SEO · 8 min read
Search intent is observed on the results page, not guessed from the words
A definition and a method. What search intent means, where the standard categories came from, why the wording of a query is unreliable evidence of what the searcher wants, and how to establish the intent behind any query in about ninety seconds.
Written by Zubair Afzal, FounderUpdated
Definition
What search intent means
Search intent is the goal a person is trying to achieve when they type a query. Not the subject of the query — the goal. Two people can type identical words while wanting completely different things, and the discipline consists of working out which of those wants dominates before deciding what kind of page to publish.
The standard categories come from a specific place. In 2002 Andrei Broder, then at AltaVista, published a taxonomy of web search dividing queries into informational, navigational and transactional. Practitioners later added commercial investigation for the stage between researching a category and buying inside it. Google’s published guidelines for its human quality raters describe query intents in comparable terms, including wanting to know something, to do something, to reach a particular website, and to visit somewhere in person.
The important consequence of that history is that intent was always a description of a population, never a label on a string. A query has a distribution of intents across everyone who types it. The engine observes that distribution at a scale nobody outside it can, then returns a results page reflecting what it found. Which is why the most reliable evidence available to you is not the wording of the query, and not an intent column in a tool. It is the page the engine returns.
The categories
Four intents, and what each looks like from outside
Categories are only useful if they change what you publish, so this table is organised around that: what the person wants, what the engine tends to show them, and what a business can realistically offer.
| Dimension | What the person wants | How the results page tends to look | What a business can offer |
|---|---|---|---|
| Informational | To understand something, or to get one fact. Often with no intention of buying anything at any point. | Explainers, definitions, question boxes, generated answers, video, and sites with no commercial relationship to the subject. | The clearest available answer, stated early. Commercial value here is indirect and arrives later, if at all. |
| Navigational | To reach a specific site, page or account they already have in mind. | One dominant result plus its sub-links, then that brand’s own properties, then directories mentioning it. | Very little, unless the destination is you. Competing for somebody else’s brand query is expensive and usually futile. |
| Commercial investigation | To narrow a shortlist. Comparisons, alternatives, reviews, price ranges, and the question of who can be trusted. | Listicles, comparison articles, review sites, forum threads, and a mixture of publishers and vendors. | An honest comparison that includes your own weaknesses, or a genuinely useful decision aid. Much B2B revenue is decided here. |
| Transactional | To act now: buy, book, call, download, sign up, or find somebody nearby who can do the job today. | Product and service pages, a map for local queries, shopping and ad units, and very few explainers. | A page that lets the action happen immediately, with price, availability, proximity or credentials visible without a scroll. |
The method in one picture
Two ways to read a query, only one of which is evidence
The intent you infer
- the verb and modifiers in the query
- what your business happens to sell
- how your team already describes the market
- the intent column in a keyword tool
The intent the results show
- the page formats that actually rank
- the kinds of organisation ranking
- the features occupying the page
- how all of that changed over a year
Where the words mislead
Queries whose wording points at the wrong intent
These are the cases where inferring from the phrasing produces a confident wrong answer. Each group is worth checking against real results before anything gets written.
Nothing in the phrasing of these tells you what they are. Only the results page does, and in each group it usually says something the wording did not.
Questions that are really buying queries
Phrased as questions and answered on the results page by service providers. The person is deciding whether to hire somebody, not writing an essay.
- how much does a commercial roof replacement cost
- do i need planning permission for a loft conversion
- how long does a divorce take
- what happens if my boiler is condemned
Product-shaped queries that are really research
They look transactional and behave like commercial investigation. The results are comparisons and category explainers, because almost nobody buys these in one session.
- crm software
- commercial insurance
- accounting software for small business
- electric van
Queries whose intent has moved
The page did not get worse. The population searching changed, the engine followed them, and nothing in your analytics announced it.
- a category term that used to return product pages
- a how-to now answered on the results page itself
- a comparison term a marketplace has taken over
- a local term that now returns a map before anything else
These are examples of how customers in this market search, drawn from keyword research and from the questions that come up on sales calls. They are illustrative, not a volume claim — the actual demand in your area is something we size before recommending anything.
Boundaries
What search intent is not
The term gets used loosely enough that the boundaries are worth stating. These six claims are all common and none of them holds.
- It is not a fixed attribute of a keyword. The same phrase can be dominated by different intents in different countries, on different devices and in different years.
- It is not a setting inside a search engine that you align yourself with. It describes what people want, which ranking systems attempt to satisfy and frequently only approximate.
- It is not the label in a keyword tool. Those labels are inferred from patterns in the words, which is precisely the method this page argues is unreliable.
- It is not decided by what you would prefer to sell. A query being valuable to you has no bearing on what the people typing it want.
- It is not a promise. Matching intent removes a common reason for failure; it does not supply a reason for success, and everything else about the page still has to be good.
- It is not permanent once diagnosed. Re-check it on your important queries at least annually, because the drift is silent and the pages suffering from it look fine.
A page can be accurate, well written and technically sound, and still fail for one reason: it answers a question nobody typing that query was asking.
This is the most common avoidable failure in content work, and it is avoidable in about ninety seconds. Search the query. Look at what is there. If your planned page resembles nothing on that first screen, you are either about to be usefully contrarian or about to waste a month, and it is worth knowing which before the work is commissioned.
The rarer and more interesting case is a results page that is genuinely poor — everything ranking is thin, dated or evasive because the market has not produced a good answer yet. That is a real opening, and it is distinguishable from a format mismatch by one question: are the existing results the wrong shape, or merely badly executed? Wrong shape is a warning. Badly executed is an invitation.
Questions
Common questions about search intent
What are the four types of search intent?
Informational, navigational and transactional come from Andrei Broder’s 2002 taxonomy of web search. Commercial investigation was added later by practitioners to describe the stage between researching a category and buying within it: comparisons, reviews, alternatives and "best" queries.
Treat them as a vocabulary for describing what you see rather than as boxes every query fits into. Plenty of real queries sit between two of them, and that ambiguity is information rather than a failure of the scheme.
How do I find the intent behind a specific query?
Search it, in a clean browser session, from the location you care about, and read what comes back. The formats that rank, the type of organisation behind them, and the features occupying the page are the search engine publishing its own conclusion about what people want.
If the first screen is comparison articles, the query is not transactional however much you would like it to be. If it is product listings and a shopping row, an explainer will not compete no matter how good it is.
Can one page serve two intents?
It can serve one fully and acknowledge the other. What does not work is splitting the page evenly, because you end up second-best at both jobs and first at neither.
The practical pattern is to lead with the dominant intent, answer it completely, then hand the minority intent off with a clear link to the page that serves it properly.
Does intent change over time?
Yes, and it is one of the least-monitored causes of gradual decline. The population searching a phrase changes as a market matures, and the results page follows them.
The signal to watch is impressions holding while clicks and average position slip on a page nobody has touched. Re-run the query and look at what now ranks. If the format has shifted, the page is answering a question people have stopped asking in that form.
Is search intent a ranking factor?
It is not a factor you can set. It is a description of the thing ranking systems are trying to satisfy, which is why matching it matters and why matching it is never on its own a promise of anything.
The useful framing is negative rather than positive. Getting intent wrong is a reliable way to make a page fail; getting it right only makes the page eligible to compete on everything else.
Check the queries your best pages are chasing
If a page you rate highly has never performed, intent mismatch is the first thing worth ruling out. It takes one search per query and the answer is usually visible immediately. Bring three pages and we will run through them.
Last updated