Search intent: what it is, the four types and how to identify it

Search intent is the goal a person has in mind when they type a query: not the word they write, but what they want to get once they click. If a page doesn't serve that goal it won't rank, however well it is written.

Every guide on the subject says that much. What almost none of them say is that since 2025 reading intent has to answer a second question, before you even pick the page format: if the AI Overview already gives the answer, is the page still worth writing? In December 2025 SEOZoom found an AI Overview on 54.15% of Italian SERPs, across a sample of almost 30 million keywords, in Italy, where Visilay is based. It is no longer a niche scenario to keep an eye on: it is what the results page looks like today.

Query, keyword and intent are not the same thing

Three terms that overlap in conversation and are worth keeping apart, because mixing them up causes half the mistakes made during keyword research.

The query is what the user actually types: "best warehouse management software for small business". The keyword is the normalised form that keyword research tools use to group those queries: "warehouse management software". The intent is what lies underneath: that person is comparing options, not looking for a definition.

Two very different queries can share the same intent, and the same keyword can hide two. That is why a keyword list sorted only by volume leads you astray, and why long-tail keywords are often worth more than head terms: they are longer precisely because the user has spelled out what they want.

The four types of search intent

The classification you'll find in every tool and every guide splits searches into four categories. It works as a first approximation, as long as you remember it is an industry convention, not an official Google category.

Intent	What the user wants	Example query (UK searches/month)	Type of page that ranks
Informational	To understand something	what is ctr (590)	Guide, glossary, tutorial
Navigational	To reach a specific site	xero login (246,000)	Home page or brand page
Commercial	To compare before deciding	warehouse management software (590)	Comparison, review, ranking
Transactional	To take the action	led screen hire (480)	Product, service or contact page

Informational intent

The user wants to know. Definitions, tutorials, recipes and technical comparisons with no immediate intention to buy all belong here. It is the largest category. It is also the one hit hardest by generative search.

Navigational intent

The user already knows where they want to go and uses Google as an address bar. These are brand queries, and for the brand that owns them they are the hardest to lose and the least useful to chase with new content.

Commercial intent

The zone where the sale is decided. "Best", "alternatives to", "X vs Y", "how much does it cost", "reviews". Whoever types these queries has a recognised problem and is choosing a supplier. These are also the queries where SEO and Google Ads overlap the most.

Transactional intent

The action is imminent: buy, request a quote, book. Volume is low; value per visit is the highest of all. A blog post almost never ranks on these queries, and when it tries, it takes strength away from the service page that should be there. It is also the most common reason a site gets traffic but doesn't convert.

Where this classification comes from

It is worth knowing that the four types were not born on a marketing blog. They come from a 2002 paper by Andrei Broder, then at IBM Research, published in ACM SIGIR Forum. Broder proposed three categories, not four: navigational, informational, transactional. He measured them on AltaVista logs in 2001: 48% informational, 30% transactional, 20% navigational.

The fourth category, commercial, was added later by SEO tools, which split Broder's transactional zone in two to separate people comparing from people buying. It is a useful distinction for funnel thinking, but it is ours, not Google's.

Google uses a different taxonomy, and it's worth knowing

In the Search Quality Rater Guidelines, the document Google uses to instruct its human raters (version of 11 September 2025), section 12.7 classifies intent differently:

• Know and Know Simple: the user wants information. Know Simple queries are the ones answered definitively in a sentence or two.
• Do: the user wants to take an action: buy, download, book.
• Website: the user is looking for a specific site.
• Visit-in-person: the user wants to go somewhere physically.
• Queries with multiple user intents: a category of its own, which Google puts in writing and which four-box models tend to hide.

Two differences matter in day-to-day work. The first is Know Simple: Google has an explicit category for questions with a short, definitive answer, and that is exactly the material an AI Overview extracts without needing to send anyone to the site. The second is Visit-in-person, which doesn't exist in the four-type taxonomy and which, for a business with physical premises, outweighs all the others put together.

What has changed now that AI gives the answers

Here the data is clear-cut, and you won't find it in most of the guides ranking for this keyword.

In April 2025 Ryan Law at Ahrefs analysed 300,000 keywords, comparing March 2024 with March 2025, and found that position one CTR drops by 34.5% when an AI Overview appears. The same study has the number that explains everything else: 99.2% of queries that trigger an AI Overview have informational intent.

Pew Research Center reached the same place starting from how people actually behave. Across 68,879 searches made by 900 US adults in March 2025, 8% led to a click on an organic result when there was an AI summary, against 15% when there wasn't. Only 1% clicked one of the sources cited inside the summary, and 26% of sessions with an AI summary ended there, against 16% without.

The agency Amsive, looking at 700,000 keywords across ten sites, added the detail that flips the reading: non-branded queries lose 19.98% of CTR, branded ones gain 18.68%, and only 4.79% of branded queries trigger an AI Overview. In other words: navigational intent is the only one coming out of this phase stronger.

Then the front moved. In July 2026 Luke Harsel at Semrush measured, across more than 600,000 keywords between November 2025 and April 2026, 71% growth in AI Overviews on commercial-intent queries, with finance up 231%. Transactional queries fell by 5% over the same period. The zone where the sale is decided is where AI is expanding now.

The December 2025 SEOZoom study on Italy, by Gennaro Mancini, adds the part that concerns publishers: of around 168 million domains monitored, only 175,000 are cited in generated answers. And 40% of cited sources are not in the top ten organic positions. Wikipedia alone appears 864,517 times. We have collected more figures on the same trend in our AI SEO statistics and in the guide to AI Overviews.

The second question: how extractable is the answer?

Those numbers produce a working rule that doesn't appear in guides on search intent. Labelling a query as informational is no longer enough to decide what to write, because within informational there is a gulf between a forty-word definition and a procedure that takes experience. The AI Overview absorbs the first, not the second.

On projects, the matrix we use to decide crosses two axes: classic intent, and how far a model can extract the answer without the user needing to come to you.

Intent	Example query	Extractability of the answer	Editorial decision
Informational, closed answer	what is ctr	High: 40-60 words are enough	Write it to be cited, not for the click. The return is the brand mention, not the session.
Informational, long answer	how to build a B2B SEO strategy	Medium: needs a process, not a fact	Long-form guide with a procedure, your own data and verifiable cases.
Commercial	best warehouse management software	Low: needs a reasoned judgement	Comparison with stated criteria and evidence. This is the zone under attack: cover it now.
Transactional	led screen hire	Very low	Service or product page, never an article.
Navigational	visilay blog	None	No new content. Protect your title and brand.
Visit-in-person	seo agency manchester	Low: the answer depends on the place	Up-to-date Google Business Profile plus a local page.

The most counter-intuitive row is the first. A short definition still makes sense, but the yardstick changes: you stop looking at sessions and start looking at whether your domain appears among the sources. That is the reasoning behind generative engine optimisation, and the reason featured snippets remain a sensible target even in a zero-click world.

How to identify intent by reading the SERP

The SERP is Google's answer to the question you're asking yourself. Five steps, in order.

1. Search in incognito, on Google for the market you're targeting, in its language. For the UK that means google.co.uk with a UK location. A US SERP says little about the British market, even in the same language. It is the mistake we see most often in content plans drawn up at a desk.
2. Count the formats of the top ten results. If eight out of ten are guides, your service page won't get in. If six out of ten are e-commerce category pages, an article is pointless.
3. Look at the SERP features. People Also Ask and featured snippets signal informational. Shopping and ads signal commercial or transactional. The local pack signals visit-in-person.
4. Read the titles. If "best", "which to choose" and "prices" keep coming up, the user is evaluating. If it's "what is" and "how does it work", they want to understand.
5. If there's an AI Overview, read which sources it cites. It tells you what Google considers a valid answer to that question, and what kind of page it expects it from. This is the new step, and for now hardly anyone does it.

If two variants return the same first page, they are the same query and one page covers both: that is the criterion for keyword clustering. It also works the other way round: two keywords that look like synonyms but have different SERPs need two pages, and trying to cover them together weakens both. On turning all this into pages, we have a separate guide on how to create SEO content and one on on-page SEO.

Two close keywords, two different intents: a real case

Macropix is a Milan-based manufacturer of custom LED screens we have worked with since 2020. In its sector, two queries live side by side that look closely related and aren't.

On Google Italy, "ledwall" gets 4,400 searches a month and has mixed intent: students, the curious, event agencies and industrial buyers. "Ledwall prezzo" (LED wall price) gets 720 searches a month and has clear commercial intent, because whoever types it is preparing a budget. They are two pages, not one, and the second is the one most manufacturers leave uncovered so as not to expose their price lists. In the UK the same pair looks different: "led wall" gets 1,000 searches a month and "led wall price" only 30, while "led screen hire" gets 480 (DataForSEO, Google UK, September 2026). Same product, different market, different page to build.

The result most worth telling is not first place for "ledwall" but the keyword "monitor pubblicitario" (advertising display): from position 88 in 2020 to position 2 in 2025. Position 88 is page nine of Google, meaning zero traffic. It wasn't a technical shortcut: it was a page written to answer what that query was really asking. The full method, with the table of eight keywords and their positions before and after, is in the case study on SEO for manufacturing and in the other projects we have published.

The mistakes we find most often in audits

• Treating a keyword as if it had only one intent. Google puts the category of multiple-intent queries in writing. On those, the SERP mixes formats and the winning page is the one that covers the most frequent case and leaves a bridge to the others.
• Aiming a service page at informational queries. It generates impressions and no enquiries, and on top of that competes with your own blog. The division of roles is clear: articles for informational and commercial, pages such as SEO services for transactional.
• Filling the text with keyword variants. Answering the intent and repeating the keyword are two different jobs, and the second easily slides into keyword stuffing.
• Never reopening the file. Intent shifts. "AI assistant" was an informational query in 2022; today it is commercial (6,600 UK searches a month, DataForSEO, September 2026). Pages should be rechecked at least once a year, together with the other ranking factors.

The query you're judged on is no longer the one the user types

There is one thing the industry still rarely says, and it is the least discussed consequence of all this. Google Search Central's official documentation states that AI Overviews and AI Mode use a technique called query fan-out: starting from one question, they run many others on subtopics and different sources, and build the answer from those. Robby Stein, VP of Product for Search, explained in July 2025 that in Deep Search mode the queries generated in the background can number in the dozens or hundreds.

It means the intent you need to satisfy is no longer just that of the query the person wrote, but that of the sub-questions the model asked itself. No tool today shows you all of them. The only practical way to catch them is to cover a topic in full rather than one keyword at a time, and that is the lens through which to read Google's AI Mode. If you want to see how to apply it to your site, it's the work we do in our AI SEO services.

Search intent FAQs