artigo

The sub-query coverage method, step by step

Camillo Dantas's method for planning and writing content that answers the searches an AI system may run behind one question. Seven steps, success criteria, a coverage matrix and a worked example.

short answerThe coverage method, created by Camillo Dantas, starts from one target question, lists the sub-queries a system could run behind it, groups them by intent, maps which page answers each group and writes a self-contained, sourced and dated block for every gap. It is an editorial discipline. It is not a Google rule and it does not promise citation, rank or traffic.
Neste texto

What is the method, and what is it not?

The sub-query coverage method is a content planning process. It borrows a mechanism that Google documents, query fan-out, and turns it into a working question: when someone asks this, which searches might the system run, and do I have a good answer for each one?

Authorship is clear. Camillo Dantas created the method and presents it as his own. The mechanism behind it, splitting a question into sub-queries, is a market concept taken from Google's public documentation. Keeping the two apart matters. The method does not come from any Google rule, and Google states that there are no additional requirements or special optimizations to appear in AI Overviews or AI Mode.

What it does well: it reduces content gaps, prevents repetition between pages and gives a rule for deciding what to write first. What it does not do: control which sub-queries a system generates, guarantee that any page is used, or replace the quality of the content. Google's own advice remains to create original, expert-led content that goes beyond common knowledge.

Who is it for?

  • Level: intermediate. Anyone who already writes SEO pages can apply it.
  • Prerequisite: access to the site's current content and to real audience questions, such as support tickets, Search Console queries and community threads.
  • Time: two to four hours for one target question the first time.
  • Output: a matrix of sub-query groups against pages, and a set of written answer blocks.

What are the seven steps?

Step 1: choose one target question

Pick a question your audience really asks and that matters to the business. Prefer compound questions that mention a criterion, a comparison or a context, because those are the ones that tend to need more than one search.

Success criterion: the question has a real author (a customer, a student, a lead) and a clear intent: learn, compare, decide, execute or fix.

Common mistake: using a loose keyword such as "site migration". Without the full question, the sub-queries stay vague.

Step 2: list the foreseeable sub-queries

Write 15 to 30 short searches a person would need to run to consider the question settled. Use seven kinds as a checklist: definition, criteria, comparison, cost and time, risk and limits, proof, and procedure. Record the origin of each item.

In the course version of the method, the origin carries a weight from 1 to 3 as evidence that someone asks the question. Search Console and support or sales conversations weigh 3. Community threads, AI answers and related searches or autocomplete weigh 2. Competitor pages, your own deduction and AI brainstorming weigh 1. A list is accepted only with at least three items of real demand and no kind above 40 percent of the items. These are working rules of the course, not market standards.

Success criterion: each sub-query is a sentence that could be typed into a search box.

Honesty rule: label the list as a simulation. Nobody outside Google sees the real sub-queries.

Step 3: group the sub-queries by intent

Join the sub-queries that call for the same answer. Twenty sub-queries usually produce five to nine intent groups. This step separates the method from a risky practice: Google treats creating many pages, one per query variation, mainly to manipulate results, as scaled content abuse. Grouping by intent produces sections, not disposable pages.

Success criterion: two sub-queries sit in different groups only when the answer to one does not serve the other.

Step 4: map each group to a page or section

Build a matrix with one row per group and these columns: page that answers, section, state (covered, partial or gap) and action. The free coverage checklist builds this matrix in the browser and exports CSV.

To decide between a section and a new page, the course version applies three tests: does the answer need more room than a section allows, does the group serve a different reader than the neighboring page, and is there real demand measured in Search Console, support or sales? A group earns its own page only with three yes answers. Otherwise it becomes a section of an existing page.

Success criterion: every group has a state and an owner.

Step 5: write the answer block

The answer block is the smallest unit of the method: a heading phrased as a question or a complete statement, a direct answer in 40 to 80 words, evidence (a source, a date or first-hand experience), a limit and a next step. It must make sense alone, because a system may use only that passage.

This is not about chopping text into small pieces. Google says there is no requirement to break content into tiny pieces for AI to understand it. The block is a discipline of clarity, and the page is still written for the reader.

Success criterion: someone who reads only the block understands the answer and knows where it comes from.

Place each block on the page where it serves the reader. Usually one pillar page covers the whole question and support pages go deeper into groups that need space, such as a long procedure or a comparison. Link them with descriptive anchor text. The pillar reaches each support page in one click, and each support page links back.

Step 7: measure and review

Record the publication date, watch impressions in the Search Console reports and repeat the simulation at 30 and 90 days. Replace outdated blocks. Coverage is measured as covered groups scoring 1, partial groups 0.5 and gaps 0, divided by the number of groups.

Success criterion: a log exists with the date, the question, the state of the matrix and the decision taken.

What does a coverage matrix look like?

Simulation. Target question: "How do I move my site to a new domain without losing traffic?" The groups and states below belong to a fictional site.

GroupSub-queries it containsPage that answersStateAction
A. Concept and decisionwhat a domain migration is; when not to migratearticle "What is a migration"partialadd criteria for when to migrate
B. Preparationmigration checklist; mapping old URLs to newnonegapwrite a "URL map" section
C. Redirects301 or 308; how long to keep them; redirect chainsarticle "Redirects"partialcorrect the redirect types and the time frame
D. Address changeSearch Console Change of Address tool; processing timenonegapwrite an "address change" section
E. Monitoring and errorstracking traffic after the move; common mistakesarticle "Monitoring"coveredreview the date

The coverage index is the sum of the scores divided by the number of groups: (0.5 + 0 + 0.5 + 0 + 1) / 5 = 0.4. Groups B and D are gaps, so they are written first.

Here is one answer block for group C, built from Google's documentation on site moves, last updated on August 20, 2026:

How long should I keep redirects after a migration? Keep permanent redirects for as long as possible, generally at least one year. Google recommends server-side 301 or 308 redirects from each old URL to its new counterpart, and advises against sending many pages to an unrelated destination such as the new homepage, which can be treated as a soft 404. Source: Google Search Central, site move with URL changes, updated 2026-08-20.

What are the limits of the method?

  • It cannot see the real sub-queries. The list is a hypothesis, and any tool that shows "real" sub-queries is estimating.
  • It does not control selection. Whether a system uses your page depends on the index, the competition and each company's own decisions.
  • It does not replace subject-matter review, especially in health and finance.
  • The weights, caps and the three-yes test are working rules of the course. They order the work; they do not predict results.
  • The matrix ages. Review it when the product, the law or the platform changes.

Where can you learn it with exercises?

The full syllabus, with exercises and a practical project, is in the course Query Fan-Out Academy, for R$ 397. The Portuguese reference page, with the full example and more detail, is Método de cobertura. For more on the author, see who Camillo Dantas is.

How many sub-queries should I list?

Fifteen to thirty is usually enough for one target question. The number matters less than the intent groups: five to nine well-answered groups are worth more than fifty loose searches.

Does the method violate Google's scaled content abuse policy?

Not when each answer is useful and original and when sub-queries are grouped into sections instead of becoming one page per variation. The policy targets many pages created mainly to manipulate rankings, without value for the user.

Does the method need a paid tool?

No. A spreadsheet or the free coverage checklist is enough. Third-party tools that list "real" sub-queries estimate what is not public.

Can I use AI to list the sub-queries?

Yes, as brainstorming. Treat the result as a hypothesis, check it against real audience questions and label it as a simulation. The final content needs a source, experience or data that a model cannot invent.

Sources

  1. AI features and your website, Google Search Central (2025-12-10)
  2. Optimizing your website for generative AI features on Google Search, Google Search Central (2026-07-10)
  3. Spam policies for Google web search (scaled content abuse), Google Search Central (2026-08-28)
  4. Site move with URL changes, Google Search Central (2026-08-20)
  5. Expanding AI Overviews and introducing AI Mode, Google (The Keyword blog) (2025-03-05)

Cubra as subconsultas

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