Query fan-out: how AI search splits one question into many
Short answer Query fan-out is how AI search handles a long question: Google's AI Mode and AI Overviews, and ChatGPT, break it into several shorter searches and gather pages for each part. For your website, it means each service page should also answer the follow-up questions (price, area, timing, options), not that you need a page for every phrasing.
Want help applying this to your business? Talk it through on a free 30-minute call.
Book a callWhat is query fan-out?
It's the step where an AI search tool turns one question into several searches, runs them at the same time, and writes its answer from the pages that come back. Google describes it for its two AI features:
"Both AI Overviews and AI Mode may use a "query fan-out" technique — issuing multiple related searches across subtopics and data sources — to develop a response." (Google Search Central: AI features and your website)
Google's guide to generative AI search gives an example. For "how to fix a lawn that's full of weeds", the extra searches might include "best herbicides for lawns", "remove weeds without chemicals" and "how to prevent weeds in lawn".
ChatGPT works the same way when it searches the web. OpenAI's help page says:
"ChatGPT search typically rewrites your query into one or more targeted queries that it sends those providers." (OpenAI Help Center: Searching the web with ChatGPT)
After reading the first results, OpenAI adds, ChatGPT "may send additional, more specific queries".
What this means for you: the searches the AI runs are short and ordinary, much like what people type into Google. A page gets pulled into the answer when it answers one of those parts well.
How many searches does one question trigger?
No platform publishes a number. Google says AI Mode issues "a multitude of queries simultaneously", and its Deep Search mode "can issue hundreds of searches" (Google, May 2025). Counts you see online, such as "5 to 11 per prompt", come from outside studies and tools, not from Google or OpenAI.
Questions are getting longer, which gives the AI more parts to split:
"While we're seeing growth in both short and long queries in AI Mode, the average AI Mode search is triple the length of a traditional Search query." (Think with Google, May 2026, US data from April 2025 to May 2026)
AI Mode reached Canada, in English, in August 2025. Google said early users were asking questions "two or three times the length of traditional search queries" (Google Canada Blog).
What query fan-out changes in keyword research
Keyword research still works, because the sub-searches are ordinary keywords. What changes is the unit you plan around: the whole question a customer asks, and the parts it splits into, not one phrase per page.
- Write down the full question. Take it from calls, emails and reviews: "My furnace is 15 years old and banging. Repair or replace, and who can come this week?"
- Split it the way the AI would. Service questions usually break into some of these parts: what's wrong, what it costs, which option, who does it nearby, how soon, and why trust them.
- Check each part against real searches. Search Console, autocomplete and Keyword Planner show the words people use for each part. How to use them: keyword research for a local business.
- Write your town in plain text. ChatGPT turns "near me" into a place name. In OpenAI's example, someone near San Francisco asks for good restaurants near them, and ChatGPT "may rewrite your prompt into the search query "top restaurants San Francisco."" If your town and service areas aren't written on the page, that search has nothing on your site to match.
You don't need to chase every phrasing. Google says its AI systems understand synonyms and general meaning:
"This means you don't have to worry that you don't have enough "long-tail" keywords or haven't captured every variation of how someone might seek content like yours." (Google Search Central: Optimizing your website for generative AI features)
Long-tail keywords are longer, more specific searches, like "tankless water heater install cost Mesa".
What should a service page cover for query fan-out?
Every part of the question, in text, so that whichever sub-search the AI runs, the page has a passage that answers it. Google lists "Making sure that important content is available in textual form" among the SEO basics that still apply to AI features (Google Search Central: AI features and your website).
| Part of the question | A sub-search it might create | Put this on the page |
|---|---|---|
| What's wrong | "furnace banging noise" | Common causes, in a few lines |
| Cost | "furnace replacement cost" | A price range and what moves it |
| Options | "furnace repair or replace" | When you'd repair, when you'd replace |
| Who, nearby | "furnace repair Burlington" | Your town and areas served, in one sentence |
| How soon | "same day furnace repair" | Hours, typical wait, emergency calls |
| Trust | "furnace company reviews" | License number, warranty, real job photos, reviews |
The sub-searches are made up to show the idea. No AI tool shows you the list it ran for a given question.
One longer page can hold all of this. Google says you don't need to cut content into small pieces for AI, because its systems "are able to understand the nuance of multiple topics on a page and show the relevant piece to users" (Google Search Central: Optimizing your website for generative AI features). Clear headings help readers find each part. For the rest of the page, see on-page SEO for service pages.
Should you make a page for each fan-out query?
No. Google's guide says so directly:
"While it might be tempting to create 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), doing so primarily to manipulate rankings or generative AI responses in Google Search violates Google's scaled content abuse spam policy." (Google Search Central: Optimizing your website for generative AI features)
A separate page earns its place when it's a different job (drain cleaning and water heaters) or a town where the work or price really differs. For towns, see city pages for service areas. If you already have near-copies, merge them into the main page.
Can you see the fan-out queries that found your site?
Only on Bing, and only partly. Google's report shows where you appeared, not which searches got you there.
| Tool | What it shows | Its limit |
|---|---|---|
| Google Search Console: Generative AI performance report | Impressions in AI Overviews and AI Mode, by page, country, device and date | No query list. A site with too few AI impressions may not see the report at all |
| Bing Webmaster Tools: AI Performance | Grounding queries and the pages they cited | Covers Copilot, Bing's AI summaries and some partner tools. Microsoft says the data is a sample |
| Bing Webmaster Tools: Topics Portfolio (beta) | "Opportunity queries" where other sites are cited and yours isn't | Microsoft says smaller sites may see a short list or no tab at all |
Grounding is the web evidence an AI uses to support and cite its answer. Microsoft defines grounding queries as "the key phrases the AI used when retrieving content that was referenced in AI-generated answers" (Bing Webmaster Blog, February 2026). That's the closest look any platform gives at fan-out. Read your grounding queries as the parts of questions your pages already answer, and the opportunity queries as parts they don't. To set up Bing Webmaster Tools, see Bing Places and Copilot.
Be careful with tools that claim to show Google's own fan-out queries. Google's guide says: "No third-party tool has access to our internal ranking or AI systems."
Example
Illustrative cases with invented numbers.
An HVAC company in Burlington, Ontario (Canadian dollars)
Its furnace page says "call for a quote" and names no towns. The fix: add "most replacements run C$5,000 to C$8,000, depending on size and venting", a repair-or-replace section, the towns served in one sentence, and the license number. One page now answers six parts of the question.
A dental clinic in Phoenix (US dollars)
A vendor proposes 30 pages: "Invisalign cost Phoenix", "Invisalign price Tempe", "clear aligners cost Scottsdale" and so on. That is the per-variation content Google's guide warns about. Better: one Invisalign page with a US$3,000 to US$7,000 range, treatment time, aligners compared with braces, and who isn't a good candidate.
A landscaper checking Bing
Bing Webmaster Tools shows "lawn aeration cost" as a grounding query, cited to a two-year-old blog post rather than the aeration service page. The fix: put the current price on the service page and link the post to it.
FAQ
Is query fan-out the same as long-tail keywords?
No. Long-tail keywords are what people type. Fan-out queries are what the AI types on their behalf, usually shorter and split by subtopic.
Does fan-out happen in AI Overviews or only in AI Mode?
Google says both "may use" it. ChatGPT does something similar whenever it searches the web.
Do I need separate "GEO" or "AEO" work for query fan-out?
Not for Google. Its guide says optimizing for generative AI search "is optimizing for the search experience, and thus still SEO."
Is it different in Canada?
No. AI Mode has used fan-out in Canada since its English launch in August 2025, and both Google's and Bing's AI reports cover Canadian sites.
Different for you?
- You don't know which words your customers use: keyword research for a local business.
- Your clicks fell after AI Overviews appeared: Google AI Overviews for local businesses.
- You want to know whether ChatGPT names you: the ChatGPT recommendation test.
- You serve many towns: city pages for service areas.
- Someone is selling you GEO services: is GEO worth it?.
- You'd rather have pages written and kept current: SEO & GEO Agent.