Imagine your content answering 48,000 questions in three months and sending 14 people to your website. That is the documented reality one digital agency found when they cross-referenced their Bing AI Performance data with Google Analytics after Microsoft's November 2025 Copilot Search update. Microsoft Copilot operates across a three-layer chain: Bingbot crawls publisher content into the Bing index, Copilot retrieves candidate pages from that index, and a language model synthesises a cited answer - all without the user ever visiting the source. The result is a new category of zero-click consumption at scale, operating across a network of more than 1 billion Windows devices, 420 million monthly active users, and every Bing search query that now triggers an AI-generated overview.
What Microsoft Copilot actually is
Microsoft Copilot is a family of AI assistants running on Microsoft's AI infrastructure, powered primarily by the Bing web index and large language models. It appears in several distinct surfaces: as a standalone assistant at copilot.microsoft.com, as a built-in Windows feature on over 1 billion active PCs, within Microsoft 365 applications (Word, Excel, Teams, Outlook), within Microsoft Edge, and - most consequentially for publishers - inside Bing search itself, where it now generates AI-powered overviews above traditional organic results.
The search integration, called Copilot Search, rolled out quietly in November 2025. It adds AI-generated summaries to the top of Bing results pages in the same way Google AI Overviews do on Google Search, compressing what would have been a click into a contained answer. This is the development most relevant to publishers because Bing had, until that point, been the one major search engine still directing meaningful referral traffic from informational queries.
How Bingbot, Bing, and Copilot interact
Understanding the pipeline matters because it determines what publishers can and cannot influence. Bingbot - Microsoft's web crawler - indexes publisher pages into the Bing index, which spans more than 10 billion web pages. When a Copilot query is submitted, the system retrieves candidate pages from that index, then a language model synthesises and cites an answer in the response.
The practical consequence is that if Bingbot cannot reach or index your content, it will not appear in Copilot responses. If Bingbot can reach it but the content is thin or poorly structured, it is less likely to be selected as a citation. And if it is cited, the current design of Copilot responses means most users read the synthesised answer without clicking through to the source.
Publishers can signal to Bingbot via robots.txt (User-agent: Bingbot). But blocking Bingbot to protest the arrangement also removes the content from both Bing search and Copilot responses entirely - the same unpleasant dilemma that robots.txt presents for every major AI crawler.
The click-through problem: 48,000 citations, 14 clicks
In February 2026, digital agency Whole Whale cross-referenced their new Bing AI Performance report with their GA4 data, and the numbers were stark. Between November 2025 and February 2026, their content was cited 48,000 times across 34 pages in Copilot responses. The number of users who clicked through to their website from those citations: 14. That is a click-through rate of approximately 0.03% - roughly 60 times lower than even a modest display ad.
Their Bing organic traffic had simultaneously dropped by roughly three quarters, almost overnight, following the November 2025 Copilot Search rollout. The pattern mirrors what publishers saw when Google introduced AI Overviews to its own results: citations rise, referral traffic falls, and the two figures do not move together.
The Bing AI Performance report - launched in public preview in February 2026 within Bing Webmaster Tools - shows Total Citations and Cited Pages, but notably contains no click data tab. To discover what those citations actually yielded in visits, publishers have to cross-reference with their own analytics separately. Microsoft has provided the citation visibility; it has not yet joined it to downstream traffic in the same tool.
Where Copilot sits in the AI referral landscape
Despite Copilot's scale at the user level, its contribution to measurable AI referral traffic is modest. As of March-April 2026, Copilot accounted for roughly 4% of measurable B2B AI referral traffic, compared to ChatGPT at 62.6%, Claude at 18.5%, Gemini at 10.6%, and Perplexity at 7.3%. The gap between Copilot's provisioned user base (160 million enterprise-licensed seats, Microsoft 365 bundled) and its referral share reflects the difference between being embedded in a software suite and being actively chosen for web research.
When Copilot traffic does arrive, quality metrics are high. Microsoft Advertising reports that Copilot-assisted customer journeys are 33% shorter on average and 76% more likely to result in lower-funnel conversions than traditional search journeys. Separately, a Microsoft Clarity study across AI traffic channels found AI-referred visits converting at approximately three times the rate of direct and search traffic. The traffic is minimal in volume and high in intent when it arrives - a profile that mirrors what other AI referral sources show.
What Microsoft's advertising integration means
Microsoft has introduced advertising into Copilot responses across Windows and Bing, with "Sponsored" recommendations appearing alongside organic AI answers when a query has commercial intent. Unlike traditional search ads, these sit within the conversational flow rather than beside it. A publisher revenue share model has been announced alongside the Microsoft Publisher Content Marketplace, which counts AP, Condé Nast, Hearst, and Vox Media among its partners - though, as with most AI licensing arrangements, participation is weighted toward larger publishers with existing commercial relationships.
The presence of ads in Copilot responses matters for publishers on two levels. First, Microsoft is capturing ad value at the AI answer layer - the same place publisher content is doing its work - without returning that value to the source sites via traffic. Second, the commercial intent queries that are most likely to trigger ads are exactly the queries that would have generated the most valuable referral traffic under the pre-Copilot model.
What publishers can do
There are four concrete actions publishers can take in response to Microsoft Copilot.
Claim the Bing AI Performance report. Sign in to Bing Webmaster Tools, add and verify your property, and check the AI Performance tab. It shows which of your pages are being cited and how often. Cross-reference with GA4 to understand the conversion rate on any traffic that does arrive. This is the only current source of direct Copilot citation data.
Use IndexNow for real-time indexing. IndexNow is an open protocol, led by Microsoft but supported by Bing, Yandex, and others, that lets a site push URL change notifications to participating search engines the moment content publishes or updates. Because Copilot retrieves from the Bing index, freshness in that index directly affects Copilot answer quality. Submitting URLs via IndexNow is more reliable than waiting for Bingbot to recrawl.
Structure content for citation. Because Copilot synthesises answers from retrieved candidates, pages that contain clear direct answers to specific questions - with explicit definitions, named figures, and structured headings - are more likely to be selected as citations. The same structural signals that help with GEO generally (direct-answer opening paragraphs, question-format headings, FAQ sections) apply here.
Do not block Bingbot without considering the trade-off. Blocking Bingbot removes content from both Bing organic results and Copilot responses simultaneously. For publishers whose Bing referral traffic has already collapsed, the loss from blocking is smaller than it used to be - but it still eliminates any possibility of capturing the high-converting minority of Copilot-referred users.
Frequently asked questions
Is Microsoft Copilot the same as Bing Chat?
Copilot is the successor to Bing Chat, which Microsoft launched in early 2023 and then rebranded. Bing Chat was a standalone conversational interface powered by GPT-4 and the Bing index. Microsoft has since extended the same underlying capability into Windows, Microsoft 365, Edge, and Bing search results under the Copilot name, making it a family of surfaces rather than a single product. For publishers, the most relevant surface is Copilot Search - the AI overview layer now built into standard Bing results pages.
Does blocking Bingbot protect publisher content from Copilot?
Yes, in that Bingbot is the crawler that feeds the Bing index Copilot draws from. If Bingbot is blocked via robots.txt, the content will not be in the Bing index and cannot be cited in Copilot responses. The practical cost is that the content disappears from both Bing organic search and Copilot simultaneously. That trade-off was more painful when Bing organic traffic was substantial; for publishers whose Bing referral traffic has already been displaced by Copilot overviews, the calculus is now narrower.
Does Microsoft pay publishers whose content Copilot uses?
For the large majority of publishers, no. Microsoft operates a Publisher Content Marketplace that includes named partners (AP, Condé Nast, Hearst, Vox Media) under licensing arrangements, but the terms are not public and participation is not open. The default situation for publishers not in that programme is that Bingbot indexes their content under the standard terms of Bing's crawling and indexing, and Copilot uses the Bing index to synthesise answers without additional compensation to individual source sites.
How does Copilot Search differ from Bing AI Overviews?
The terms are used interchangeably by most commentators, and in practice they describe the same user-facing feature: an AI-generated summary appearing above traditional Bing search results. Microsoft introduced the feature in November 2025 under the Copilot Search branding. Some reporting distinguishes between the standalone Copilot assistant (at copilot.microsoft.com or in Windows) and the Bing in-results integration, but both draw on the same Bing index and synthesise answers using the same underlying models.
What does Copilot mean for a publisher's Bing referral traffic going forward?
The Whole Whale data from early 2026 - 48,000 citations yielding 14 clicks in roughly three months, alongside a three-quarters drop in Bing organic traffic - is the clearest published measure of the Copilot Search effect. It is a single case study, not a universal benchmark, and the scale of the impact will vary with query type, industry, and how prominently a publisher's content appears in citation panels. The structural direction, however, is consistent with what happened on Google: AI overviews consolidate answers within the results page, and click-through rates on cited sources fall sharply. Publishers should assume Bing referral traffic will continue to decline and plan content strategy around citation visibility and the high-intent minority of visits that do arrive, rather than volume recovery.