Microsoft made Web IQ, its web grounding service for AI agents, generally available to every Azure customer on Tuesday, November 17, and cut the price. Web IQ runs on Bing’s own index, which means the crawl controls publishers already manage for Bing now also govern what agents can retrieve. That moves a search-engine decision into the agent supply chain.

Microsoft’s Bing Search Blog, in a post by Jordi Ribas, president of Search & AI, described the launch alongside Microsoft Ignite. Web IQ Auto becomes the default mode. One call handles several retrieval steps over web pages, images, video, news, weather, and places, and the result is trimmed evidence in place of full pages. A second tier, Auto Advanced, evaluates a weak result, refines the query, and retrieves again for harder questions.

Three new knowledge domains arrive with it. Weather supplies current conditions and forecasts. Places offers structured data for over 207 million businesses worldwide, including addresses, phone numbers, and opening hours. Commerce, in private preview, spans over 40 billion structured offers. Microsoft also says Web IQ now detects health questions and favors sources it considers clinically credible, with credibility informed by principles from the National Academy of Medicine.

The publisher-facing details matter most to search teams. Microsoft says Bingbot, the documented crawler that Microsoft says feeds Web IQ, Copilot, and Bing Search, respects robots.txt, so disallowed pages never reach the index feeding the service. NoIndex keeps pages out of the Search index. A third control, announced October 30, lets publishers set AI preferences specific to grounding at the page level. Microsoft calls it industry-first, and that claim is Microsoft’s own.

Those three levers are separate. Blocking Bingbot removes a page from every Microsoft surface at once. A grounding preference, by contrast, is meant to leave a page crawlable and indexed while limiting its use in AI answers. Teams should decide which outcome they want per template, because one robots.txt rule cannot express both.

Local teams have a specific reason to look at Places. Agents using it will pull a business’s address, phone number, and hours from Microsoft’s structured data, not from the business’s own site. Mismatched listings in Bing Places or on aggregators could surface in agent answers, such as a booking assistant quoting the wrong closing time.

The performance claims all come from Microsoft. A test of 500 product-category queries found its commerce offering produced 57 percent more matches than its nearest rival, according to Microsoft’s internal assessment, and the rival is unnamed. Microsoft also open-sourced a FreshSearch benchmark that scores how current results are and how quickly a provider finds new content. Anyone can replicate it from the GitHub code sample, which is the more checkable of the two claims. No independent measurement is cited.

Customer figures are similarly self-reported. Genspark co-founder Ray Zhong said Web IQ brought his voice assistant’s latency down from 6 seconds to 0.6. Publicis Groupe and National Grid are also named as users. Microsoft did not publish the new pricing in the post and points readers to the Web IQ website.

For the next ninety days, audit your Bingbot rules and any page-level AI preferences, then check how your business listings read in Bing Places. Agent builders will treat that data as ground truth, and Web IQ makes Bing’s index a source of answers well beyond Bing’s own results page.

Microsoft’s Bing Search Blog, post by Jordi Ribas, published November 17, 2026.