What's new with Web IQ: Ignite 2026 edition
Ignite week has arrived, and itโs an exciting time for Web IQ. Today on the Bing Blog, Jordi Ribas shared some big news. Web IQ is now accessible to all Azure users at new, reduced prices. We’re also enhancing its functionality with features like Auto, Auto Advanced, and fresh data sets including Weather, Places, and Commerce, along with support for health-related applications.
This article serves as the first in an upcoming series of detailed explorations into Web IQ. Here, youโll discover whatโs newly available, how each feature operates, the costs associated, and crucial limitations you need to be aware of before implementing it. Letโs take a look at the key highlights:
- Auto is now the standard method to access Web IQ. Just one request understands what the agent needs, exploring various sources like Web, News, Images, Videos, Local, and Weather, and retrieves only the quoted sections. For complex queries, Auto Advanced performs iterative retrieval attempts.
- Agents receive structured knowledge, moving beyond just web pages. Weather and Places data is available to everyone, while Commerce is entering a private preview with over 40 billion offers.
- Control the sources your agents pull from. Scopes enable you to add or exclude specific sources at no extra cost. With Premium Content, you can enhance retrieval using our list of premium publishers or incorporate your own licensed data alongside web results in the same ranked output.
Capability | What’s new | Status |
Auto | One call, multi-source retrieval with cited passages | Generally available |
Auto Advanced | Iterative retrieval for complex questions | Generally available |
Weather | Current conditions, forecasts, and government alerts | Generally available; through Auto |
Places | Structured information on over 207 million locations | Generally available |
Commerce | Structured product and offer data | Private preview |
Health | Health-focused ranking and source filtering | Generally available |
Scopes | Named include and exclude lists for sources | GA on Web and News; preview on Auto |
Web, Images, Videos, News | Lower pricing | Generally available |
Browse | Lower pricing | Generally available |
If you enjoy a hands-on approach, you can dive right in!
Those attending Ignite, be sure to check out
For everyone else, grab a coffee, set your agents to autopilot, and letโs get started!
When building grounding for an agent, youโre likely familiar with the routine: plan the next inquiry, choose the right sources, gather results, evaluate the evidence, identify any gaps, and search again. Each of these steps adds model calls, tokens, network trips, and costsโnot to mention the orchestration code needed to manage it all. With Web IQ Auto, this whole web-research loop can be streamlined into a single endpoint with two user-friendly modes, allowing you to delegate search effectively.
By using Web IQ Auto, a single inquiry expressed in natural language can represent your agentโs intent. For instance, asking “What is the latest price of crude oil and what factors are influencing it?” enables Auto to choose the correct underlying sources for the inquiry and provide a range of evidence including text, images, videos, weather, or time zones, all backed by citations.
Some inquiries may require more than one retrieval pass for effective answersโthis is where Auto Advanced steps in. It assesses whether the gathered evidence is adequate, adjusts the query to address any gaps, and conducts further retrieval attempts before presenting the cited evidence. With this feature, you can focus less on coordinating retrieval steps and more on how your agent responds and acts. All it takes is setting “mode”: “advanced”.
To further refine your results, you can employ responseFilter to determine which evidence types to include, and maxResponseLength to limit the total size of the evidence (default is max 30K, with higher limits available for Azure managed customers).
Generally, web pages present unstructured prose that agents must dissect. In contrast, Web IQโs passages contentFormat only delivers relevant tokens. In various domains, structured data is often a better match: agents can reason about it immediately, reducing the amount of parsing code you need to write. At Ignite, weโre adding three new knowledge fields to Web IQ: Weather, Places, and Commerce, along with more formats coming soon.
Provide agents structured data for real-world applications using Web IQ Weather, Places, and Commerce (private preview).
Weather is a key factor in many decisions: determining which customer locations a maintenance crew should visit tomorrow, understanding which deliveries might be impacted by severe weather this week, or deciding if an event on Thursday needs to be relocated indoors.
Web IQ Weather incorporates Microsoftโs award-winning weather intelligence into the same Web IQ Auto API previously introduced, providing structured data on current weather, daily forecasts for the next week, hourly forecasts for the next 24 hours, and governmental safety alerts to help agents consider potential disruptions.
This service spans 243 countries and territories, supported by 140,000 observation stations. Government warnings are standardized across 65 countries, precipitation forecasts are updated every 5 minutes, and weather forecasts refresh hourly.
To begin, simply set “responseFilter”: [“weather”] when using Auto, or leave it open for a richer context for your agent.
For location-aware agents, itโs essential to gather accurate details like addresses, phone numbers, and opening hours rather than pulling unstructured data from a directory page. Web IQ Places delivers structured information on more than 207 million businesses, landmarks, attractions, and service-area companies across over 200 markets, utilizing a retrieval and ranking method tailored for Places, resulting in a compact, model-ready schema. Key details encompass addresses, phone numbers, operating hours, and geographical coordinates. Enhanced content is returned where available, varying by location and market.
When it comes to AI-driven shopping, the accuracy of the answers depends on the quality of the product data behind them. Scraping product pages can involve latency, incomplete pricing information, confusing page layouts, and bot restrictions. Web IQ Commerce provides agents with structured product and offer data sourced from over 40 billion options, including product IDs, images, descriptions, prices, availability, and offers from numerous merchants.
Currently, weโre launching two Web IQ Commerce APIs in private preview: the Commerce Search API transforms shopping inquiries into ranked products with details on price, availability, merchant information, and links. Meanwhile, the Commerce Offers API lists all applicable purchasing options for a product across different retailers, including pricing, shipping, availability, and data freshness. You can request access to the private preview here.
Health-focused agents need to rely on credible sources. However, conventional web ranking systems prioritize relevance and freshness over source quality. With Web IQ, we now automatically identify health-related inquiries and apply health-aware ranking to elevate reputable health sources, guided by principles recognized independently by the National Academy of Medicine. You can combine this with Scopes to restrict your sources to government health websites or medical journals.
Web IQ Health is tailored for offering consumer and member health information and education, suited for health plans, digital health providers, and consumer health applications. Itโs important to note that it does not provide personalized medical advice, diagnoses, or clinical decision support.
Note: Web IQ is not covered by Microsoftโs BAA, so please avoid including Protected Health Information (PHI) in queries.
Grounding with open-web data can pull from sources that may not be appropriate for your organization for various reasons, impacting the quality of the outcome. Whether you need to stick to approved sources within a regulated environment, utilize a tenant-wide list to block certain domains, or develop an agent that exclusively searches your own companyโs sitesโScopes allows you to apply constraints directly in the retrieval call with minimal latency implications.
To get started, create a named list of sources to include and exclude directly in the Web IQ portal, or via a REST API, and specify the name as scopeName in your API request. Each scope can feature up to 10,000 inclusive and 10,000 exclusive selectors. Selectors can include domains (matching subdomains), path prefixes like docs.example.org/products, or exact URLs. Note that wildcards, ports, and query strings are not currently supported. In case of conflicting rules, exclusions take precedence over inclusions.
If the scope canโt be applied, the call will fail by default. To allow a more graceful fallback, set failOnMissingScope: false to default to an unscoped search and check scopeApplied in the response to confirm if it was in effect.
Creating and managing scopes via the API requires a separate Microsoft Entra grant (CustomerApiAccess on your Web IQ billing account). Consequently, a search API key can access scopes but cannot create them. Keep in mind that any edits may take a bit to come into effectโtypically around 30 minutes, without a guaranteed service level.
With Zero Query Logging, customers with regulatory or compliance needs can request the suppression of query text from service logs. Operational and billing data will remain, and each request will be reviewed. You can make this request through Web IQ support (aka.ms/microsoft-webiq-support).
Since Microsoft builds and manages its own index, we retain control over content discovery and sourcing, ensuring that every result points back to its original source. Bingbot, our identified crawler, respects robots.txt directives, and since October 30, publishers have been able to set page-level preferences for AI grounding.
The quickest way to create Web IQ-powered agents is by using the Foundry Responses API, where Web IQ now enhances the web search tool, replacing Grounding with Bing and broadening integration throughout Foundry experiences scheduled for rollout in 2027.
Apart from the REST API, Web IQ also accommodates MCP integration, allowing you to configure it as a custom MCP tool within Foundry or any MCP client that supports remote HTTP servers, like VS Code or GitHub Copilot CLI.
If youโre eager to incorporate it into your coding process, simply install with pip install webiq (Python 3.11 or later) or npm install Microsoftโ/webiq (Node.js 22 or later). Both methods accept an API key or any Azure Identity credentials, such as DefaultAzureCredential. The SDKs currently support Web, News, Images, Videos, and Browse; you can call Auto via REST.
Weโre excited to see what you create! Feel free to share your questions and feedback in the comments below, and weโll keep you updated as we reveal more of whatโs in the pipeline.
Share this content:
Discover more from Qureshi
Subscribe to get the latest posts sent to your email.