Query Fan-Out Mapping
Google’s AI Mode breaks one query into several before answering it, and a page built for the literal wording covers only one of them. Ivan Sp decomposes a target query into the sub-questions a real searcher asks before, during, and after it. Google’s AI Mode fans a query into related sub-queries and assembles an answer from pages covering them, so a page built for the literal query text alone covers a fraction of the demand.
The Before, During, and After Journey
Sub-questions get grouped by where they sit in the searcher’s task. Before questions cover what someone needs to understand first. During questions cover the decision itself. After questions cover what happens next, including what to expect and what it costs.
Six to ten adjacent questions get generated per target query. Each one then gets checked against the existing page.
Assigning Every Sub-Question
Each sub-question receives one of four assignments:
- Keep. The page already answers it adequately.
- Improve. The page touches it, and the answer is thin or buried.
- Add section. The page should answer it and currently does not.
- Separate page. The question carries enough distinct intent to deserve its own page.
The fourth assignment matters most. Loading every adjacent question onto one page overloads it with unrelated intents, which dilutes its topical focus and often creates cannibalization against pages that already exist.
Fan-Out as a Build Order
The map that comes out of this process doubles as a content plan. Questions marked “add section” become the outline for the current page. Questions marked “separate page” become the next pages in the queue, already scoped and already justified by demand.
Local intent gets flagged separately from question type. “Implant cost” and “implant cost Edmonton” carry different intent despite sharing a topic, and merging them into one page serves neither well.
Checking Coverage Against Rankings
Before new sections publish, the sub-questions get checked against terms the site already ranks for. Keywords in the upper quarter of impressions count as protected territory and never get diluted by a competing new section. The full method sits on the AI search optimization service page.
Decompose One Query Yourself
I suggest an exercise that takes about ten minutes.
- Search your most valuable query in a private browser window.
- Write down every People Also Ask question, expanding each to reveal more.
- Add the related searches shown at the foot of the page.
- Check how many of those your own page answers directly.
Most pages answer two or three of a dozen. The rest are being answered by whoever did, and that is who gets quoted.
Where the Standard Keyword Advice Breaks Down
How many questions does your target query really contain?
- The usual advice
- Target the query as written, since that is the phrase people actually type.
- What holds up in practice
- The typed phrase is the entry point rather than the whole question. A surface that expands a query assembles its answer from pages covering that expansion. A page matching only the literal wording covers a fraction of what is being asked.
Consider a page built precisely for one commercial phrase. The optimisation is textbook and the page ranks respectably for that exact term.
The surface answering that query is pulling from eleven sub-questions. The page addresses two, and competitors supply the other nine.
Ivan Sp maps the fan-out before building. Some sub-questions belong on the same page and some need their own, and the map decides which.
Optimising for the literal phrase produces a page that ranks and contributes almost nothing to the assembled answer buyers actually read.
The answer your buyers see is being assembled from competitors’ pages right now. Each sub-question you do not cover is a paragraph somebody else supplied.
Four Things Commonly Assumed About Query Targeting
Which of these describes your current keyword targeting?
I meet these four assumptions whenever a page is built for a single phrase.
| Commonly assumed | What actually holds |
|---|---|
| One query means one question. | A conversational surface expands a query into related sub-queries. The assembled answer draws on pages covering that expansion. |
| Fan-out mapping needs special tooling. | People Also Ask prompts and related searches are visible on any results page, and both expose part of the expansion. |
| Every sub-question needs its own page. | Some belong on the same page as sections. Splitting a page per sub-question produces thin pages competing with each other. |
| Covering more sub-questions always helps. | Coverage matters where the sub-question has demand and the site can answer credibly. Padding a page with unrelated answers dilutes it. |
| Rankings show whether coverage is working. | Position measures the page against a query. Whether a sub-question is answered anywhere on the site is a separate audit. |
Every sub-question you leave uncovered is a place a competitor gets quoted instead. The map costs an afternoon and it decides what the next six months of content does.
A Worked Example: One Query, Eleven Sub-Questions
Consider a commercial query decomposed into its journey. The mapping below is illustrative, written to show how assignment works. No client engagement is described here.
| Sub-question | Stage | Assigned to |
|---|---|---|
| What the service actually involves | Before | The service page, as its opening section |
| Whether the problem needs this service at all | Before | The service page, as a qualifying section |
| How long the work takes | During | The service page scoping section |
| What it costs to scope | During | The service page scoping section |
| How it differs from the adjacent service | During | A dedicated comparison page |
| What happens if it is done wrong | After | A dedicated blog post |
| How to verify the work was done | After | The service page verification section |
I assign rather than splits. Five of the seven rows belong on one page, and two earn pages of their own.
Fan-out mapping is planning a conversation rather than answering a question. Somebody asking one thing has six more queued behind it.
The assignment column is the deliverable. A list of sub-questions with nowhere to put them is a keyword list with extra steps.
How a Query Decomposes
A single query carries several dimensions the searcher cares about but never types. Take “best dentist in Edmonton” as a worked example. The dimensions underneath it are reputation signals such as reviews and awards, professional credentials and specialties, practical factors including location, hours, insurance, and pricing, and patient experience factors such as anxiety-friendly or family-friendly care.
Each dimension generates its own sub-queries. “Dentist Edmonton good for anxiety” and “pediatric dentist Edmonton highly rated” both sit under the same head term, and a page covering neither answers only the surface question.
How Fan-Out Mapping Runs
I decompose the query before deciding what to build.
- Search the target query and record every People Also Ask prompt, expanding to reveal more.
- Record the related searches shown at the foot of the results page.
- Ask the target query in a conversational surface and note the sub-topics the answer covers.
- Sort every collected sub-question into before, during, and after the main question.
- Remove sub-questions with no genuine demand or no commercial relevance.
- Remove sub-questions the business has no standing to answer credibly.
- Assign each survivor to a page, either as a section or as a page of its own.
- Split into a new page only where the sub-question carries its own demand and intent.
- Check which sub-questions the site already answers, and which competitors answer instead.
- Order the build so the main page publishes before the pages that support it.
How This Engagement Is Scoped
I scope fan-out work by the number of target queries rather than by page count.
- What arrives
- A decomposition per query, sorted by journey stage, with every sub-question assigned.
- Typical duration
- Two to three weeks for a set of ten target queries.
- What the client supplies
- The target query list, and commercial priorities for assignment.
- Included
- Decomposition, demand filtering, assignment, coverage audit, and a build order.
- Not included
- Writing the pages, quoted separately once the map is agreed.
- What changes the estimate
- Several service lines, or queries spanning very different buyer stages.
- How it is quoted
- After a short scoping call, once the query list is agreed.
Tools and Primary Sources
The references below cover the surfaces and the evidence. Each opens in a new tab.
- Google AI features documentation describes how conversational surfaces assemble an answer from web sources.
- Search Console Performance report shows which sub-questions the site already earns impressions for.
- Google Trends separates sub-questions with real demand from those that merely sound plausible.
- Creating helpful, people-first content is the standard each answer still has to meet on its own.
Common Questions About Query Fan-Out Mapping
How is query fan-out different from keyword research?
Keyword research finds terms with search volume. Fan-out mapping finds the questions surrounding one term, including questions nobody types into a keyword tool because they get asked conversationally to an assistant.
How many sub-questions should one page answer?
Enough to complete the task the searcher arrived with, and no more. A page that fully answers four sub-questions outperforms one that half-answers ten.
Does fan-out mapping work for e-commerce product pages?
Yes. Product queries fan out into sizing, compatibility, materials, delivery, returns, and comparisons. Most product pages answer the first and ignore the rest.
Does every sub-question deserve its own page?
No. Sub-questions that share an intent belong on the same page as sections. Splitting them apart creates thin pages that compete with each other for the same ranking.
Discuss This Work
The first call is a free 30-minute consultation with no pitch attached. You receive an honest assessment of where the site stands and what can realistically be improved.