Query Fan-Out Does Not Mean You Need 100 Pages

Query fan-out expands a complex question into related searches. It helps a search system gather information needed to answer different parts of the request.

For content planning, use those related questions to understand the reader’s needs. Group questions that belong together, and create a separate page when it serves a distinct purpose with enough substance to stand on its own.

The exercise below helps you turn a question set into a useful page plan.

What query fan-out actually means

Google describes query fan-out as a process where a model generates related queries to gather information needed for a user’s broader question. That makes sense. A complex question often contains multiple information needs.

If somebody asks:

“What’s the best way for a small law firm to improve local visibility while reducing PPC dependency and showing up in AI search?”

The system may need information about:

• local SEO,
• law-firm marketing,
• paid search,
• organic traffic,
• AI search,
• reviews,
• technical site quality,
• and perhaps geography.

The retrieval system can gather relevant sources for those needs. The important word is sources. Not necessarily separate pages manufactured for every sentence variation.

Why one page per subquery breaks down

Prompts are effectively infinite.

Consider:

• AI search for law firms
• AI search optimization for law firms
• GEO for law firms
• ChatGPT visibility for lawyers
• how can lawyers appear in AI results
• AI SEO for attorneys
• AI optimization for personal injury firms
• technical AI SEO for law firms

Should those be eight pages?

Probably not. They may all reflect the same underlying subject.

If you build one page for every phrase, you create:

• duplication,
• keyword cannibalization,
• maintenance debt,
• thin differentiation,
• conflicting canonicals,
• weak internal architecture,
• and a deeply unpleasant editorial calendar.

Google now explicitly warns against generating separate content for every possible fan-out variation primarily to manipulate generative Search. We agree. Rare moment of peace on the internet.

Build a topic system instead

A strong architecture has:

• a canonical parent concept,
• supporting subtopics,
• clear page roles,
• and contextual relationships.

For this Mithril cluster:

• AI Search Optimization is the broad pillar.
• How AI Search Finds Sources handles retrieval.
• Query Fan-Out handles this specific concept.
• Internal Linking explains the architecture connecting the resources.
• Entities and Evidence handles machine understanding.
• Measurement handles observation.

Those pages overlap conceptually. They do not duplicate each other’s job. That is the target.

One URL should have a reason to exist

Before creating a page, finish this sentence:

“This page deserves its own URL because...”

Good answers:

• It serves a distinct user intent.
• It owns a durable topic.
• It contains a substantial body of unique information.
• It represents a distinct service.
• It represents a distinct entity.
• It solves a problem that would make another page unwieldy.
• It has a meaningful conversion path.

Weak answers:

• Semrush showed another keyword.
• ChatGPT suggested it.
• The phrase contains “near me.”
• We needed something to publish Thursday.

Query coverage is not page count

A strong page can cover multiple related questions when they belong together. Google’s current guidance explicitly says its systems can understand nuance across topics and synonyms and that there is no need to create content for every query variation.

This should be liberating. You can write an actually useful article. Imagine.

A page about robots.txt for AI can naturally answer:

• What is OAI-SearchBot?
• What is GPTBot?
• Does blocking GPTBot block ChatGPT search?
• What is Google-Extended?
• Can robots.txt enforce blocking?
• Do AI agents obey robots.txt?

These are related questions. They strengthen one canonical resource. Breaking them into six 500-word pages would make the site worse.

When a subtopic does deserve its own page

Separate the content when the information need becomes meaningfully independent. Our AI training article deserves its own URL because it is not merely crawler syntax.

It requires:

• business-model analysis,
• publisher distinctions,
• risk tradeoffs,
• policy recommendations,
• and a separate strategic decision.

The JavaScript article deserves its own URL because rendering is a substantial technical problem with many implementation patterns. The llms.txt article deserves its own URL because it has enough hype, technical context, and evolving standards to justify dedicated treatment.

The test is depth and independence. Not whether another keyword exists.

The query fan-out content model

Instead of creating pages by prompt, create them by information object.

For example, a law-firm marketing cluster might include:

• Law Firm SEO
• Local SEO for Law Firms
• Law Firm PPC
• Legal Website Technical SEO
• AI Search for Law Firms
• Law Firm Analytics and Attribution

Those are meaningful disciplines. Within them, answer the fan-out questions naturally. Then connect them. This gives retrieval systems multiple useful sources without filling the index with near-duplicates.

Internal links become more important

When pages represent distinct pieces of a larger subject, internal links explain the relationship. The anchor text should tell the reader why the destination matters.

The link becomes part of the explanation. This is how human navigation and machine context align.

What about long-tail keywords?

Long-tail research is still useful.

It reveals:

• language,
• specific pain points,
• questions,
• commercial intent,
• and content gaps.

The mistake is assuming every row in a keyword export requires a URL. Use query research to improve coverage. Not multiply pages.

What about location pages?

Location intent can legitimately justify separate URLs. A law firm with real offices in Phoenix and Tucson may need distinct location pages because each represents a real entity and local intent.

A company creating:

• Phoenix SEO
• Mesa SEO
• Tempe SEO
• Scottsdale SEO
• Gilbert SEO
• Chandler SEO

with the same paragraph and city name swapped is solving a different problem. Usually badly. Entity reality should drive the architecture. Not find-and-replace.

What about service + industry pages?

These can also be legitimate when the combination changes the information. SEO for law firms can be materially different from SEO for ecommerce.

Different:

• search behavior,
• regulations,
• local signals,
• content models,
• conversion flows,
• competition,
• tracking,
• and sales cycles.

But if the page is simply the generic SEO service page with “law firm” inserted twelve times, it has not earned independence.

Again:

the page needs a job.

How query fan-out changes content research

It should make research broader, not thinner.

Instead of asking:

“What exact phrase should this page rank for?”

also ask:

“What adjacent facts would somebody need to make a good decision?”

For a technical SEO service page, that might include:

• crawlability,
• rendering,
• indexation,
• site architecture,
• structured data,
• performance,
• migration risk,
• analytics,
• and remediation planning.

You do not need nine pages automatically. You need enough useful coverage and a clear architecture where deeper topics deserve their own resources. This is research as information design.