A quiet release from Microsoft this week, flagged by Search Engine Journal, matters more than the model launches around it. It is called Web IQ, and it is a search engine built for machines instead of people.
Web IQ is a set of grounding APIs that let AI agents pull information directly from Bing's index. Instead of returning ten links or a full page, it hands the model passages and structured evidence objects. Microsoft claims sub-165-millisecond responses and lower token cost than crawling, and it respects the same robots rules Bing already honors.
Grounding Is the New Ranking
For twenty years, the game was ranking. Get your page to position one, capture the click. The click is dying, which I have written about more than once. What replaces it is grounding, the process by which an AI model decides which sources to pull into the answer it generates.
Web IQ is infrastructure for that decision. When an agent answers a buyer's question, it is no longer sending them to your site. It is retrieving passages, your passages if you are lucky, and synthesizing an answer. Whether your brand appears depends on whether the grounding layer selected your content as evidence. That is a different target than a blue-link rank, and most brands have not changed targets. I have made the case that nine out of ten brands are invisible in AI, and grounding APIs are the mechanism that decides who stays invisible.
The mental model that helps is to stop thinking about a results page and start thinking about a citation. In the old web, you competed for a slot a human would click. In the grounded web, you compete to be the source a model quotes, often without the human ever seeing your name. Being quoted and being clicked are different games with different rules, and the rules of the second one are still being written this quarter.
Why Structured Evidence Changes the Work
The detail that matters is structured evidence objects. Models do not want your hero image, your navigation, or your 1,800-word preamble. They want a clean, attributable claim they can cite cheaply. Content that is structured, specific, and verifiable gets selected. Content that buries one fact under a narrative gets skipped, because pulling it costs more tokens for less certainty.
This is why the old content reflex fails in AI answers. I argued that AI search will not cite you when your content is built for dwell time instead of extraction. Web IQ makes the cost explicit. The grounding layer literally optimizes for fewer tokens and higher certainty. Write for that, or get left out of the evidence set.
This rewards a kind of writing most marketing teams have trained themselves out of. Lead with the claim. State the number. Attribute the source. Make the unit of value a single, liftable sentence rather than a mood that takes three paragraphs to land. The irony is that this is also better writing for humans, who were skimming for the answer anyway.
The Measurement Gap and the Move
Microsoft is also giving publishers AI citation data in Bing Webmaster Tools, which means you can start to see whether you are being used as a source. That visibility is the first half of the work. The second half is producing content the grounding layer wants: clear claims, with numbers, attributed, structured so a machine can lift one sentence and trust it. Seeing whether you are the source across ChatGPT, Gemini, and Perplexity, not just Bing, is the cross-model picture GEOflux (geoflux.ai) is built to map.
The measurement gap is real and worth naming. Microsoft is showing citation data, Google is testing impression data for AI features, and neither yet gives you the full click and query picture the old search reports did. That is uncomfortable, but it is not an excuse to wait. The brands that learn to read the partial signals now will be fluent by the time the data matures, while the ones waiting for perfect dashboards will start two years behind.
The brands that adapt will treat their site as a source database for AI systems, not only a destination for human clicks. That is a real reframe, and it touches everything from how you write to how you mark up a page. The discovery surface keeps moving, and the work is to be the thing the machine pulls from, not the thing it routes around.
None of this means the open web disappears. It means the open web becomes a supply chain for machines as much as a destination for people, and your content is inventory in that chain. Inventory that is clean, labeled, and easy to verify gets pulled. Inventory that is messy and self-promotional gets passed over, the same way a buyer skips the supplier whose catalog is impossible to read.
Ranking was about being the best link. Grounding is about being the most liftable fact. Microsoft just built the pipe. Decide whether your content fits through it.