Query Fan-Out: Your Prompt Is Not the Query

The single most misunderstood concept in AI search, and why most AEO strategies are chasing the wrong question.

At Google I/O 2025, Elizabeth Reid, Google’s Head of Search, walked on stage and quietly blew up how SEO works. She introduced a concept called query fan-out. Most people in the room clapped politely, went back to their desks, and carried on optimising for keywords. A year on, most marketing teams are still doing the same thing.

This is a problem. Because if you’re pouring budget into AEO (answer engine optimisation) on the assumption that the prompt your customer types is the query Google actually searches for, you’re working from a broken map. I’ve provided affordable SEO services for 15 years, and the brands that get their heads around this concept now are going to have a serious head start over the ones that don’t. In this post I’ll explain what query fan-out is, why it breaks everything you think you know about ranking, and what to actually do about it.

The prompt your customer types is not the query Google actually runs.

What is query fan-out?

Query fan-out is the technique AI search engines use to decompose a single user prompt into multiple related sub-queries, run those sub-queries in parallel, and then synthesise one answer from the combined results. The fan-out bit is literal: one question goes in, and the system fans it out into a spread of related searches behind the scenes.

Say someone types “best running shoes for flat feet” into Google’s AI Mode. The AI doesn’t just search for that exact phrase. It breaks it down into sub-queries like “top stability running shoes,” “running shoes for overpronation,” “flat feet arch support 2026,” and half a dozen others. Each sub-query pulls different pages. The AI stitches all those sources together into a single answer.

Google itself confirms this in its own developer documentation: 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. Research from Surfer SEO analysing over 173,000 URLs found that 68% of pages cited in AI Overviews aren’t in the top 10 organic results for the original query. Let that sink in. Your page can rank number one for your target keyword and still be invisible in the AI answer sitting above it.

How the Google query fan-out technique actually works

Here’s Reid at I/O, in her own words: Google’s systems recognise when a question needs advanced reasoning, call on a custom version of Gemini to break the question into different subtopics, and issue a multitude of queries simultaneously on your behalf. That’s the mechanism. It’s not conceptual. It’s running every time someone asks a complex question.

A few things are worth understanding about how this works in practice:

• A typical fan-out generates somewhere between 3 and 12 sub-queries, depending on how complex the original prompt is. Simple factual questions like “capital of Spain” usually don’t trigger fan-out at all. Complex or open-ended questions almost always do.
• The sub-queries aren’t stable. The same Surfer study found that only about 27% of fan-out queries stay consistent across repeated searches. Run the same prompt twice and you’ll likely see different sub-queries. This is part of why AEO is so hard to measure the way SEO was.
• The mechanism isn’t exclusive to Google. ChatGPT, Perplexity, Claude, Gemini, Copilot, Grok, they all use some form of query decomposition. The specifics vary, but the principle is the same: one prompt in, many queries out, synthesis back out to the user.

This is why Lily Ray describes query fan-out as the AI “extrapolating what the next steps might be when searching for this search.” The system isn’t passively matching your content to a query. It’s actively predicting what a thorough answer should cover, then going to find it.

Your page can rank number one for your target keyword and still be completely invisible in the AI answer sitting above it.

Why “the prompt is not the query”

This is the mindset shift the whole industry needs to make. I’ve been saying this to clients a lot recently, and the brands getting ahead are the ones taking it seriously.

Classic SEO worked on a one-to-one logic. You picked a keyword. You wrote a page to match it. Google matched the page to the query. Done. Query fan-out smashes that logic. The user’s prompt is now just the starting point. Google, or ChatGPT, or Perplexity, takes that prompt and asks a whole list of other questions on the user’s behalf. Your content needs to satisfy those questions, not the one on the screen.

So when you sit down to write a piece targeting “best CRM for small businesses,” you’re not really competing for that phrase anymore. You’re competing to be the source the AI trusts enough to cite when it fans out into “CRM pricing under £50/month,” “CRM with simple onboarding,” “CRM for teams of five,” and a dozen other sub-questions you never saw. If your content only addresses the surface query, you’re fighting the wrong battle.

So is SEO replaced by AI? No. But it’s not the same game.

Every week someone asks me if SEO is dead. It isn’t. If anything, the fundamentals matter more. Surfer’s study found that pages ranking for fan-out queries are 161% more likely to be cited in AI Overviews. Translation: if your page isn’t already showing up in Google’s index for the sub-queries the AI generates, you’ve got zero chance of being the source it synthesises from.

So classic SEO signals (authority, technical health, content quality, backlinks) are still the foundation. They’re just no longer the whole building. What’s changed is what you need to rank for. It’s no longer one keyword. It’s the whole cluster of related sub-questions the AI might fan out into. Which means the old approach of “one page, one keyword” is dead. Topical authority has replaced keyword targeting as the core strategic unit.

The iceberg beneath every AI answer

There’s a second layer to this that most AEO tools are completely missing. Tom Rudnai at Demand-Genius recently published some research that I think is genuinely important. His team ran hundreds of prompt clusters through ChatGPT across 14 B2B categories, and found that

• At the awareness stage of a buyer journey, 0% of AI responses invoked any external retrieval. No search. No citation. The AI answered entirely from its training.
• At the consideration stage, the same pattern held. 0% retrieval. 0% citation.
• Only at the conversion stage (“which CRM should I buy?”) did the AI actually go looking for sources, and even then only 48% of the time.

Source: Demand-Genius, Dark AI study (2026). Hundreds of prompt clusters across 14 B2B categories run through ChatGPT.

In other words, the vast majority of AI interactions that influence a buying decision happen before any citation ever occurs. Rudnai calls this Dark AI, and it’s a good name for it. Most of the conversations shaping how your brand is perceived are invisible to the tools we’re using to measure AEO. We’re watching the tip of the iceberg and pretending it’s the whole thing.

Most of the conversations shaping how your brand is perceived in AI never produce a citation at all.

The practical implication: if your only AEO goal is “get cited when someone asks which product is best,” you’re competing for the 16% of prompts where citations actually happen, after the AI has already largely made up its mind. The real battle is upstream, in the exploratory phase, where the AI is still figuring out what your category looks like and who the credible players are.

What to actually do about it

Right. So if the prompt isn’t the query, and most of the influence happens in conversations you can’t see, what does a sensible strategy actually look like? Here’s where I’d focus:

Build topical authority, not keyword pages

Stop thinking in terms of individual pages targeting individual keywords. Start thinking in terms of content clusters that cover a subject in genuine depth. If you want to show up when the AI fans out into 12 sub-queries around your topic, you need to be a credible source across all 12. This is what we’ve always banged on about at Wrise: small brands can outrank household names by owning a topic deeply. Query fan-out makes this more important, not less.

Create content that actually adds something

Rudnai’s research keeps coming back to a concept called information gain. Content that contributes new knowledge, original research, specific perspective, real expertise, rather than rewording what’s already out there. AI models are trained on summaries. They can produce summaries themselves. What they can’t produce is genuine insight from someone who’s actually done the work. This is why

• E-E-A-T isn’t a checkbox, it’s the whole strategy. Surfacing your team’s real expertise onto your website is now the single highest-leverage content move you can make.
• Generic AI-written blog posts are worse than useless. They dilute your authority and give the LLMs nothing to cite.
• One strong piece of original research beats 50 summary articles.

Get mentioned on sources AI already trusts

Reddit’s share of AI Overview citations grew by 450% between March and June 2025. G2, Trustpilot, industry publications, and expert LinkedIn posts are heavily over-represented too. If you’re only publishing on your own website, you’re giving the AI one signal. You need to show up across the sources it already treats as authoritative.

Stop measuring the wrong things

Citation tracking is useful but limited. It only captures that tiny slice of prompts where the AI actually searches. It tells you nothing about the much larger pool of interactions where the AI answered from memory (and either mentioned you or didn’t). Don’t make budget decisions based on a metric that’s watching 16% of the game.

The bigger picture

Query fan-out isn’t a hack. It’s not a trick you can game. It’s a fundamental change in how AI-powered search works, and it rewards the exact opposite of the quick-win mindset. Brands that win in AI search are the ones with deep, consistent, genuinely expert content across a topic. Brands that pump out thin AI-generated listicles are actively damaging their position.

This is a long game. It was always a long game, but query fan-out makes it a slightly different long game. If you want to understand where this all sits in the broader picture of how search is changing, I’ve written about that too in SEO in 2026 and the online search shift.

The short version: the brands that treat their team’s knowledge as a content asset are pulling ahead. Everyone else is optimising for a query Google never actually ran.

Is your website invisible?

Got a team full of experts but nothing on your website to show for it?

That’s exactly the kind of problem we love solving at Wrise. We work in partnership with businesses to surface the knowledge your team already has and turn it into the kind of content that builds real authority across the sub-queries AI actually cares about.

No contracts, no setup fees. Get your free SEO review →

Affordable SEO Services for Small Businesses 🚀

Outsmart your competition with a low-cost SEO package that delivers real results. Stop dreaming. Start ranking!  

✅ No Contracts

✅ No Set Up Fees

✅ Free Review

Get Your Free SEO Audit

Get a free SEO review of your website! 🧐

Want to boost your website’s search engine rankings and organic traffic? Tell us about your project to get started. We’ll provide a free preliminary review and SEO action plan.

Name *

Email *

Phone Number

Website URL *

How Can We Help?

Boost My Rankings >

Written by Rob Truslove

Rob is the founder of Wrise and a specialist in SEO and search strategy. He has over 15 years’ experience helping more than 500 brands grow their online visibility and performance.

With an academic background in creative writing and storytelling, Rob combines narrative-led content with data-driven search strategy to help websites rank, engage and convert. Over his career, he has worked with household brands including P&O, The AA, Trevor Sorbie and Purple Bricks, as well as ambitious SMEs looking to scale quickly.