Search intent is the job a search is trying to get done. AI answers didn't retire it. They run searches of their own before they cite anything, and since June, Bing's AI Performance report puts an intent label on each of those searches. That means I can see the intent behind a citation, not only the citation.
What is a grounding query?
A grounding query is a search an AI answer runs to find its sources. Microsoft's definition, from the launch of the AI Performance report on February 10, 2026: grounding queries show “the key phrases the AI used when retrieving content that was referenced in AI-generated answers”. The report covers Copilot, Bing's AI summaries and select partner sites.
So a person asks, and the AI writes the search. On WireRef, my electrical reference site, one of those searches reads “what size wire for 50 amps”, and Bing counted 488 citations of WireRef for it from Apr 13 to Oct 11, 2026.
On June 16, 2026, Bing added intents: grounding queries are now sorted into “broader categories such as Informational, Commercial, Navigational, Learn and Solve, Research, Creation, Local, and more”. On my sites it used 11 of them. My search intent guide uses the classic four.
Does search intent still apply when AI does the searching?
Yes, and Google says so from its side. Its guide to optimizing for generative AI search, updated July 10, 2026, describes query fan-out as “a set of concurrent, related queries generated by the model to request more information”, and says optimizing for generative AI search is “still SEO”.
A fan-out search has a job to do like any other search. The difference is who wrote it. That's why intent sits at the front of the stage of my system that wins the citation: Intent, information gain, extraction-ready structure. The work that wins the citation.
Which intents do AI answers search for on my sites?
It depends on who the site serves. Find the row closest to your own site.
| Who the site serves | My site | AI searches | Most common label | Next most common |
|---|---|---|---|---|
| For fans | The Crawler's Cookbook 7 months in | 178 | Informational, 74% | Media, 12% |
| For fans | The Quack Report 7 months in | 113 | Live Event, 51% | Informational, 38% |
| For fans | Title Town PDX 7 months in | 68 | Live Event, 62% | Informational, 28% |
| For people mid-task who need a number | WireRef 8 months in | 1,076 | Informational, 55% | Learn and Solve, 23% |
| For people mid-task who need a number | GageRef 8 months in | 159 | Informational, 58% | Comparison, 15% |
| For people mid-task who need a number | RatingPlate 7 months in | 7 | Informational, 43% | Learn and Solve, 29% |
| For people deciding what to buy or use | CheckMyTap at least 8 months in | 187 | Informational, 32% | Research, 31% |
| For people deciding what to buy or use | UpOrbit at least 8 months in | 20 | Commercial, 40% | Informational, 30% |
Both sports sites lead with Live Event, while The Crawler's Cookbook, a fan site about books, leads with Informational. WireRef, GageRef and RatingPlate lead with Informational. CheckMyTap leads with Informational at 32%, with Research at 31%; UpOrbit leads with Commercial at 40%, with Informational at 30%.
Bing Webmaster Tools, AI Performance, grounding-query exports for eight of my public sites, Apr 13 to Oct 11, 2026. A share is of that site's searches in the export. The full count, label by label, is at the end of this post.
Why do sports sites get live-event searches?
Because their answers change by the week. On Title Town PDX, 42 of its 68 AI searches carry Bing's Live Event label (62%), and they drew 89% of its citations in the export. On The Quack Report it's 58 of 113 (51%).
They are rosters, schedules and kickoff times: “portland fire wnba roster 2026”, “cascade softball schedule”, “oregon vs oklahoma state kickoff time”. Each of those answers can change between one game and the next.
So the pages have to change just as often. Title Town PDX's results bot updates the site twice a day, on the schedule in the site's own repository (read Oct 13, 2026). The Title Town PDX case study shows what that bot does.
Does Bing's label match mine?
My search intent guide sorts a search by its words. Here are its rules run over all 1,808 searches, with Bing's labels beside them.
| My rule | It says | Searches it matches | Bing's labels for them |
|---|---|---|---|
| Includes your brand name | Navigational | 1 | Informational 1 |
| Asks "what," "how," or "why" | Informational | 219 | Learn and Solve 112, Informational 76, No label 14, 17 other |
| Includes "best," "top," "vs," or "review" | Commercial | 123 | Comparison 80, Commercial 20, Live Event 11, 12 other |
| Includes "buy," "price," "hire," or action words | Transactional | 0 | none |
Where my rule says Commercial, Bing mostly says Comparison: 80 of 123. Where my rule says Informational, Bing splits it, and 112 of 219 are Learn and Solve, the how-to kind.
My rule has a miss, and it's the word "vs" on a sports site. Of the searches it calls Commercial, 11 carry Bing's Live Event label, all of them on The Quack Report, and each one is a game. “ucla vs oregon on cbs” is a question about a TV channel, and Bing has it right.
Bing's labels miss too. It labels “dante moore injury update” Live Event and “dante moore injury report” Informational, two phrasings of one question. Its topic labels filed “portland fire home schedule” under Public Safety & Emergency Services; the Portland Fire is a WNBA team.
Bing says as much. Its intents and topics come from “evolving AI/ML classification systems” (June 16, 2026), and its newest view warns that “a query may land in a topic you would have labelled differently” (October 13, 2026).
Should you build a page for every AI search?
No. Google's guide is plain about it: creating “separate content for every possible variation of how people might search (for example, by focusing on other queries that people have asked, or fan-out queries)” to manipulate rankings or AI responses “violates Google's scaled content abuse spam policy”.
My own rule is older than AI answers: Every URL should justify being there. On The Crawler's Cookbook I checked before building a separate Halloween page. The costume guide already drew 99.9% of costume and Halloween search impressions, so the Halloween angle went inside the guide instead of onto a page of its own (Google Search Console, query and page data, Sep 1 to Oct 2, 2026).
Which intent never shows up?
Navigational: 0 of 1,808. Bing's list of intents includes it, and not one search behind a citation of my sites carries it. Only one of those searches contains one of my sites' names, and Bing didn't call that one Navigational either.
A person who already knows the site goes straight to it. Google says its AI features already “show links front and center when you're looking to go to a particular site” (October 15, 2026). The page that person lands on is the subject of navigational search intent.
What can't grounding queries tell you?
- It's a sample. The exports hold 166K of the 619K citations Bing counted for these sites over the same days, 27%. Microsoft calls it “a sample of overall citation activity”.
- It's Bing only. Copilot, Bing's AI summaries and select partner sites. It says nothing about ChatGPT or Google's AI Mode.
- It counts citations, not visits. A citation means an answer used the page; it doesn't mean anyone clicked.
- It doesn't join a search to a page. The search export and the page export are separate reports.
- It leaves some of my sites out. This post reads eight of my sites, the ones whose exports hold enough searches to read.
- The labels are a preview. 63 of the 1,808 searches carry no label at all.
Written for this site and first published October 17, 2026. Every number in it is computed from my own exports when the site builds, with its data-through date beside it.