Search intent analysis: reading the results page, classifying the query, and catching a mismatch

Search intent analysis is the process of deciding what somebody wants when they type a query, and the evidence is already on the results page. Google has spent years testing which page types satisfy each query, so the formats currently ranking are the best available statement of intent: ten product listings means the searcher wants to buy, ten guides means they want to learn, a map pack means they want somebody nearby. Read the page before you read any tool.

The analysis matters because intent decides the format, and format decides whether the page can rank at all. A beautifully written guide aimed at a query where everybody else shows a product grid will not break through, however good it is. Match the format first, then compete on quality inside that format.

The four classes, and the page each one needs

Class	What the searcher wants	The page that satisfies it	Tell tale signals
Informational	To understand something	Guide, explainer, how to, definition	Featured snippet, People Also Ask, video carousel
Commercial	To choose between options	Comparison, review, best of list, buyer's guide	Listicles from publishers, review sites, comparison tables
Transactional	To buy or sign up now	Product page, pricing page, category page	Shopping results, ads above the fold, product listings
Navigational	To reach a specific place	The brand's own page	One dominant result plus sitelinks

Local intent cuts across all four. A map pack above the organic results means the searcher wants somebody within driving distance, and the page that satisfies it is a location page plus a well kept Business Profile. The longer explanation of how these classes relate to keyword types sits in what a keyword is in SEO.

The method: eight minutes per query

1. Search it in a clean window. Incognito, with the location set to your target market. Personalised results will tell you about your own browsing history rather than about the query.
2. Write down the format of the top ten. Count them: four guides, three product pages, two listicles, one forum thread. The majority format is the intent.
3. Read the titles as a group. The repeated words across ten titles are the words the searcher expects to see. If eight of ten say "for small business", that qualifier is part of the query even when it is not typed.
4. Note the SERP features. Shopping carousel means transactional. People Also Ask means the question has sub questions you should answer. A video carousel means somebody wants to be shown rather than told.
5. Open the top result and find the answer. Where it sits on the page tells you how impatient the searcher is. An answer in the first sentence means a quick answer query; an answer at section four means a considered one.
6. Check the dates. Where every result was updated in the last six months, freshness is part of what the searcher wants, and a page from 2021 will struggle whatever it says.
7. Check the depth of the sources. Original research, named authors and first hand testing in the top results mean the bar is expertise rather than word count.
8. Write one sentence. "Someone searching this is a ___ who wants to ___ and will judge the page on ___." If you cannot write that sentence, the analysis is not finished.

Mixed intent, and how to handle it

Plenty of queries have no single answer. "CRM software" returns product pages, comparison listicles and definitions all at once, because different people type the same two words for different reasons. Google is hedging, and the practical response is to pick one intent, serve it completely, and link clearly to the others.

The choice should follow your business rather than the majority format. A software vendor takes the transactional slot with a product page and links to a comparison. A publisher takes the comparison slot and links out to the vendors. Trying to satisfy all three on one page produces a page that satisfies none, and it is the most common reason a long, careful article underperforms a thinner competitor.

Diagnosing a mismatch on a page you already have

A page with impressions and almost no clicks is usually an intent problem rather than a quality problem. Three checks in order:

• Pull the real queries. In Search Console, filter the performance report to that URL and read the query list. These are the questions the page is actually being shown for, which is frequently not the question it was written for.
• Compare position against click through. A page averaging position four with a very low click through rate is being shown to people who look at the title and decide it is not what they wanted.
• Compare your format against the top three. If they are all comparison tables and you have a 2,000 word essay, that is the fix: restructure, do not rewrite from scratch.

Intent also explains conversion gaps that look like a landing page problem. A page converting at a fraction of the site average often ranks for an informational query while the page is built for a transactional one, and measuring organic conversion by query intent is the way to see it. The cheap fix is usually to add the information the searcher came for and keep the offer further down, rather than moving the offer up.

Recording the analysis so it is reusable

Keep the output in the same sheet as your keyword work, with four columns: the query, the class, the required format, and the page that owns it. That turns intent analysis from a one off exercise into a filter you apply to every new idea, and it exposes the case where two of your own pages are chasing the same intent. The keyword analysis template has the full column set; intent is the column that most teams add last and then wish they had added first.

One note on scale. Classifying intent by hand works up to a few hundred queries. Beyond that, start with a rough automatic pass, using the words in the query as a first guess, and then hand check every query you intend to build a page for. The automatic pass sorts the list. The hand check is what stops you building the wrong format for the twenty queries that matter.

Intent calls that are hard to make, and how to make them

Can intent be read from the keyword alone?

Modifier words give you a strong first guess: "how", "what" and "why" lean informational, "best" and "vs" lean commercial, "buy", "pricing" and "near me" lean transactional or local. The results page confirms or overturns that guess, and it is the results page that decides what ranks.

What do I do when the top ten are split evenly between formats?

Pick the format that matches what you sell, cover that intent completely, and link to the other formats from inside the page. A clear answer to one intent beats a compromise across three.

How often does intent for a query change?

Slowly for evergreen topics and quickly for anything tied to a product cycle, a season or a news event. Re-check the results page for your twenty most important queries once a quarter, and immediately after any large ranking movement.

Does intent analysis still matter if I write for people rather than search engines?

It matters more. Intent analysis is a structured way of asking what the reader actually came for, which is the same question a good writer asks. The results page simply gives you evidence rather than an opinion.

How do I analyse intent for a brand new product category?

Analyse the problem rather than the product. Search the words a customer would use to describe the pain, read the forums and the questions under competitors' content, and build for the query they have today rather than the one you wish they typed.

Should one page ever target two intents?

One page, one primary intent, with a clear path to the others. The exception is a long buyer's guide that genuinely answers the informational question and then presents the choice, which works when the two intents belong to the same person at the same moment.

Intent after the search box: what people type into an assistant

Everything above reads intent from a results page. A share of the same intent now arrives somewhere else entirely. Someone who would have searched "best crm for small business" types a paragraph into ChatGPT, Gemini, Perplexity or Claude instead, explaining their team size and budget, and gets three names back with reasons attached. The intent is identical and the evidence has moved: there is no results page to read, no title pattern to count, and no click through rate to diagnose.

Running your top queries through an assistant as a full sentence is the equivalent exercise, and the answers tell you two useful things: which format the assistant drew from, and which companies it decided were worth naming. Checking it is now a normal part of a research cycle, and AI Knows Us is a tool that does the asking for you, keeping the answers over time so a change is visible. The analysis habit is the same one described above. Only the surface being read has changed.