Search intent: why the grid of four intents is no longer enough
Google documents that it rephrases a question and breaks it into subtopics, then runs several searches to answer it. A page therefore no longer answers a single intent attached to one keyword: it has to cover the cluster of sub-questions the engine will ask on your behalf.
- The classic grid of four intents, informational, navigational, commercial and transactional, still describes the buyer but no longer describes how the engine works.
- Pew measures that 8% of one-to-two-word searches trigger a generated summary, versus 53% from ten words up. Query length decides how much traffic is left.
- 60% of questions phrased with who, what, when or why trigger a summary. These are exactly the ones the market targets first.
- The budget consequence: one page that covers a cluster of sub-questions is worth more than five pages fighting over the same query.
- The only place you can observe these rephrasings is the Bing citation report, which publishes the anchor queries.
On this page
Query fan-out
Query fan-out is the technique by which an engine rephrases the question a person asks, breaks it into subtopics, then runs several distinct searches to gather what it needs to build its answer. Google documents it for its generative features. Its practical consequence is that a single question from the user triggers several queries against the index, so the idea of a single intent attached to one keyword no longer matches what the engine actually runs.
Why the classic grid no longer works
For twenty years, the method has been to sort each keyword into one of four boxes: the person wants to learn, to go somewhere, to compare, or to buy. You then produce one page per box, and you measure. This grid structured content budgets and still describes a buyer's state of mind correctly. It no longer describes what the engine runs.
Google in fact documents a query fan-out technique: it rephrases the question, breaks it into subtopics, then runs several searches to gather what it needs to answer. One person asks a question, the engine asks five. Assigning a single intent to one keyword therefore amounts to describing only one of the five.
The second shift comes from the user side, and it is measured. The Pew Research Center observed the real browsing of 900 American adults: 8% of one-to-two-word searches trigger a generated summary, versus 53% from ten words up. And 60% of questions phrased with who, what, when or why trigger one. Queries are getting longer, and the longer they get, the more the answer appears without a click. The fewer visits happen, the fewer customers the content brings in.
These two figures have an immediate budget consequence. Long questions phrased with question words are exactly the ones the market has advised targeting for three years. They are also the ones that produce the fewest clicks. They keep their value in presence and awareness, but a budget funded on a promise of traffic will be disappointed by them.
What to settle before funding a content production
How you select topics decides the return on a content budget more than the quality of the writing does. A list of keywords sorted into four boxes produces pages that compete with each other and sometimes target zero-click queries.
- Do our topics come from questions customers actually ask, or from a keyword tool?
- How many of our pages target the same question under different phrasings?
- Which queries do we want to win for traffic, and which for presence alone?
- On which pages do we accept being cited without being visited, and why?
- If none of these pages has produced a customer within nine months, what do we stop funding?
The answer that holds up distinguishes the queries that bring visits from those that bring only presence, and says so before producing. A hollow answer sells a list of keywords sorted into four boxes, promises traffic on long questions, or treats every variant as a page to write.
Deciding which topics deserve a page, which ones group together and which will only earn presence is settled from your real customer questions. A 90-minute consultation settles it, with a written summary your team can execute.
What this changes in production
The shift fits in one sentence: you no longer produce one page per keyword, you produce one page per cluster of sub-questions. This change of unit directly alters the cost of a content plan.
| Old method | What it produces | New method |
|---|---|---|
| One page per keyword variant | Pages that cannibalize each other | One page per real, distinct question |
| Classify the intent before writing | Copy that answers a third of the need | List the expected sub-questions |
| Target search volume | Traffic with no buying intent | Target the questions that precede a purchase |
| Optimize keyword density | Copy no one reads aloud | Answer first, then develop |
| Measure rankings | A report disconnected from sales | Measure the leads generated per page |
The first row is the one that cuts a plan's cost the most. Two pages answering the same question deprive each other of ranking, and that is the most frequent cause of failure for a batch of content bought by volume, as detailed in publishing 100 pages at once. The structure that lets you answer first and then develop is covered in structuring content to be quoted.
What stays in-house: the list of questions your customers ask before buying, in their own words, and those that come back after the sale. Your salespeople and your customer service supply them in an hour, and no tool replaces them. What can be delegated: grouping these questions into clusters, checking the existing content for duplicate intent, the page structure, the writing and the measurement. A company that supplies thirty questions it has heard gets a more profitable content plan than a company that buys a list of a thousand keywords.
Where to observe the real phrasings
The least known practical point on this subject is that there is a place where you can see the rephrasings the engine produces, and it is not at Google.
The Bing tools citation report publishes the anchor queries, that is, the searches the system runs internally to build its answer. It is the only source that shows the real vocabulary triggering a pickup, rather than the one you assumed. The account is free, which avoids committing a production budget to a hypothesis. Google's report, for its part, gives impressions without queries. The full comparison of the two is in optimizing for answer engines.
A keyword tool gives volumes, not intents. It does not know whether a query precedes a purchase, whether it comes from a competitor watching the market, or whether it is asked by someone who has already chosen. Building a content plan on these volumes alone amounts to funding visits without knowing what they are worth or what they cost per customer won.
The general framework of search is laid out in what search optimization really covers, the inventory of existing content in the audit that helps you decide, the writing rules in structuring content to be extracted, and the metrics that drive decisions in marketing metrics.
Building a presence that produces customers rather than an inventory of pages is at the heart of the Attract customers with SEO and AI goal.
Already running a marketing team? See how we plug in as reinforcement on search engine optimization.
Frequently asked questions about search intent
Are the four types of intent outdated?
They still describe a buyer's state of mind correctly, but they no longer describe what the engine runs. Google documents that it rephrases a question, breaks it into subtopics and runs several searches. Assigning a single intent to one keyword amounts to describing only one of those searches.
Should you target long questions?
For presence, yes. For traffic, with caution. Pew measures that 8% of one-to-two-word searches trigger a generated summary, versus 53% from ten words up, and 60% of questions using who, what, when or why. These queries therefore produce fewer clicks, while keeping their awareness value. The call is made by objective, not by volume.
Do you need one page per keyword?
No, it is the most frequent cause of failure for a content plan. Two pages answering the same question deprive each other of ranking. The right unit is the real, distinct question, with its sub-questions handled on the same page. Grouping costs less than rewriting.
Where can you see the phrasings the engine actually uses?
In the Bing tools citation report, which publishes the anchor queries, that is, the searches run internally to build an answer. It is the only source that shows the real vocabulary triggering a pickup. Google's equivalent report gives impressions, without queries.
- Google Search Central, Optimizing your website for generative AI features on Google Search, on query fan-out, accessed July 2026.
- Pew Research Center, Google users are less likely to click on links when an AI summary appears in the results, browsing study of 900 American adults, March 2025.
- Bing Webmaster Blog, AI Performance in Bing Webmaster Tools, on the publication of anchor queries, February 2026.
Geneviève puts the strategy for your engagement into action. She leads all our web development projects: Shopify, WordPress and the new ways of building a site with AI. She manages our team of developers and translates your business needs into technical language. She runs your organic search (SEO), your visibility in AI answers (GEO) and your site's conversion rate optimization (CRO). Her work is at the heart of three goals: Attract customers with SEO and AI, Improve your site's conversion, and Strengthen your visibility in AI answers. With Gabriel, she also builds the landing pages for your advertising campaigns. She writes mainly about SEO, AI visibility and web design.
About Falia →
SEO
Paid advertising
Conversion
AI visibility
Web design