Query fan-out does not mean you need more pages

Query fan-out has become the most quoted mechanism in AI search, and the most consistently misapplied. Google defines it in its guidance on generative AI features: a set of concurrent, related queries the model generates to request more information and fetch additional relevant results before answering. Google's example is a question about a lawn full of weeds producing side queries about herbicides, chemical-free removal and prevention.

That is a description of how the system gathers material. It reads, to a certain kind of SEO tooling, like a keyword list.

The move everyone makes, and why it fails

The reflex is to enumerate the fan-out queries and publish a page for each. It feels like coverage. It has the shape of something that used to work, when a page per long-tail query was a reasonable bet.

Google addresses this directly and unusually firmly. Its guidance says that creating separate content for every possible variation of how people might search, including fan-out queries, primarily to manipulate rankings or generative responses, violates the scaled content abuse spam policy. It then adds the practical objection: a high quantity of pages does not make a website higher quality or more relevant, and its systems have improved at understanding relevance even where there is no exact match between the query and the page's primary content.

There is a second problem that has nothing to do with policy. Twenty thin pages built from one topic will compete with each other, dilute whatever authority the topic had, and give you twenty pages to keep current instead of one. That cost lands whether or not anything is ever penalised.

What fan-out actually implies

If the model asks several questions around the original one and assembles an answer from what it retrieves, then the page that gets used is the page that can supply more than one of those pieces credibly.

That points at depth on a decision rather than breadth across phrasings. A surrogacy agency does not need separate pages for what compensation is, when it is paid, what affects it and what is not included. It needs one page that answers all four in a way a model can lift a clean passage from, because those four are the fan-out of one question a real person is asking.

It also points at structure that a human would want anyway. Clear headings that name the actual question. A direct answer near the top rather than after four paragraphs of context. Specific figures, dates and conditions rather than ranges hedged into uselessness. None of that is AI formatting. It is what makes a page quotable, and being quotable is the whole mechanism.

Where the effort should go instead

Start from decisions, not queries. List the four or five things a buyer has to settle before they can commit, and build one strong page per decision. The query variations collapse into those decisions on their own.

Then make each page answer its question outright in the first breath. Google's guidance says you do not need to write in a specific way for AI systems, and that is true about phrasing. It is not an argument for burying the answer.

Finally, keep them current and say when you checked. A model assembling an answer from several sources has to choose between them, and specificity plus a visible date is the part of a page a summary cannot fake.

The honest uncertainty

Nobody outside Google can observe which fan-out queries ran for a given search or which passage was selected. Search Console reports that a page appeared, not what it was asked. So the reasoning above is inference from documented mechanism plus ordinary editorial judgment, not a measured result.

What makes it a safe bet is that the failure mode of being wrong is a small number of genuinely good pages. The failure mode of the other approach is a site full of near-duplicates and a policy the platform has already named.

What this means for an operator

Build one page per buying decision, not one page per query variation, and put the answer in the first paragraph. If a proposal recommends a page for every fan-out query, that is the tactic Google names in its spam policy.