Search intent: questions first

The four categories, and where they really come from

Before the categories, the thing that makes them usable: intent is a property of the search, not of the keyword. The same word changes intent depending on how it is used — “motorbike insurance” serves both the person working out how it works and the person after a quote, and those two people need different pages.

Almost every guide lists four categories. They are useful, but it is worth knowing they are not a Google classification.

Three of those categories — informational, navigational and transactional — come from a 2002 academic paper, A Taxonomy of Web Search by Andrei Broder, which analysed AltaVista's logs. The commercial one is a later addition by the SEO industry, invented to separate the people comparing from the people buying.

What Google actually uses

In the quality rater guidelines — the document Google uses to instruct the human raters who judge result quality — the categories are different. The section on intent distinguishes:

• Know and Know Simple, people after information, with the “simple” variant for short, definitive answers;
• Do, people who want to accomplish something;
• Website, people looking for a specific site;
• Visit-in-person, people looking for something physical near them;
• and queries with multiple intents at once.

Two things matter more than the labels. The first is that these are instructions for people assessing results, not a label the algorithm attaches to queries: there is nowhere to read “the official intent” of a search. The second is that Google explicitly allows for the multi-intent query, which is exactly what breaks editorial plans built on one keyword per page.

Why a keyword list is not enough

A keyword list sorted by volume tells you how much something is searched, not what is wanted. The typical outcome is one page per keyword — dozens of pages answering the same question in slightly different words.

The damage is not only the wasted work. When several pages on the same site answer the same question you get cannibalisation: the engines have to pick which one to show, the choice wobbles over time, internal links split across two addresses, and none of them is as strong as one page would have been.

Grouping by intent produces the opposite: fewer pages, each with a clear reason to exist. It is also the most reliable way to decide which page to update when a new phrase turns up.

What the grouping changes

• Different phrases with the same answer share a page.
• Similar phrases with different answers need separate pages.
• Volume decides the order of the work, not the number of pages.
• Two pages competing for the same question get merged, not optimised harder.

How to read intent

No data source declares a query's intent. It is inferred, and the reliable signals are few. This is a procedure that can be repeated the same way on every group of phrases.

1. Look at the kind of page holding the top positions. It is the most reliable signal, because it is how the engines have already interpreted that question. If they are all guides, the phrase is informational even when it looks commercial. If they are all product pages, a guide is not getting in there.
2. Read the related questions block. It is the block Google calls a related questions group, also known as “People also ask”. It tells you what the person has not understood yet, and it is the material you build a page's internal structure from.
3. Look at the other elements on the results page. A product carousel, a video result, a local results block: each one tells you what kind of answer the engine considers right.
4. Check the phrases where the site takes impressions without clicks. If a page shows in a good position in Search Console and almost nobody comes in, one possible explanation is that it answers a different question from the one asked. The others are a low average position, or an element on the results page absorbing the click.
5. Record the outcome and the date. Intent is to be re-read, not settled once and for all.

When the SERP is mixed

Sometimes the top positions mix page types: three guides, two product pages, one comparison. That is not an engine error, it is the multi-intent query the quality rater guidelines allow for.

In that case there are two routes. If one of the two questions is clearly in the majority among the results, build the page on that one and pass the other on with an internal link. If they are balanced, one page each is better: any page trying to answer both ends up answering both badly.

Intent changes over time

A question that is informational today can become commercial tomorrow, often because a market grew up in the meantime. The reverse happens too: as a product becomes ordinary, searches move from “how does it work” to “how much does it cost”.

This does not need continuous monitoring. It needs a rule: before rewriting a page that is underperforming, re-check the kind of results holding the top positions. If they have changed, the problem is not the text — it is that the page answers a question people have stopped asking.

An example, to see how it works

Take eight phrases collected on one topic:

• what is warehouse management software
• what is warehouse management software for
• warehouse management software how it works
• best warehouse management software
• warehouse management software comparison
• warehouse management software pricing
• warehouse management software demo
• warehouse management software for small businesses

A list sorted by volume produces eight pages. Read for intent it produces three, plus one to check.

The first three close the same question, “what it is and what it does”: one page, with the other two phrases as internal sections. The next two are both a request to compare and go together on a comparison. “Pricing” and “demo” look close to comparison, but they are a decision already made: they belong to the product page.

The last one is the case worth looking at closely. “For small businesses” looks like a variant of the first, but if the top positions show pages dedicated to that segment then it is a different question and deserves its own page. If they show the same generic guides, it is just a section of the main page.

The check is always the same: open the top positions for one phrase from each group. If the kind of page that ranks is the same, the group holds. If it changes, the group has to be split.

From the map to the structure

Once the questions are grouped, the structure almost writes itself: a group of neighbouring questions is a page, the adjacent groups are its internal links, and the groups no page covers are the editorial plan.

This is also the point where a map of questions starts to pay off in AI answers: a page that closes a question completely is a page that can be cited.

What the map produces

• A list of pages to create, with the question each one closes.
• A list of pages to merge because they answer the same question.
• The internal links between neighbouring question groups.
• The order of the work, set by volume and by distance from the buying decision.

The rule for each page's internal order is the same: the answer to the main question up top, in a paragraph that stands on its own, and the related questions as the sections that follow. Someone in a hurry finds what they came for immediately, and a system that has to cite the page finds a self-contained block to cite.

Search intent and AI answer engines

There is a time-saving piece of news here. In its guide to optimising for generative AI, Google says no dedicated techniques are needed: an llms.txt file is not used by Search, there is no need to chop content into smaller pieces so the AI understands it, writing for machines does not help, and structured data is not a requirement for appearing in generative answers, though it remains useful for rich results.

What Google points to as decisive is something else: creating content people find unique, interesting and useful will likely influence your site's presence in generative AI search, over the long run, more than any other suggestion in that guide. The technical recommendations are the usual ones: crawlable and indexable content, semantic HTML, good page experience, no duplicates. On content, Google asks that information be organised with clear paragraphs and headings.

Translated into this guide's terms: the work on intent is the work on citability. A page built on a precise question, which closes it and declares it in the title, is exactly what an answer system can retrieve and cite. A page built on a keyword, touching five subjects and closing none, is not, and no technical trick saves it.

It is also consistent with what Google asks about content: write for people rather than for search engines, and avoid producing content across many topics hoping some of it performs. That is the same outcome you get by starting from the questions instead of the keyword list.

Common questions