the mechanism in more detail than any other engine vendor, which makes Copilot the most legible of the engines and the one most companies ignore.

This page sets out what Microsoft has published about how Copilot grounds answers in Bing, what the AI Performance report measures, how IndexNow gets new pages into Bing quickly, and where Copilot sits in a multi-engine plan. It is written for an owner or marketing director whose buyers work in Microsoft 365, where Copilot is already in the toolbar.

Does Copilot search the web, or does Bing?

Bing does, on Copilot's behalf. Microsoft's Copilot web search documentation says that when web search is enabled Copilot "may fetch information from the Bing search service when information from the web helps to provide a better, more grounded response," and it describes the process: Copilot parses the prompt, identifies terms where web information would improve the response, and "generates a search query that it sends to the Bing search service asking for more information." That generated query "is different from the user's original prompt" and "consists of a few words informed by the user's prompt," after which the information Bing returns "is used to compose the response returned to the user."

Microsoft also shows its work. The same documentation says web search query citations appear "in the linked citation section of the Copilot response" and that the section "shows the exact web search queries (derived from the user's prompt) that were sent to the Bing search service," available in the chat thread for 24 hours in Microsoft Copilot Chat. For a company this is unusually useful: Copilot tells you the few-word query it ran, which is the query your page needs to rank for in Bing. We log those queries when we run a client's prompt bank, and they are the closest thing any engine offers to a public list of the hidden searches behind a buyer's question.

Which Bing tools show whether Copilot cites you?

Bing Webmaster Tools, and specifically its AI Performance report, which Bing's AI Performance announcement introduced in public preview on February 10, 2026. The report shows "how publisher content appears across Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations," and it defines its two main metrics as Total Citations, "the total number of citations that are displayed as sources in AI-generated answers," and Grounding Queries, "the key phrases the AI used when retrieving content that was referenced," with citation counts broken down by URL. Any site verified in Bing Webmaster Tools can see it, at no charge, which makes it the only first-party AI citation report any engine vendor currently provides to site owners.

Bing's generative search announcement from July 2024 described the underlying answer layer as combining "the foundation of Bing's search results with the power of large and small language models," generating answers that cite their sources while "retaining traditional search results and increasing the number of clickable links, like the references in the results." Microsoft has also said Copilot uses structured data for grounding, which is why we treat schema as a real lever on Bing even though the evidence that it drives citations elsewhere is weak; the schema page sets out where it helps and where it does not.

How does IndexNow get a page in front of Copilot faster?

By telling Bing directly instead of waiting to be crawled. Bing's IndexNow announcement in October 2021 described it as "a new protocol created by Microsoft Bing and Yandex, allowing websites to easily notify search engines whenever their website content is created, updated, or deleted," with the promise that "by notifying one search engine you will notify all search engines that have adopted IndexNow." The IndexNow documentation adds the mechanics: you host a key file on your domain to prove ownership, you can submit up to 10,000 URLs per post, and a 429 response means you are being rate-limited for sending too much. Most major CMS platforms and CDNs have a plugin or a switch for it, and for a company correcting a wrong fact about itself, submitting the corrected page through IndexNow is the difference between Copilot seeing the fix this week and next month.

Bingbot's handling of JavaScript is the other thing to know. Bing's 2018 post on Bingbot and JavaScript said that bingbot "is generally able to render JavaScript" but that "it is difficult for bingbot to process JavaScript at scale on every page of every website," and it recommended dynamic rendering, serving pre-rendered static HTML to the crawler. That advice is older than Copilot and still sound, since a page Bing renders slowly or partially is a page Copilot may never be handed.

Do Copilot's citations change over time?

About as much as ChatGPT's. Profound's July 2025 volatility study compared identical prompts a month apart on roughly 80,000 prompts per platform and found 53.4% of Copilot's cited domains had changed, against 54.1% for ChatGPT, 59.3% for Google AI Overviews and 40.5% for Perplexity, with the drift growing to 70 to 90% over six months. Copilot's citations move because Bing's results move and because the generated query is a few words that can vary between runs, so the AI Performance report, which aggregates over time, is a better read than any single answer. Cyrus Shepard's May 2026 synthesis of 54 studies scores URL accessibility at 9.5 out of 10 as a citation factor, and on Bing that factor is checked in a tool you can open today.

Should a mid-sized company bother with Copilot?

If your buyers work inside Microsoft 365, yes, and the cost of finding out is an hour. Verify the site in Bing Webmaster Tools, open the AI Performance report, and read the grounding queries; if they match the questions your sales team hears, Copilot is already part of your buyer's process and the report tells you which pages are doing the work. We include Copilot in every prompt bank run because its citations overlap only partly with the other engines, and because the grounding queries it exposes help us write pages that answer the hidden searches the other engines keep private. The tools comparison places the free Bing report alongside the paid trackers, and the audit is where we run the whole set.

• Verify the site in Bing Webmaster Tools and confirm every important page is indexed, since Copilot cannot cite what Bing has not read.
• Open the AI Performance report, export the grounding queries, and compare them with your prompt bank.
• Turn on IndexNow through your CMS or CDN and submit corrected or new pages the day they publish.
• Serve the main text in plain HTML or pre-rendered, following Bing's own dynamic rendering advice.
• Keep Organization schema accurate, because Microsoft has said Copilot uses structured data for grounding.

Frequently asked questions