Microsoft separates AI, human search metrics in Bing tools

Image credit: Search Engine Journal

Microsoft Corp. introduced distinct reporting mechanisms within Bing Webmaster Tools Tuesday, separating traditional human search performance from citations generated by artificial intelligence, signaling a significant shift in search paradigms.

This move underscores a re-architecture of Microsoft’s search infrastructure to cater specifically to AI agents, which the company anticipates will generate substantially more queries than human users.

Microsoft launched two separate reports: ‘Search Performance’ for human-driven queries and ‘AI Performance’ for citations within AI contexts. The AI Performance report, which includes a ‘Citation Share’ metric, became publicly available in preview in February and saw expansions in March and June.

Jordi Ribas, who leads Search and AI at Microsoft, said while Bing was initially built for human users, the next era of search is designed for AI agents. Ribas projected that these AI agents could potentially generate 1,000 times more queries than humans.

The company’s re-architected Web IQ infrastructure, built upon the existing Bing index, now serves AI agents by returning passage-level evidence rather than conventionally ranked documents. This system powers the grounding for AI models like Copilot and ChatGPT.

Web IQ operates with a P95 latency of 164 milliseconds and scores passages using a GDSAT (grounding satisfaction) metric, according to Microsoft. The core distinction, the company noted, is that factors determining a page’s relevance for a human searcher differ from those making a passage useful to an AI.

Currently, Web IQ is accessible on a limited, invitation-only basis, indicating a strategic direction rather than immediate widespread deployment.

Both Microsoft and Google are rolling out separate AI-focused reports within their respective Webmaster Tools and Search Console platforms. These reports specifically address the performance of content on each platform’s proprietary AI surfaces.

Industry analysts emphasized that first-party reporting from platforms and third-party measurement tools serve different analytical purposes and should not be conflated when evaluating content performance.

Source: Search Engine Journal