AI Visibility · 1 July 2026

Google's 'Query Fan-Out' Explained, and How to Write Pages That Catch Every Version of a Question

For twenty years, search optimisation taught a simple habit: pick the phrase your customer types, then build a page around it. One keyword, one page, one target. A plumber wanted "emergency plumber Leeds", so the page said "emergency plumber Leeds" a few times and hoped to rank.

That habit is now quietly out of date, and the reason has a name: query fan-out. When someone asks an AI tool a question, it does not run that one phrase and read the top result. It breaks the question apart into many smaller searches, runs them all, and assembles an answer from the best sources for each piece. Write for the single headline phrase and you catch one of those searches. Write for the whole cluster and you catch a dozen.

This guide explains what query fan-out is, in plain terms and backed by Google's own words, then spends most of its length on the part that actually changes your Monday morning: how to write pages that answer every version of a question, not just the obvious one. It is a writing method, not a technical manual.

If you want the broader picture of how to show up specifically in Google's AI features, our guide to appearing in Google AI Overviews covers that ground and introduces fan-out briefly. This page goes deep on the writing craft fan-out demands, wherever the answer appears.

What query fan-out actually is

Query fan-out is the technique an AI search tool uses to answer a question by running several related searches at once, rather than a single search for the exact words you typed.

This is not a guess about how the technology works. It is how Google describes its own system. When Google introduced its AI Mode, it explained that the feature "uses a query fan-out technique, breaking down your question into subtopics and issuing a multitude of queries simultaneously on your behalf" (Google, AI in Search, May 2025). The tool does the work of asking many questions so the person only has to ask one.

A worked example makes it concrete. Suppose someone asks an AI tool:

"Who's a good accountant in Sheffield for a small limited company, and what should I expect to pay?"

Behind the scenes, the tool does not search for that whole sentence. It fans the question out into something like:

• accountants in Sheffield
• accountants who handle limited companies
• typical fees for small-company year-end accounts
• reviews of accountancy firms in Sheffield
• what a small limited company needs from an accountant
• qualifications to look for in an accountant

Then it gathers the strongest sources for each strand and writes a single answer that stitches them together. The business that gets named is the one whose online presence answered the most of those strands clearly. A firm with a page that only says "Sheffield accountants: trusted, professional, reliable" answers almost none of them. A firm with clear, separate material on limited-company accounting, on fees, on what the process involves, and a stack of recent reviews, answers most of them, and walks into the answer.

What this changes about how you write, not just how you optimise

The old keyword model treated a search as a single target you aimed at. Fan-out replaces that target with a scatter of smaller ones. Three practical consequences follow, and all three are about content rather than code.

You are no longer competing for one phrase. You are competing to be the clearest answer to a cluster of related questions. Ranking first for "Sheffield accountant" is worth far less than being the source that answers six of the ten sub-questions fan-out generates around that topic. Breadth of genuine, specific coverage beats repetition of one keyword.

Vague, brochure-style copy now actively costs you. When a tool fans a question into specifics (fees, process, qualifications, area covered), it needs pages that answer those specifics. A page written in personality and adjectives gives it nothing to lift for any of the strands. The more concrete your writing, the more strands you catch.

Sub-questions you have never explicitly targeted become winnable. Because fan-out searches for things like "what does the process involve" and "how long does it take", a page that plainly answers those everyday questions can be pulled into an answer even though you never thought of them as "keywords". The questions your customers actually ask on the phone are now optimisation targets.

There is reassurance buried in this for a local business. Fan-out tends to preserve the things that matter most to you. When the AI-visibility firm Profound studied 10,000 prompts across ChatGPT, Perplexity and Copilot to see which parts of a question survived the rewrite, it found that location was "almost always preserved" across all three engines, and that prompts framed as "best [product or service]" were the most stable of all (Profound, What AI Engines Actually Search For, 2026). Profound sells AI-visibility tools, so treat its figures as vendor research rather than neutral fact; here the finding lines up with the wider pattern that these tools hold tightly onto geography and "best in [place]" intent. The practical reading: "best emergency electrician in Bristol" is a question shape fan-out handles cleanly and keeps your town inside, so the sub-questions you most want to win are the ones the technology is least likely to throw away. For the wider picture of how to get named in those answers, see our complete guide to AEO for local businesses.

How to find the sub-questions before you write

You cannot answer the cluster of questions if you do not know what is in it. Mapping the cluster is the step most people skip, and it takes an afternoon, not a budget. Here is how to build the list for any service you offer.

1. Write down the headline question first. The plain thing a customer would ask: "who installs heat pumps near me?" That is your starting point, not your destination.

2. Add the five practical strands fan-out almost always generates. For any local service, the tool reliably searches around:

• What it is / what's included: what does the service actually cover?
• Cost: how much does it cost, or what is the range?
• Time: how long does it take, how quickly can you come out?
• Trust: qualifications, accreditations, guarantees, reviews.
• Area: which towns or districts you cover.

Write the customer's version of each. "How much does a heat pump cost to install in [town]?" "How long does heat pump installation take?" "Do I need an MCS-certified installer?" These are your sub-questions.

3. Mine the questions you already get asked. Your inbox, your call notes and your quote conversations are full of the exact sub-questions fan-out is running. Every "but does that include..." and "how soon can you..." is a strand to answer on the page.

4. Read the "People also ask" boxes and AI answers yourself. Search your headline question on Google and in AI Mode, and note every follow-up question and sub-point the answer raises. Those are literally the strands the engine cares about for your topic.

5. Check what a named competitor covers that you don't. If a competitor is named in AI answers and you are not, read their site against your sub-question list. The gaps are usually strands they answer and you don't.

By the end you should have, for each service, a headline question and eight to fifteen specific sub-questions. That list is the brief for the page.

How to write the page so it catches the whole cluster

Now turn the list into a page. The goal is a page that answers the headline question at the top and every sub-question plainly somewhere on it, in language a machine can lift cleanly.

Lead with a direct, quotable answer

Open the page by answering the headline question in one or two plain sentences, before any throat-clearing. "We install air-source heat pumps across Sheffield and Rotherham, typically in two to three days, from around £8,000 to £13,000 depending on the system." That single sentence answers four strands at once (service, area, time, cost) and is exactly the kind of statement an AI can quote directly. Compare it to "Welcome to our heat pump page, where quality meets passion", which answers nothing.

Give each sub-question its own clear answer

Work down your sub-question list and make sure each one is answered explicitly on the page, ideally with the question as a heading and a direct answer beneath. You do not need a wall of text; you need a clear statement per strand. A reader skims it easily and a machine reads it cleanly. The independent research behind "generative engine optimisation" found that pages improved their visibility in AI answers when they added concrete material like relevant statistics, cited sources and direct quotations, the opposite of vague filler (Aggarwal et al., GEO: Generative Engine Optimization, KDD 2024).

Use the customer's words, not your industry's

Fan-out searches in the language people actually use. If customers say "boiler service" and your page says "annual thermal maintenance", you miss the match. Name services the way customers say them, ask the questions the way customers ask them, and keep the trade jargon for where it genuinely adds precision (like a named certification).

Put a frequently-asked-questions block at the foot

An FAQ section is the most natural home for the smaller strands: the "do you...", "how soon...", "what if..." questions that do not deserve their own page but absolutely get searched. A genuine FAQ block, where each question is one a real customer asks and each answer is direct, maps almost perfectly onto how fan-out decomposes a query. Do not stuff it with questions nobody asks; answer the real ones.

Split, don't cram

Fan-out rewards specificity, which means one giant page trying to cover every service in every town answers each sub-question worse than a focused page would. Give each important service its own page and each area you serve a real page with genuine local detail. A page that tries to be about everything is quotable for nothing.

Keep the facts consistent with the rest of your web presence

The fees, areas and qualifications you state on the page should match what your Google Business Profile, directories and listings say. Fan-out pulls from many sources and cross-checks them; contradictory facts make the engine less confident and less likely to name you. Decide your numbers once and keep them identical everywhere.

A worked before-and-after

To make it concrete, here is the same service page written two ways.

Before (one headline keyword, brochure voice):

Driving Lessons in Norwich
Welcome! We are Norwich's friendly, professional driving school, passionate about getting you on the road. With our experienced instructors and bespoke approach, your success is our priority. Get in touch today!

It targets one phrase, answers no sub-questions, and gives an AI nothing specific to lift. It catches one strand of the fan-out at best.

After (written for the cluster):

Driving Lessons in Norwich
We provide manual and automatic driving lessons across Norwich and the surrounding villages, with DVSA-approved instructors. Lessons are £38 an hour, or £360 for a block of ten. Most learners take 40 to 47 hours of tuition before passing, in line with the national average. We cover NR1 to NR16 and can usually start within a week.

How much do driving lessons cost in Norwich? £38 per hour, or £360 for ten hours booked together.
Manual or automatic? Both. Automatic lessons suit learners who want to pass faster on a narrower licence.
How many lessons will I need? Most people need 40 to 47 hours; we will give you an honest estimate after your first assessment lesson.
Are your instructors qualified? Every instructor is DVSA-approved and fully insured.
Which areas do you cover? All of Norwich and the surrounding villages, NR1 to NR16.

The second version answers cost, transmission type, duration, qualifications and area as separate, plain, quotable statements. When fan-out breaks a learner's question into those strands, this page is a clean source for each of them. Same business, same facts, written for how AI search actually reads.

Common mistakes

• Optimising for one phrase and ignoring the cluster. The headline keyword is the start of the work, not the whole of it.
• Brochure language over plain facts. Fan-out searches for specifics; adjectives answer none of them.
• One mega-page for every service and town. Specific pages catch specific strands; a catch-all catches nothing well.
• Using industry jargon instead of customer wording. If the words on your page are not the words customers use, the match never happens.
• Inventing an FAQ nobody asks. The block works only if the questions are real ones your customers actually ask.
• Letting facts drift between your page and your listings. Contradictory numbers across the web make every engine less sure of you.
• Writing for the machine instead of the reader. The same plain, specific, honest writing serves both; do not keyword-stuff in the name of fan-out.

Frequently asked questions

What is query fan-out in simple terms?

It is the way AI search tools answer a question by running many smaller, related searches at once instead of one search for the exact words you typed. Google describes its AI Mode as "breaking down your question into subtopics and issuing a multitude of queries simultaneously on your behalf" (Google, May 2025). The tool answers a cluster of sub-questions and assembles one reply.

How is writing for fan-out different from old-style SEO?

Old SEO aimed a page at one keyword. Fan-out means a page is judged on how many of the smaller sub-questions around a topic it answers clearly. The shift is from repeating one phrase to genuinely, specifically answering the whole cluster of related questions a customer's query contains.

Do I need special tools to do this?

No. The method is research and writing: list the sub-questions for each service (using your own customer conversations and the "People also ask" boxes), then answer each one plainly on a focused page. Tools can speed up the research, but the core work is done by hand.

Does fan-out keep my location, or does it get lost?

Location tends to survive. A study of 10,000 prompts found location was "almost always preserved" when AI tools rewrote and fanned out a query across ChatGPT, Perplexity and Copilot (Profound, 2026). For a local business that is good news: the strand you most want to win is the one the technology is least likely to drop.

How long should a page be to cover the cluster?

Long enough to answer the headline question and each genuine sub-question clearly, and no longer. That is usually a focused page per service or area, not an enormous catch-all. Depth of specific answers matters more than raw word count.

Is this only relevant to Google, or to ChatGPT and Perplexity too?

The principle applies across AI search tools, because they all decompose and rewrite questions to some degree. The writing method (answer the cluster plainly, in the customer's words) helps you wherever the answer is assembled. Our guide to how each AI engine picks its sources covers where the engines differ.

Where to start

If you take one habit from this guide, make it this: before you write or rewrite any service page, list the sub-questions a customer's request really contains (cost, time, what's included, qualifications, area, plus the things they actually ask you on the phone), and make sure the page answers every one of them in plain, quotable sentences. That single change turns a page built for one keyword into a page built for the whole fan-out.

If you would rather see which sub-questions are surfacing your competitors and not you, that is part of what a QBiz AI Visibility audit does. We run the fanned-out questions your customers ask across the engines, show which versions you already answer and which you miss, and hand you the gaps to fill in order. It is the honest starting point before you rewrite a single page.

Get your AI Visibility audit →