Search intent: the four types, and how to read it off the results page
What search intent is, the four types, and how to read it off the results page instead of guessing from the query, with the AI-exposure angle.
Auditing a 312-page recipe site, I found a chocolate dessert recipe ranking at position 1 for the query "how to make dessert", with 124 impressions and zero clicks. The query asked for a general guide and the page was one specific dessert, so none of those 124 people came in. That impression was not a missed opportunity: it was worth nothing, and counting it as one delayed the diagnosis.
What is search intent?
Search intent is what a person wants from a query, and you read it off the type of page that ranks in the top results, not off the words. On a new site it decides what is worth attempting; on an established one it flags when a query quietly changed hands.
Quick answer
- Intent is not deduced from the query, it is read off the results page. Google has already published its verdict: what it shows there is what it considers answers the query.
- The four types are informational, navigational, commercial and transactional. Commercial is the one most often dropped, and where the margin usually sits.
- Intent decides your exposure to generative AI. Informational is where the loss concentrates; transactional holds up far better.
- Aiming at one intent and showing up for another is not fixed with more content. It is fixed by accepting the query was never yours.
| Type | What the searcher wants | How it shows in the SERP | Best for |
|---|---|---|---|
| Informational | To understand something | Guides, videos, featured snippet, AI overview | Authority and citations, not sales |
| Navigational | To reach a specific site | The brand first, with sitelinks | Your own brand only |
| Commercial | To choose between options | Comparisons, reviews, prices, ads | A near purchase with margin |
| Transactional | To act now | Product pages, forms, maps | Revenue, and what AI erodes least |
What kind of problem this is
The dominant model treats intent as a property of the query text. You read the phrase, look for a revealing verb, assign it a box, and move on. It is fast, and it fails because it assumes the person classifying and the person searching share a vocabulary.
The correct model is different. Intent is not decided by whoever writes the query or whoever classifies it: it is decided by Google every time it returns results, and that decision is published. If the top 10 for your target query are ten product pages, that query is transactional even if it looks informational to you. Google has spent years watching what satisfies that search. Your intuition took five seconds.
And classifying stopped being a tidying exercise. With zero-click searches rising from 56 % to 69 % between May 2024 and May 2025 according to Similarweb, two queries with the same volume no longer carry the same value: it is an exposure calculation.
The four types, with the one that usually goes missing
Informational
The searcher wants to understand: "what is search intent", "how crawling works". It is the type that produces the most content and the one having the hardest time, because it is where featured snippets and AI summaries appear, answering without anyone reaching the first result. It still has a function. The function moved. It is no longer a traffic channel, it is the material a generative engine can cite, which is the argument in how to get cited by ChatGPT.
Navigational
The searcher already knows where they are going and uses Google as an address bar: "search console", "vercel login", a brand name. The only navigational intent you can work is your own name, and the work is not losing it to directories and aggregators.
Commercial
This one falls out of most classifications because it gets confused with informational. The searcher has already decided to buy and is choosing: "X vs Y", "reviews of", "how much does it cost". It looks like a question and it is a purchase stage. You recognize it because the top 10 fills with comparisons and reviews with ads on top, and advertisers pay where the money is.
Transactional
The searcher wants to act. Buy, hire, book, call. In the Search Console of marketingnativo.com, a site I run, "agencia de seo estados unidos que ayudan a negocios locales latinos" shows 154 impressions at position 28.5 with no page written for it. Nobody types that out of curiosity. Less volume, more competition, better economics. It is the type that best withstands AI summaries, for a mechanical reason: a summary can explain what a CRM is, but it cannot sign up for one on your behalf.
How to read intent without guessing
The procedure takes two to five minutes per query and needs no paid tool.
1. Search the query and see what type of page holds the top 10
Not which domains: which type of page. Ten long guides is informational. Ten product pages is transactional. Seven comparisons and three reviews is commercial. If the result surprises you, the one who was wrong was you. Country matters: the keyword research I ran across 11,220 queries in Colombia, Spain and Mexico returned different lists for the same seeds, and the wider method is in free keyword research.
2. Read the SERP features, which are declared signals
Google does not just choose results, it chooses formats, and each format says something. A product carousel with prices declares transactional intent. A knowledge panel with a single brand declares navigational. A local pack with a map declares nearby execution. A featured snippet with a definition declares informational. The majority rules, not the exception. When two incompatible formats coexist, the query is split.
3. Check against what is already reaching you
Search Console has half the work done. Export twelve months of queries, sort them into the four types, and calculate what share of your impressions falls in each box. That split is your site's risk profile, and almost nobody has looked at it. For a local business the same read is sharper still, because the queries are local by construction: local keyword research.
When NOT to do this
When the SERP is mid-volatility. During a core update the top 10 reorders from one day to the next: 2026 had four confirmed, and the March one changed the share of the top 3 to 79.5 % according to SE Ranking. Classifying on an unstable SERP is reading noise.
Also when the problem is not topics but clicks. A position 1 with zero clicks does not have an intent problem if the title does not contain the word people type, and that diagnosis is in why organic traffic drops.
Intent decides how much generative AI affects you
Here the taxonomy becomes a business decision. AI summaries appear on around 43 % of Google searches, 48 % by March 2026, and organic CTR with a summary present fell from 1.76 % to 0.61 % according to Seer Interactive, which also measured a drop from 2.72 % to 1.62 % on queries without a summary when one sits above. That loss is very unevenly distributed: informational takes almost all of it, and transactional holds up much better.
Three profiles with the same impressions and different fates:
- Pure informational site (blog, publisher, recipe site). Maximum exposure, and the worst at resisting it, because its ad revenue depends on the volume that evaporates.
- Mixed site (services with a blog). Informational loses traffic and commercial holds. The decision is not to stop writing guides, it is to stop counting them as an acquisition channel.
- Transactional site (store, agency, local service). Low exposure. The profitable work rarely sits here.
The counterpart changes the conclusion. The page cited inside the summary keeps visibility where the others lose it, and in AI Mode, where 93 % of queries end with no external click, the citation is all that is left. What those summaries are and how they count is in what AI Overviews are.
The mismatch you cannot write your way out of
Back to the dessert. The mismatch was structural, and it has three exits, only one of which is writing. Create a page of the correct type and let the recipe compete for its own specific query. Accept the query is not yours and stop counting those impressions as opportunity, which is what we did. Or rewrite the page for the other intent, which almost always ruins the one that already worked. Publishing more of the same is none of the three. Google is not showing you by mistake: it is showing you as a second-choice candidate for an intent you do not serve, and that is a stable position.
Mistakes that repeat
- Classifying by reading the query instead of looking at the SERP. You write a guide for something Google answers with product pages, and that page never enters the top 10.
- Folding commercial into informational. You lose the box with the best economics: someone comparing has already decided to buy.
- Competing for other brands' navigational queries. Budget burned with a predictable return: none.
- Treating every impression as opportunity. Wrong-intent impressions sink the average CTR and misdirect the diagnosis.
- Serving two intents on one URL. It answers neither well and loses to specialists in both.
- Classifying once and never again. Intent shifts with the market and with every core update.
Data and transparency
The method of reading intent from the type of page holding the top results is operating judgment from work across a portfolio recording more than 300 million impressions a year in Search Console. No share-by-intent figure appears here: the ones that circulate come from third-party keyword samples and vary enormously with the set chosen.
The dessert case (position 1, 124 impressions, zero clicks for "how to make dessert") comes from an audit of a 312-page WordPress recipe site: client work, not named. The query "agencia de seo estados unidos que ayudan a negocios locales latinos", 154 impressions at position 28.5, is from the Search Console of marketingnativo.com, my own site, which is why it is named. The 11,220 queries across Colombia, Spain and Mexico are own research from August 2026. AI summary presence comes from Similarweb (Generative AI Landscape 2026) for the 43 % and BrightEdge (commercial verticals, March 2026) for the 48 %; the CTR drop with a summary, from 1.76 % to 0.61 %, from Seer Interactive, published 4 November 2025, over 3,119 informational queries. The 93 % no-click figure in AI Mode comes from Semrush, over 69 million US desktop sessions between 1 May and 5 July 2025, and from Seer Interactive, over 25.1 million impressions. Zero-click searches rising from 56 % to 69 % are Similarweb (May 2024 to May 2025); the March core update volatility is SE Ranking. Verified as of September 2026.
Primary sources, opened 13 August 2026: the Seer Interactive study; the SE Ranking survey.
What this changes
The four types are the most repeated part of any SEO manual, taught as an exercise in sorting words into boxes. That was their function when every impression converted into a visit at the same rate.
Not anymore. The same scheme is now a map of exposure: it says which part of your traffic sits in the path of AI summaries and which part nobody touches. The classification did not change, what it is for did. And the exercise that returns the most costs no new plan. Do not classify the queries you want. Classify the ones already reaching you, and see how many of your impressions live in the box that is emptying.
Frequently asked questions
How do you identify the search intent of a query?
By searching it on Google and looking at which type of page holds the top 10, not by interpreting the query text. If guides dominate, the intent is informational; if product pages dominate, transactional; if comparisons dominate, commercial. The SERP features Google shows, such as a product carousel or a local pack, confirm the read.
What are the four types of search intent?
Informational, navigational, commercial and transactional. Informational wants to understand something, navigational to reach a specific site, commercial to compare options before buying, and transactional to act now. The commercial type is the one most often omitted and usually has the best return, because someone comparing has already decided to buy.
Can a query have more than one search intent?
Yes, and you recognize it because the SERP mixes incompatible formats: long guides sitting next to product pages. Google splits the space when no single intent dominates. The practical decision is to serve the majority intent of the top 10 with one URL, rather than covering both and ending up mediocre at each.
Which intent should I work if generative AI is eating traffic?
Transactional and commercial hold up much better: the CTR loss concentrates in informational. That does not mean abandoning informational, but changing its metric. An informational page in 2026 is measured by citations in generative engines and by authority, not by direct visits.
How do I know if my page ranks for the wrong intent?
By the ratio of impressions to clicks in Search Console. High impressions, a good position and a CTR near zero mean Google shows you and nobody recognizes you as the answer. If the query also calls for another format, such as a general guide against a specific case, the mismatch is one of intent.
Most sites do not have a ranking problem
They have a what-happens-next problem. You can rank first and still sell nothing. The diagnostic looks at both and tells you which one is costing you money.
See the diagnostic