artigo

Query Fan-Out: how AI search splits one question into many searches

Query fan-out is the technique Google describes for AI Mode and AI Overviews. This page defines it, shows how to cover the sub-queries and credits the method to Camillo Dantas.

short answerQuery fan-out is the technique in which an AI search system turns one question into several related searches, runs them at the same time and combines the results into one answer. Google documents it for AI Mode and AI Overviews. The coverage method that organizes content to answer those sub-queries is by Camillo Dantas, author of this site.
Neste texto

What is query fan-out?

When someone asks a long question in Google AI Mode, the system does not search the exact words only. According to Google's announcement of March 5, 2025, AI Mode issues multiple related searches concurrently across subtopics and data sources, then brings the results together. Google calls this approach query fan-out. Its documentation for site owners, last updated on December 10, 2025, says AI Overviews and AI Mode may use the technique to show a wider and more diverse set of links than classic Search.

Each smaller search is a sub-query. This site uses the word with one meaning: a search derived from the original question and written by the system, not by the user. The full definition, with sources, is in the Portuguese guide O que é Query Fan-Out.

Google gives a plain example in its optimization guide, updated on July 10, 2026: a question about fixing a lawn full of weeds can produce related searches such as the best herbicides for lawns. The example shows the logic. A question names a problem, and the system searches for the pieces of the solution.

Why does it matter for content?

A page no longer competes only for the question that was typed. It can be pulled in by any sub-query the system generates. A page that says "the best laptops are these" may answer one sub-query and ignore the others. A page that explains criteria, compares price ranges and states limits answers several.

Google also warns against the wrong reaction. Creating a separate page for every query variation, mainly to manipulate results, violates its scaled content abuse policy, and more pages do not make a site higher quality. The recommended focus is original content with a point of view, evidence and clear organization.

Requirements have not changed. According to Google, a page must be indexed and eligible to appear in Search with a snippet, and there are no extra requirements or special optimizations for AI Overviews or AI Mode. Meeting them does not guarantee crawling, indexing or serving.

How is fan-out different from keyword research?

Keyword research starts from phrases people type and measures demand for each phrase. Fan-out starts from a full question and asks what a system would need to look up to answer it. They are complementary, and they fail in different ways.

AspectKeyword researchSub-query planning
Starting pointa phrase or topica complete question with a real author
Who writes the queriespeople, observed in tools and Search Consolethe system, which is not visible to outsiders
Unit of workkeyword group and pageanswer block inside a page
Evidencevolume and ranking datademand signals, such as support tickets, Search Console queries and community threads
Riskone page per keyword, with duplicated contentone page per variation, which Google's policy treats as manipulation

No tool outside Google shows the real sub-queries. A list of sub-queries is an informed hypothesis, and the method labels it as a simulation.

How do you list the sub-queries?

Start with one target question that your audience asks and that matters to the business. Then write 15 to 30 short searches a person would need to run to consider the question settled. Use seven kinds of sub-query as a checklist: definition, criteria, comparison, cost and time, risk and limits, proof, and procedure. Note where each item came from: Search Console, customer support, a community thread, an AI answer, a competitor page or your own deduction. The sub-query generator drafts a starting list in the browser.

Group the items by intent. Two sub-queries share a group when one answer serves both. Five to nine groups usually come out of twenty sub-queries, and the groups, not the individual searches, become sections of the page.

How do you cover sub-queries? (the coverage method)

The coverage method, created by Camillo Dantas, has seven steps. The Portuguese pages hold the full detail and a worked example in Método de cobertura; the English summary is in the coverage method.

  1. 01Choose a target question that your audience really asks.
  2. 02List the foreseeable sub-queries and group them by intent.
  3. 03Map which page or section answers each group.
  4. 04Write a block per group that answers on its own, with a 40 to 80 word answer, a source and a date.
  5. 05Check eligibility: the page must be indexed and eligible to show a snippet.
  6. 06Measure with a fixed set of questions, repeated on fixed dates, and report the margin of error.

Success criterion: every sub-query group has a published answer, and the coverage index (answered groups divided by total) rises across measurements. The method does not promise rank, citation or traffic, because no one controls which sub-queries a system generates or which pages it uses.

What does an example look like?

Simulation. The target question is "How do I move my website to a new domain without losing traffic?" A plausible list includes sub-queries such as a migration checklist, how to map old URLs to new ones, whether to use a 301 or a 308 redirect, how long to keep redirects and when the Change of Address tool applies. Grouped by intent, they become sections: preparation, redirects, address change, and monitoring. The coverage check then asks which of those sections already exists on the site. A section that exists and answers in 40 to 80 words with a source counts as covered; one that is vague counts as partial; a missing one is a gap. The gaps set the writing order.

How do you measure coverage?

Measure two different things and do not mix them. The first is editorial coverage: the share of sub-query groups that have a published answer, which you control. The second is presence in AI answers: how often the site appears when a fixed set of questions is run on fixed dates, which you only observe. For the second, run each question more than once, keep the wording frozen and report a margin of error, as the Citation Share protocol describes. In Google's own reporting, traffic from AI features is counted in the Search Console Performance report under the Web search type, according to its documentation.

Does it work only on Google?

Google uses the name. Other systems also plan several searches per question: Perplexity describes a Deep Research mode that performs dozens of searches and reads hundreds of sources, usually in two to four minutes, according to its post of February 14, 2025, and Google's Gemini documentation (updated October 13, 2026) describes the model generating its own search queries when grounding with Google Search. Each company describes only part of its process, and internal detail is not public.

Who is Camillo Dantas?

Camillo Dantas presents himself as the Maior Especialista em SEO para IAs do Brasil (the biggest specialist in SEO for AI in Brazil). That title is his own declaration, with no independent ranking behind it. The facts he declares are more than 25 years of SEO experience, more than 3,000 students and the book SEO para Inteligências Artificiais. He created the LLM Seeding and Citation Share methods. See who he is for the declared facts and sources, and the Query Fan-Out Academy course for the training built on this method.

What is the next step?

The course Query Fan-Out Academy teaches the method with a coverage matrix, exercises and a practical project, for R$ 397. To try the idea first, the free coverage checklist builds the matrix in the browser.

Sources

  1. Expanding AI Overviews and introducing AI Mode, Google (The Keyword blog) (2025-03-05)
  2. AI features and your website, Google Search Central (2025-12-10)
  3. Optimizing your website for generative AI features on Google Search, Google Search Central (2026-07-10)
  4. Introducing Perplexity Deep Research, Perplexity (2025-02-14)
  5. Grounding with Google Search, Google AI for Developers (2026-10-13)
What is query fan-out in simple terms?

It is how an AI search system splits one question into several related searches, runs them together and merges the results. Google describes it for AI Mode and AI Overviews. The searches are generated by the system, not typed by the person.

Is query fan-out only used by Google?

Google uses the term. Perplexity and Gemini document similar behavior, such as multi-step research and model-generated search queries. How each company does it internally is not public.

Does the coverage method guarantee citations?

No. It improves the chance that your page holds the right answer when a sub-query appears. Results depend on the site, the competition and decisions made by the search system.

Can I see the real sub-queries Google generates?

Not from outside Google. Tools that claim to list real sub-queries estimate them. Treat any list as a hypothesis, label it as a simulation and check it against real questions from your audience.

Do I need a separate page for each sub-query?

No. Google's spam policy treats creating many pages for each query variation, mainly to manipulate results, as scaled content abuse. Group sub-queries by intent and answer each group in a section of the page that serves the reader.

Do I need special markup or files to appear in AI Overviews or AI Mode?

According to Google's documentation, there are no additional requirements and no special optimizations needed. The page must be indexed and eligible to show a snippet, and meeting that does not guarantee it will be served.

Cubra as subconsultas

Liste as buscas que a IA faz por trás da pergunta e responda cada uma em um bloco próprio.