Microsoft Copilot holds 1.3 percent of the AI chatbot market, according to First Page Sage's September 2026 tracking report, the smallest share among ChatGPT, Gemini, Claude, Grok, and Perplexity. It also runs inside Windows, Edge, and Microsoft 365 for roughly 218 million active users, per company data compiled by Business of Apps. Those two numbers describe the same product, and neither tells a brand whether Copilot is worth optimizing for. The real answer sits in how Copilot retrieves and cites content, infrastructure Microsoft has been rebuilding all year.
How Copilot actually retrieves and grounds its answers
Microsoft Copilot answers most consumer queries by drawing on Bing's search index, the same index Bingbot builds for ordinary Bing search results. There is no separate Copilot index. Blocking Bingbot removes a page from Bing search and from Copilot's grounding at the same time, a distinction Microsoft made explicit in a May 2026 clarification of how its systems differ from a standard ranking index.
Microsoft explained the distinction this way, as trade outlet ppc.land reported in May 2026:
An index built for AI grounding differs from a ranking index across five dimensions: factual fidelity, source attribution quality, freshness, coverage of high-value facts, and contradiction handling.
In practice, Copilot pulls a passage from that shared index, runs it through a grounding check against the user's question, and attaches a citation before the passage reaches the reply. Enterprise-built assistants work differently. Microsoft's Copilot Studio, the product companies use to build their own branded assistants, restricts retrieval to a set of domains the company configures and runs its own provenance and semantic-similarity checks before summarizing, according to Microsoft's own documentation. A brand optimizing for the public Copilot in Windows or Edge is optimizing for Bing's general index. A brand whose support bot runs on Copilot Studio is optimizing for a narrower, self-selected one.
Why blocking Bingbot blocks Copilot everywhere at once
Bing has no equivalent of Google-Extended, the crawler control that lets a publisher allow Google search indexing while blocking Google's AI training separately. Bingbot is the only production crawler Microsoft uses for both jobs, so a robots.txt rule written to keep an AI crawler out often removes a site from Copilot and Bing search in the same stroke.
That single-crawler design catches out publishers who copy a blocklist built for OpenAI's GPTBot or Anthropic's ClaudeBot and add bingbot to it, assuming it behaves like a separate AI-only bot. It doesn't. VizibleAI's audit of what each crawler actually reads found the same pattern across most major bots: publishers block user agents by name without checking what traffic that name actually serves, and Bingbot is the one most likely to carry consequences beyond what they intended. The safer move is checking server logs for Bingbot's actual request volume before writing a disallow rule, not copying a template built for a different vendor's crawler.
Microsoft just handed publishers a map of which pages Copilot cites
Bing Webmaster Tools added an AI Performance report in February 2026, giving publishers their first view of which prompts triggered a Copilot or Bing Chat answer that cited their site. A March 2026 update added query-to-page mapping, letting a publisher click a query to see every page cited for it, or click a page to see every query that cited it, according to Search Engine Land's coverage of the rollout.
Microsoft followed that in June 2026 with Web IQ, a grounding API that returns passage-level evidence to AI agents at under 165 milliseconds at the 95th percentile, about two and a half times faster than the benchmark Microsoft cited for competing services, Search Engine Journal reported. Web IQ runs on Bing's existing index and respects the same robots exclusion rules Bing already honors, so it doesn't create a new crawler to block or a new opt-out to configure. What changes is visibility: a brand can now see the exact page Copilot pulled from for a specific prompt instead of inferring it from a referral-traffic spike that may or may not be real. Brands tracking citations with VizibleAI can cross-reference what Bing's own dashboard shows against what's happening across ChatGPT, Gemini, and Perplexity at the same time, since Bing's tool only covers its own engine.
The market-share number is the wrong way to judge whether Copilot matters
Copilot's 1.3 percent share of the standalone AI chatbot market, per First Page Sage, measures how often someone opens an app specifically to chat with an AI. It says nothing about how often Copilot answers a question a user never framed as an AI conversation at all, inside a Word document, an Outlook draft, or an ordinary Bing search box.
Business of Apps put Copilot's active users at 218 million across Windows, app, and web in its most recent company-data compilation, a base far larger than its current chatbot-share ranking suggests. The pattern isn't new. VizibleAI has made the same case about Mistral and DeepSeek, engines with tiny measured share that still show up in enough answers to matter for specific audiences and regions. Copilot's version of the argument runs through distribution rather than geography: it ships inside software hundreds of millions of people already have open, so the marginal cost of a citation showing up there is closer to zero than the cost of earning one in an app someone had to choose to download.
What actually earns a citation inside Copilot
Copilot's citation criteria track Bing's own ranking signals more closely than any other major answer engine's, because it draws from the same index rather than a separate retrieval layer built for AI. Structured data, clear entity names, and content freshness carry weight for the same reason they help in ordinary Bing search, not because Copilot applies a distinct AI-specific ranking model on top.
JSON-LD markup that identifies an organization, product, or FAQ block gives Bing's index a cleaner entity signal to match against a Copilot prompt, the same argument VizibleAI's guide to schema markup for GEO makes for ChatGPT and Perplexity. Freshness matters more for a conversational answer than for a static ranked list, since a wrong or outdated fact is more visible when Copilot states it directly than when it sits in the third organic result.
The retrieval source varies enough between engines that a single technical checklist can't cover all of them:
ChatGPT: OpenAI's own retrieval index, plus Bing's index for web-search queries through a licensing arrangement.
Google Gemini and AI Overviews: Google's search index and Knowledge Graph.
Perplexity: a crawler-built index of its own, supplemented by direct API retrieval.
Claude: Brave Search's live index, queried at response time rather than a pre-built store.
Microsoft Copilot: Bing's index, built and served by Bingbot.
How Copilot's citation mechanics compare to ChatGPT and Claude
Copilot is the only major answer engine whose citations trace back to a general-purpose search index that existed for a different product first. ChatGPT built a retrieval layer specifically for chat. Claude queries a live search partner at response time. Copilot inherited Bing's index, which predates any of this by two decades.
VizibleAI's breakdown of how ChatGPT chooses its sources found OpenAI weighting recency and domain-level trust signals distinct from Bing's own ranking factors, while VizibleAI's Claude playbook found Brave's live retrieval rewards pages that already rank well in a separate index entirely. Bing SEO and Copilot optimization overlap by default, so a brand with strong Bing rankings already has most of the groundwork Copilot rewards. A brand that skipped Bing entirely is the one with catching up to do.
Frequently Asked Questions
Does Microsoft Copilot use a different index than Bing search?
No. Copilot draws on the same index Bingbot builds for Bing search results. Microsoft has confirmed there is no separate crawler or index reserved for Copilot's AI answers, so a page indexed by Bing search is already eligible to be cited by Copilot, and a page blocked from Bing search is blocked from Copilot too.
Can I let Bing crawl my site for search but block it from Copilot?
Not currently. Bing has no equivalent of Google-Extended, the setting that lets a publisher separate search indexing from AI training and grounding access. Bingbot serves both purposes with one user agent, so blocking it removes a page from Bing search and Copilot's grounding at the same time.
How many people actually use Microsoft Copilot?
Business of Apps put Copilot's active users at 218 million across Windows, app, and web, based on company data compiled through mid-2026. That figure covers everyone using Copilot inside Windows, Edge, or Microsoft 365, a much larger base than the 1.3 percent share Copilot holds among people who open a dedicated AI chatbot app.
What is Bing Web IQ and does it change GEO for Copilot?
Web IQ is a grounding API Microsoft launched in June 2026 that returns passage-level evidence from Bing's index to AI agents at sub-165-millisecond latency. It runs on the same index and robots rules Bing already uses, so it doesn't add a new opt-out to manage, but it gives AI agents faster, more structured access to the same content Copilot already cites.
Is it worth optimizing for Copilot specifically, given its small chatbot market share?
Usually yes, because the marginal cost is low. Since Copilot draws on Bing's existing index, most of the work is standard Bing SEO plus structured data, work that also supports Bing search itself. The exception is a brand that has never invested in Bing at all; for them, Copilot's citation mechanics start from the lowest baseline of any major engine.
See whether Copilot is already citing you, or your competitors
Bing's own dashboards only show citations inside Bing and Copilot, not what's happening across ChatGPT, Gemini, Claude, and Perplexity at the same time. VizibleAI tracks brand mentions and citation sources across all of them, so a low share in one engine doesn't stay invisible. Start a free trial to see where your brand currently shows up, and where a competitor is showing up instead.