While large language models are trained on data up to a specific cutoff date, they lack access to information published after this point. To ensure their AI assistants and agents deliver current and accurate responses, there is a need for them to tap into real-time web data. Microsoft has been at the forefront of this development with its Grounding with Bing Search feature, enabling agents to incorporate up-to-date web information in their answers generation.
At Build 2026, Microsoft unveiled Web IQ, a novel suite of AI-native grounding APIs tailored for AI applications and autonomous agents. This innovative tool empowers AI apps to base their responses on data derived from web pages, news articles, images, and videos. Notably, Web IQ leverages Bing’s extensive global index but has been re-engineered to cater specifically to AI apps and agents.
Instead of merely providing simple web search links, Web IQ delivers relevant passages and structured evidence objects. This approach minimizes the amount of extraneous context sent to an AI model, thereby reducing token usage and enhancing the quality of grounded responses. In a blog post announcing the development, Knut Risvik, Distinguished Engineer of Search & AI at Microsoft, revealed that Web IQ underwent a comprehensive redesign across indexing, retrieval, ranking, passage selection, and orchestration. It employs Microsoft’s Harrier embedding models and technology derived from DiskANN, its large-scale nearest-neighbor search system.
In a significant stride towards enhancing gaming performance, Microsoft’s internal tests have revealed that Web IQ delivers a remarkable latency of under 165ms (p95) across five data center regions. This speed is approximately twice as fast as the fastest alternative solutions under similar configurations, marking a substantial leap in efficiency.
Moreover, Microsoft highlighted that Web IQ not only offers superior speed but also maintains quality using significantly fewer tokens compared to its competitors. However, the company refrained from disclosing the specific rival services mentioned in their announcement.
Interestingly, Web IQ will inherit Bing’s publisher and web ecosystem policies, encompassing support for robots exclusion protocols, publisher controls, and content access preferences. This integration aims to ensure a seamless transition and compatibility with existing systems.
While Web IQ is not yet available to the general public, developers and organizations with interest can register for it through the provided link.
It’s worth noting that since Web IQ is designed for AI agents and multi-step workflows, Grounding with Bing will continue to serve as an accessible entry point for traditional search experiences for existing customers.
News Source: Neowin
Comments
Be the first to comment.