How Microsoft Copilot chooses the pages it cites

Copilot is built on Bing

Bing's guidelines cover Bing search, Copilot and grounding results together, and say Copilot relies on the same crawling, indexing and ranking foundation as traditional search, so good search practice also supports being cited.[1]

Bing finds pages through IndexNow submissions, XML sitemaps, crawlable internal links and links from relevant websites, and asks site owners to use IndexNow when pages are added, changed or removed.[1]

What makes a page likely to be cited

Microsoft says pages are more likely to be selected for grounding and citations when they stand on their own: facts and definitions are explicit, important information is visible on the page, and key information comes early rather than after a long introduction.[1]

It also asks for one main topic per page, clear and consistent names for people, organisations, products and places, accurate structured data that matches what's visible, and content kept up to date.[1]

What reduces your chances

The NOARCHIVE directive stops content being used in Copilot answers, and NOCACHE limits Copilot to the page's address, title and snippet. NOINDEX keeps a page out of Bing and Copilot entirely.[1]

Microsoft lists practices that reduce visibility, including large-scale automatically generated content without editorial review, keyword stuffing, link schemes, misleading structured data, and content written to manipulate AI answers.[1]

Microsoft is clear that none of this guarantees citations in AI answers, and that a fall in clicks doesn't always mean a loss of visibility, since content can appear as citations without a click.[1]