Microsoft has unveiled Web IQ, a new set of grounding APIs designed to help AI agents access and use information from Bing’s search index more efficiently. The company describes Web IQ as a “search engine for AI systems” – a tool built not for human users, but for AI models that need reliable information while completing complex tasks.
What Is Web IQ?
Unlike traditional search engines that return full web pages, Web IQ delivers relevant passages and structured evidence objects. This means AI models receive only the information they need, rather than processing entire pages of content.
According to Microsoft, this approach reduces the number of tokens AI models must process, helping lower costs and improve response speed. The company summarizes the benefit as: “Fewer tokens in, better answers out, lower cost per call.”
To support this, Microsoft rebuilt its retrieval system on top of the Bing search index, redesigning how content is indexed, ranked, and selected for AI use cases.
Performance and Efficiency Claims
Microsoft says Web IQ performs strongly in both speed and information quality.
The company uses a metric called Grounding Satisfaction (GDSAT) to evaluate whether retrieved information is fresh, relevant, and trustworthy. Based on a test involving 3,000 sample queries, Microsoft claims Web IQ achieved higher scores than competing solutions.
In terms of speed, Microsoft reports response times below 165 milliseconds at the 95th percentile (P95), which it says is nearly 2.5 times faster than competing systems tested across five data centers.
The company also claims that Web IQ maintains answer quality while requiring fewer tokens, even as the amount of retrieved information increases.
Publisher Controls Remain in Place
Microsoft says Web IQ respects the same robots.txt rules and publisher preferences already honored by Bing. The company is also collaborating with the Internet Engineering Task Force (IETF) and other industry organizations to help develop standards for how AI systems access and use web content.
How the Technology Works
Web IQ relies on Microsoft’s open-source embedding model to identify relevant content, along with additional ranking and selection models that determine which passages are most useful.
Microsoft notes that these models are trained specifically for AI reasoning workflows rather than optimized for standalone benchmark performance.
To support fast searches across massive datasets, Web IQ also extends DiskANN, Microsoft’s technology for searching large-scale indexes without loading all data into memory.
Why It Matters
Web IQ represents the latest step in Microsoft’s broader effort to connect AI systems with web content more effectively.
Earlier this year, Bing Webmaster Tools introduced AI citation reporting, followed by features that mapped grounding queries to cited pages and previewed Citation Share. Those tools help publishers understand how AI systems use their content, while Web IQ provides the mechanism AI systems can use to retrieve that content in the first place.
Because Web IQ focuses on delivering individual passages rather than complete web pages, the factors that make content useful for AI grounding may differ from those that influence traditional search rankings.
What’s Next?
Microsoft is currently accepting expressions of interest from potential users, but several details remain unknown. The company has not yet announced pricing, API documentation, or a timeline for general availability.
It also remains unclear whether Microsoft’s existing AI products, such as Copilot or Bing Chat, already use Web IQ behind the scenes or whether the new platform will operate separately from those services.
You may love to read:
