One Question, A Dozen Searches: What Google's Query Fan-Out Means for Your Business

There is a sales pitch doing the rounds at the moment, built on a "leaked Google AI Mode system prompt." It points at something real and then steers you toward buying sixty articles and a stack of backlinks. The underlying idea is worth understanding. The conclusion is worth ignoring. Here is the honest version.

You don't need a leaked prompt for any of this, by the way. Google has documented the mechanism in public since it launched AI Mode. We will use the official version.

What query fan-out actually is

When you ask AI Mode or trigger an AI Overview, Google's systems analyse your question and, if it is complex enough, expand it into multiple related searches across different subtopics and data sources, then assemble the results into a single answer. Google's own description, from its May 2025 Search announcement, is that the AI "breaks down your question into subtopics and issues a multitude of queries on your behalf," exploring the web more deeply than one traditional search could.

Google's patents fill in the mechanics. They describe a model that takes one query and generates variants emphasising different intents: comparative ones ("A vs B"), exploratory ones ("how does X work"), and decision-making ones ("best X for situation Y"). For very involved questions, a Deep Search mode can issue dozens or even hundreds of background queries before answering. Google has said its AI search experiences now serve around 1.5 billion people a month, so this is not a fringe feature.

A simple published example: ask "best sneakers for walking" and AI Mode may also search behind the scenes for "best sneakers for men," "walking sneakers for different seasons," "sneakers for trail walking," and "best slip-on walking sneakers." One question becomes a small research project.

Why this changes the maths for your website

Here is the consequence most businesses have not absorbed. You can rank beautifully for your main keyword and still be absent from the answer, because Google went looking for things you have no page about.

Picture a real buying question: "What's the best CRM for a 20-person roofing company that needs estimates, automated follow-up and QuickBooks integration?" Fan-out might quietly research roofing-specific CRMs, automated lead follow-up, estimate tools, QuickBooks integrations, CRM pricing, and competitor comparisons. A generic "our CRM" page answers maybe one of those. A competitor with useful pages across several may appear again and again as Google builds the answer, even if you outrank them on the single head term.

This also explains something SEOs have noticed: a page sitting on page three of normal results can show up in an AI answer. It wasn't ranking well for the obvious keyword. It was the best match for one of the hidden sub-queries. Google now assesses relevance at the level of individual passages across many queries, not whole documents against one.

So the unit that matters has shifted from the keyword to the decision. Your homepage says what you do. An industry page says who you serve. A comparison page says how you differ. A pricing page says whether you fit the budget. A case study supplies evidence. A technical guide answers the big objection. Each is a separate door into the same answer.

The trap: this is not permission to spam

Here is where the sales pitches go wrong, and where you should be careful.

The tempting reading of fan-out is "so I should publish a page for every possible sub-query." Do not. Google explicitly warns that creating content for every fan-out variation, primarily to influence its AI responses, can fall foul of its scaled content abuse policy. A business that floods its site with sixty near-identical thin articles is not building coverage; it is building exactly the footprint Google's spam systems are designed to catch.

Which is worth sitting with, because the loudest advice right now, including the pitch that prompted this article, is to buy bundles of dozens of articles and a set of authority backlinks. That is selling the thin-page play that Google's own guidance tells you to avoid, dressed up in fan-out language. Fan-out rewards genuine decision coverage and punishes manufactured volume. Those are not the same thing, and the difference is the whole game.

What "decision coverage" actually looks like

Coverage means identifying the genuine parts of your customer's decision and answering them properly, once each, with something only you can provide. For most businesses that is a handful of strong pages, not a content farm:

A clear core category page. Audience or industry pages for the segments you actually serve. Honest comparisons against the alternatives buyers weigh you against. Real pricing information. Original data or customer results that nobody else has. An implementation or technical guide that answers the question that stalls a sale.

The test for each page is simple: does it contribute something Google cannot already find in ten other places? "10 Benefits of Payroll Software" fails that test. "What multi-state payroll compliance actually costs, with our own figures" passes it, and it tends to attract a more serious buyer, because the person reading it has already declared their problem.

Why fan-out breaks how you measure visibility

If one question triggers a dozen searches, "what do I rank for?" stops being a meaningful question. As one analyst put it, an AI Mode answer is a function of synthesis, not of what ranks for a single query. You cannot see fan-out in a rank tracker, because most of the sub-queries are synthetic and never typed by a human.

This is the real reason AI visibility needs its own measurement. To know whether you show up across a decision, you have to test the actual questions buyers ask, watch which of them surface you, and track that over time and against competitors, across the engines people use. That is precisely what SearchScore's Tracker does: it runs the real questions in your category across ChatGPT, Gemini, Claude, Perplexity, Grok and DeepSeek, and shows where you appear, where a competitor appears instead, and how it moves. A single keyword rank cannot tell you any of that.

What to actually do

Map your customer's real decision, the genuine questions they ask before buying, and make sure you have one strong, distinctive page for each meaningful part. Get the fundamentals right first, because Google has confirmed AI Mode and AI Overviews run on its core ranking systems: a page has to be indexed and genuinely useful to be eligible at all. Earn real mentions on credible third-party sources, not bought volume. Then measure where you actually appear across the decision, and fix the gaps.

Coverage, not keywords. Useful, not manufactured. Measured, not assumed.

One note on that "leaked system prompt": it was published by a public prompt archive and appears to be one captured AI Mode configuration, not an official Google document. Its key instructions happen to match Google's published guidance, which is why we relied on the documentation instead. You never needed the leak.