Query Fan Out: What It Is, How AI Search Uses It and How to Plan Content for It

Query fan out (Google writes it “query fan-out”) is the name Google gave to the way its AI search features break a question apart before answering it. Someone types one long question; behind the scenes, the system runs a set of narrower searches, reads what comes back and writes a single response with links to the pages it used.

That shift matters for anyone who depends on search for leads. The page that gets cited is not always the page that ranks first for the original question. Often it is the page that answers one specific sub-question clearly. This guide explains the technique using Google’s own descriptions, separates what is documented from what is speculation, and shows how to plan content so your site has a good answer for the sub-questions that matter to your business.

What is query fan out?

Google introduced the term when it launched AI Mode in March 2025. In its announcement of AI Mode, Google described query fan-out as issuing multiple related searches concurrently across subtopics and multiple data sources, then bringing those results together into an easy-to-understand response. Its Search Help page for AI Mode uses almost the same words: the question is divided into subtopics, each one is searched simultaneously, and the results are combined.

Put simply, query fan out is the step between your question and the AI answer. The system decides what it would need to know to answer well, turns each need into its own search, and gathers material for every branch.

Three features define it:

• One question becomes many. A single prompt produces several sub-queries, each aimed at a part of the question.
• The searches run in parallel. Google stresses that they run concurrently, not one after another.
• The results are merged. The user sees one response, usually with links to the sources that support it.

Query fan out means your page competes for a sub-question, not only for the keyword someone typed. A clear answer to one narrow need can earn a citation that a broader page misses.

How does query fan-out work, step by step?

Google has not published the internal details, so any step-by-step account is a simplification. What it has said publicly supports this sequence:

1. The question is interpreted. The model reads the full prompt, including constraints such as budget, location or a comparison between options.
2. Subtopics are generated. The question is broken into the pieces needed for a complete answer. Google’s own example compares sleep tracking across a smart ring, a smartwatch and a tracking mat, which naturally splits into one search per device and one per feature.
3. Searches run concurrently. Each subtopic becomes a query. Google says these can hit the web index and other sources such as the Knowledge Graph, real-world information and shopping data.
4. Supporting pages are identified. Google’s documentation on AI features says that while responses are generated, its models identify more supporting web pages, which lets it show a wider and more diverse set of links than a classic search.
5. The answer is composed. The results are combined into one response, with links to the pages that support specific statements.
6. Follow-ups repeat the cycle. AI Mode is built for follow-up questions, and each follow-up can start a new fan-out with the earlier context in mind.

The important point for site owners sits in steps 3 and 4. Retrieval happens per sub-query. If your page is the best result for one of those narrower searches, it can be pulled into the answer even when it would never rank for the long original question. Our guide on how to appear in AI Overviews covers the selection side of that process in more detail.

Which search systems use query fan-out?

Google is the one company that has named and described the technique in public. According to its documentation, these features may use it:

• AI Mode. Google’s launch post and help page both say AI Mode uses query fan-out.
• AI Overviews. The Search Central page AI features and your website says both AI Overviews and AI Mode may use the technique.
• Deep Search. At Google I/O in May 2025, Google said Deep Search uses the same technique taken further, issuing hundreds of searches and reasoning across them to produce a fully cited report.
• Agentic tasks in AI Mode. Google’s ticket-buying example describes AI Mode starting a query fan-out across sites to compare options with real-time pricing and inventory.

Other AI assistants that search the web, such as ChatGPT search and Perplexity, also retrieve several pages for one prompt and cite them. They do not publish their retrieval steps in the same terms, so it is safer to say the general idea applies to AI search broadly while the specific name and details are Google’s. If your goal is to be named by those assistants as well, our guide on how to get your business cited by ChatGPT covers the signals they rely on.

One more nuance: not every question is fanned out to the same degree. A simple factual question may need one lookup. A question with several conditions, a comparison or a decision behind it is where the technique does the most work, and those are usually the questions that come before a purchase.

Query fan-out vs classic search and keyword research

Classic search maps one query to one ranked list. Keyword research mirrors that: you pick a term, build the best page for it and measure where it ranks. Query fan-out adds a layer in between, and the differences are worth seeing side by side.

Keyword research does not go away. It changes shape: instead of one keyword per page, you research the cluster of questions around a decision and make sure each one has a clear home on your site. If you are still deciding how SEO, answer engines and generative AI search fit together, our explainer on AEO vs SEO vs GEO lays out the three layers.

If you want a second pair of eyes on which questions your site answers today and which ones it leaves to others, a Blueprint Briefing is a short call where we look at that with you.

What does a query fan-out look like? A worked example

Google does not show the sub-queries it runs, so any example is an illustration of the logic, not a log of real searches. Take a question a small business owner might type into AI Mode:

“Should I rebuild my company website from scratch or redesign the one I have, if I don’t want to lose my Google rankings?”

To answer that well, a system would need several separate pieces of information. A plausible fan-out looks like this:

No single page answers all eight well. A site that has a strong website redesign process guide, a clear page on migrating a website without losing SEO rankings and an honest custom website vs WordPress comparison has three chances to be pulled into that one answer. A site with a single “Our Services” page has, at best, one weak chance.

That is the practical meaning of fan-out: coverage of the decision, spread across pages that each do one job well.

The facets most fan-outs cover

Across most service topics, the same facets come up again and again. They work as a checklist: for any important topic on your site, ask whether each facet has a clear answer somewhere.

The examples are illustrations for a generic service. The point is the structure: ten facets, each a likely branch of a fan-out.

Does query fan out change SEO?

Less than the hype suggests, and more than nothing. Google’s position is clear. Its page “AI features and your website” says there are no additional requirements to appear in AI Overviews or AI Mode, and no special optimizations necessary. The same SEO fundamentals apply: crawling allowed, content findable through internal links, good page experience, important content in text form, structured data that matches the visible text and an up-to-date Business Profile.

So fan-out does not introduce a new ranking trick. What it changes is where the competition happens. Each sub-query is ranked on its own, with the usual systems. Your page needs to be a strong result for the narrow question, and the passage that answers it needs to be easy to lift and quote.

Google’s Search Central blog post on succeeding in AI search, published May 21, 2025, makes the same point from the other direction. It asks site owners to focus on unique, non-commodity content, because people in AI search ask longer and more specific questions, plus follow-ups. Generic pages that repeat what every competitor says give the system no reason to pick them for any branch.

Three consequences follow for a business website:

• Depth across a topic beats one perfect page. A cluster of focused pages, linked well, covers more branches.
• Quality signals still decide. Pages that show who wrote them and why they can be trusted are easier to choose. Our guide to E-E-A-T explains those signals.
• Classic rankings still count. Sub-queries are searches. A page that ranks poorly for every narrow question will not be retrieved by fan-out either.

What query fan-out does not mean

Because the term is new, a lot of advice built around it goes further than the evidence. These are the claims we see most often, and what Google’s documentation actually supports.

If you want to know more about the file mentioned in the third row, our guide on what an llms.txt file is explains where it helps and where it does not.

If you suspect your site was built on one of these myths, a Blueprint Briefing is a quick way to find out which pages are doing real work and which are only adding noise.

How to map the fan-out for your own topics

You cannot see Google’s sub-queries, but you can see the questions people ask around a topic, and those are a close proxy. This is the method we use, and it needs no paid tool.

1. Pick one decision, not one keyword. Start with a question that comes right before someone contacts a business like yours, such as “should I hire an agency or a freelancer for my website.”
2. Collect autocomplete variants. Type the main phrase into Google and note the completions. Add “what”, “how”, “vs”, “cost” and “best” in front of it and note those too.
3. Collect People Also Ask questions. Open the boxes on the results page; each opened question reveals more. These are questions Google already groups with your topic.
4. Pull long queries from Search Console. Filter the Performance report for queries of six or more words. Long, specific queries are the ones that look most like fan-out branches.
5. Add the questions your customers ask. Sales calls, emails, chat logs and reviews hold the questions search tools miss, often in the exact words people use.
6. Ask AI assistants for the sub-questions. Ask a few assistants what they would need to know to answer the main question. Treat the output as brainstorming, not data.
7. Sort everything by facet. Group the questions under the ten facets above. Gaps show up quickly.
8. Record the evidence. Next to each question, note where it came from. A question seen in autocomplete, PAA and Search Console is stronger than one an AI suggested.

There is no target number of questions. The right size for a map is whatever covers every facet without repeating itself; when two questions would get the same answer, merge them.

The best source of fan-out questions is rarely a tool. It is the list of questions your customers already ask before they hire you, written in their words.

Query fan out tools and simulators: what can they tell you?

Search for “query fan out tool” and you will find generators, browser extensions and simulators. Most work the same way: they send your query to a language model and ask it to list the sub-queries it would run. Some then check which of those sub-queries your page answers.

These tools can be useful for brainstorming, especially when you are new to a topic. Their limits are worth keeping in mind:

• They estimate. Google does not expose its sub-queries, so no tool can show the real ones for a given search.
• They vary. Run the same query twice and you may get a different list, because the model generates it fresh each time.
• They lack demand data. A generated sub-query may be logical and still never be searched. Cross-check with autocomplete, PAA and Search Console.
• They do not know your customers. They cannot add the questions only your sales calls reveal.

A simple rule works well: use a tool to widen the list, then keep only the questions that show up in at least one real data source or in your own customer conversations.

Where should each sub-question be answered?

Once you have the map, every question needs a home. This is where many sites go wrong: they either cram everything into one page or spin off a thin page for each question. A simple decision table keeps it clean.

Before creating any new page, check that no existing URL already answers the same question. A simple spreadsheet with one primary keyword per URL is enough to keep track. Our guide to what pages a business website should have covers the core pages every site needs before adding topic pages.

How do you write passages AI search can retrieve?

When an AI answer links to your page, it is usually supporting one statement with one part of your text. Writing for that is mostly writing clearly. These habits make a passage easy to retrieve and safe to quote:

• Answer first. Put the direct answer in the first one or two sentences under a heading, then add detail.
• Make it stand alone. A passage should make sense without the paragraph before it. Name the subject instead of writing “it” or “this.”
• Use question headings where natural. A heading that matches the sub-question tells readers and systems what the section answers.
• State entities plainly. Who, what, where. “GoSocial is a custom web design company in Boca Raton, Florida” is easier to use than a clever tagline.
• Prefer specifics. Steps, criteria, named standards and sourced figures are more useful than adjectives.
• Use tables for comparisons. Comparison sub-queries are common, and a table answers them in a compact, readable form.
• Cite primary sources. When you state a fact from a standard, a law or official documentation, name it and link to it.

A quick test: copy one section into a blank document and read it alone. If it still answers its heading, it is ready. If it starts with “As mentioned above,” rewrite the opening. The same discipline applies to the snippet people see in classic results; our guide to writing a meta description shows how to state a page’s answer in one or two sentences.

If you would like us to run that test on your key service pages, bring them to a Blueprint Briefing and we will go through them with you.

How does site structure help with query fan-out?

A fan-out pulls from many pages at once, so the way your pages connect matters. Search systems discover and understand pages through links, and Google lists making content findable through internal links among the fundamentals that apply to its AI features.

The structure that works best is a hub with spokes. The hub page covers the decision at a high level and links to each focused page. Each focused page answers one branch in depth and links back to the hub and to its closest siblings. The anchor text of each link describes what the target page answers, which gives systems a second, consistent label for that page.

Three checks keep the structure healthy:

1. No orphans. Every page in the cluster receives at least one link from another page in the cluster.
2. Descriptive anchors. “Website redesign process” says more than “click here” or “learn more.”
3. Short paths. Important pages are reachable in a few clicks from the home page or a main menu item.

Our internal linking strategy guide walks through building and auditing these links on a business website.

What technical basics does a page need to be retrieved?

Fan-out cannot retrieve a page that search cannot see. Google’s documentation says that to be eligible as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Google Search with a snippet. That one sentence carries several practical checks:

The snippet row works in both directions. Google notes that the same preview controls used for classic results, such as nosnippet and max-snippet, also limit how content appears in its AI features. If you have content you do not want quoted, those controls are the documented way to handle it. Page speed belongs in the same checklist; our guide to Core Web Vitals explains the metrics Google reports.

How does query fan-out apply to service businesses?

Most of the published examples of fan-out are about products: phones, sleep trackers, tickets. Service businesses face the same mechanism with different questions. Someone choosing a contractor, a clinic, a law firm or a web developer rarely asks one short question. They ask about fit, process, risk and proof, often in one long prompt.

For a service website, that usually means the following pages carry fan-out branches:

• Service pages that explain what is included, who it suits and how the process runs, in plain language.
• An About page that states who you are, where you are based and why you can be trusted; our guide to writing an About Us page covers this.
• Pricing factors that explain what drives cost even if you do not publish prices.
• Proof: case studies, reviews and credentials that can be verified elsewhere.
• Guides that answer the questions customers ask before they call.
• Your Google Business Profile, which Google lists among the things to keep up to date for its AI features.

For a GoSocial example: our web development page explains the process from discovery to measurement, while separate guides answer narrower branches such as platform choice, redesigns and site structure. Each page owns one job, and they link to one another.

How do you cover comparison and buying queries?

Comparison is one of the most common branches in a fan-out, because almost every decision has an alternative. Google’s own AI Mode examples are comparisons. If your site never compares options, it has no answer for a whole family of sub-queries.

Good comparison content for a business has four parts:

1. Name the options fairly. Describe each alternative, including the ones you do not sell, in neutral terms.
2. Compare on criteria the buyer cares about. Ownership, maintenance, flexibility, time to launch, risk. A table works well.
3. Say who each option suits. “Option A fits a business that…” is more useful and more quotable than “Option A is better.”
4. Be honest about trade-offs. A comparison that always favors the author reads as an ad and gives systems less reason to trust it.

Examples on our own site include landing page vs website and web agency vs freelancer. Each compares options on stated criteria and ends with who each choice fits, which is exactly the shape a comparison sub-query is looking for.

If you are not sure which comparisons your buyers make before contacting you, we can work that out together in a Blueprint Briefing, using the questions your customers already ask.

How do you measure results from query fan-out?

Measurement is the weakest part of the picture, and it helps to be honest about that. Google’s documentation says sites that appear in AI Overviews and AI Mode are included in the overall search traffic in Search Console, reported in the Performance report under the “Web” search type. There is no separate filter for AI features, and the sub-queries themselves are not shown.

That means you measure topics and outcomes rather than single positions:

• Long queries. Track the number of distinct queries of six or more words that bring impressions to a cluster. Growth suggests the cluster answers more specific questions.
• Pages per topic. Watch whether more pages in a cluster earn impressions, not only the hub.
• Referrals from AI assistants. In GA4, look at sessions whose source is an AI assistant domain, such as chatgpt.com or perplexity.ai.
• Engagement and leads. Google says clicks from results with AI Overviews tend to be higher quality, with users spending more time on the site. Track form submissions, calls and bookings from organic visits to see if that holds for you.
• Manual spot checks. For your most important questions, look at the AI answer every few weeks and note which sources it cites.

None of these numbers proves that a specific sub-query retrieved a specific page. Together, they show whether your coverage is growing and whether it brings business. Our post on website analytics covers the basic reports to set up first.

How do you keep fan-out coverage current?

Fan-out branches change when the world changes: a new regulation, a new platform option, a price shift in your industry, a new question your customers start asking. Coverage that was complete last year can have gaps today.

A light routine keeps it current:

• Quarterly question review. Rerun autocomplete and PAA for your main decisions, and pull new long queries from Search Console.
• Customer question log. Ask whoever answers the phone or email to note new questions as they appear.
• Update, do not duplicate. When a new question fits an existing page, add a section there instead of starting a new URL.
• Show the date. Keep a visible “last updated” date on guides and change it only when the content changes.
• Check facts against sources. When you cite documentation, recheck it during the review; official pages are updated too.

What are the most common query fan-out mistakes?

Most of these come from treating fan-out as a trick rather than as a description of how people now search. The fixes are ordinary good practice: clear pages, clear structure, honest answers. That is also why the same work tends to help in classic results, which is where most SEO, AEO and GEO work still starts.

How GoSocial plans websites for query fan-out

GoSocial is a custom web design company in Boca Raton, Florida, building custom-coded and custom WordPress websites for businesses across the United States since 2016. We optimize websites for traditional search (SEO), answer engines (AEO) and generative AI search (GEO) on one foundation, and fan-out planning is part of that foundation rather than an add-on.

Our web development process runs from discovery and strategy to architecture, design and content, development, testing, launch and measurement. Fan-out shows up at two points. In strategy, we map the decisions your buyers make and the questions behind them. In architecture, we decide which questions get their own page, which become sections and how the pages link, with one owner for each keyword.

Every GoSocial website launches with GA4 and Google Search Console connected and tracks form submissions, phone clicks, WhatsApp clicks and bookings, so you can see which topics bring visits and which bring leads. No one can guarantee an AI citation, and we do not. We build for the metrics and show you the measurements. The domain, hosting account and code belong to you, so another team can maintain the site if you ever choose to.