Search intent is the reason behind a query: what the person wants to know, do or find when they type it into Google or ask an assistant. Get the intent right and a modest page can win. Get it wrong and a brilliant page will not rank, will not be quoted, and will not convert, because it answers a question nobody asked.
For the definition alone, see the search intent entry in our glossary. This post is the practical part: how to read intent, how to match a page to it, and what our own tracked questions show about which kind of page wins each intent in AI answers.
What kinds of intent are there?
Marketers usually sort intent into four types:
- Informational: the person wants to learn something. "What is generative engine optimization."
- Navigational: they want a specific site or page. "Attensira login."
- Commercial: they are comparing options before a decision. "Best AI search visibility tools."
- Transactional: they are ready to act. "AI search visibility starting price", "sign up".
Google's own Search Quality Rater Guidelines, the instructions its human raters follow (the September 11, 2025 edition), use a slightly different set: Know queries, some of which are Know Simple queries with a short, agreed answer; Do queries, where the user is trying to accomplish a goal; Website queries, looking for a specific site or page; and Visit-in-person queries, looking for a business or a category of businesses nearby. The guidelines are explicit that a query can carry more than one intent.
The labels matter less than the habit: before writing anything, decide which of these the person is in.
How do you identify the intent of a query?
Read what already wins
The fastest and most reliable method. Search the query and look at the page types on the first page, not just the domains. If it is all guides, the intent is informational. If it is product and pricing pages, transactional. If it is "best X" lists and comparisons, commercial. If it is a map, visit-in-person. Google has watched millions of people choose; the results page is its summary of what they wanted.
Read the words
Modifiers give intent away: "what is", "how to" and "why" lean informational; "best", "vs", "alternatives", "review" lean commercial; "price", "buy", "trial", "demo" lean transactional; a brand plus "login" or "pricing" is navigational. They are clues, not rules.
Read the AI answers too
Ask the question to the assistants your buyers use, several times. Look at the shape of the answer (a definition, a list of options, steps, a price range) and at the kinds of sources it cites. Do not assume the sources are Google's top results: assistants often cite pages that do not rank for the query at all, so read the sources as well as the shape. The answer's shape tells you what the engine thinks the person wants.
Read the prompt as a whole
Prompts to assistants are longer than search queries and often state their intent outright: "best AI visibility tool for a seed-stage B2B startup, under a small budget, that does the work rather than just reporting". That is commercial intent with three constraints. A generic "best tools" page does not match it; a page that addresses those constraints does.
How do you match a page to the intent?
Each intent has a page type that fits it. Use the type the winners use, then do it better.
Informational, short answer
- Page that matches
- A definition or reference entry
- What the first lines must do
- Give the answer in one or two sentences
Informational, in depth
- Page that matches
- A guide with question-shaped sections
- What the first lines must do
- Answer the core question in 40 to 60 words, then go deep
Commercial
- Page that matches
- A comparison, an alternatives page, a "best X for Y" with a method
- What the first lines must do
- State the verdict and who each option suits
Transactional
- Page that matches
- A pricing page, a product page, a sign-up
- What the first lines must do
- State the price, the terms and the next step
Navigational
- Page that matches
- The page itself, reachable and clearly titled
- What the first lines must do
- Be the obvious result for your own name
Do
- Page that matches
- A tool, a template, a step-by-step how-to
- What the first lines must do
- Let the person do the thing, with the steps first
Two rules hold across all of them. Answer the question in the first lines, in a sentence that stands alone, because that is the passage both a reader and an AI answer reach first. And give each intent its own page: a page that defines, compares and sells at once does all three badly.
Which kind of page wins each intent in AI answers?
We track 61 questions about our own category in our own product, on ChatGPT and Perplexity. Reading which sources decided each answer over the 30 days to 4 October 2026 shows a pattern by intent. The numbers come from the attensira.com workspace over the Attensira MCP on 4 October 2026, and every per-question count below (answers, citations, times we were named) is from the same read, 5 September to 4 October 2026.
The table at the top of this page sorts four of these questions by intent. Those intent labels are this post's own, using the four marketer types above; they are not values the product filed. Attensira's agent files a question as informational, comparison, commercial or transactional, and has no navigational value, so a question like "what is Attensira" would not be filed under that name.
Informational questions are decided by authorities. For "how to get cited by ChatGPT" on ChatGPT, OpenAI's own help centre held 1,205 of the 1,831 citations in 57 answers from 5 September to 4 October 2026. For "what is generative engine optimization", Google's developer documentation led. Informational questions about a platform are answered from the platform's own docs, and an independent page has to offer something the docs do not, such as measured data, to be quoted at all.
The same holds for a definition. For "what is GPTBot", OpenAI's own crawler documentation was the first deciding source on both ChatGPT and Perplexity, followed by a long tail of glossary pages that each restate it.
Commercial questions are decided by lists. For "best answer engine optimization tools", the most influential source on ChatGPT was a competitor's blog, and on Perplexity it was HubSpot's roundup, followed by Profound's own list, Meltwater, AIMultiple and G2's category page. We were named in none of the 115 answers from 5 September to 4 October 2026. A commercial answer leans on pages that list options, so the work is getting onto those lists, as well as having a comparison page of your own. The band below shows the shape in one reading: asked for Ahrefs alternatives for AI search on 3 October 2026, Perplexity answered with a five-item list and cited ten "Ahrefs alternatives" pages, Zapier's and Backlinko's among them. Across 118 answers to that question from 5 September to 4 October 2026, we were named in none.
Transactional questions are decided by pricing pages and price write-ups. For "how much does AI search visibility cost", the first deciding source on ChatGPT was a competitor's pricing page, followed by articles that benchmark what tools charge. We were cited in 1 of 108 answers from 5 September to 4 October 2026. A transactional answer wants a price stated plainly, and it takes it from whoever states it most clearly.
Navigational questions are where we win. For "what is Attensira", we were named in all 66 answers from 5 September to 4 October 2026 and cited in 63, and our own site was the first deciding source on both engines. Navigational intent is the easiest to win and the least valuable for finding new buyers: nobody asks it until they already know your name.
The lesson is practical: once you know a question's intent, you can predict what kind of page decides it, and therefore whether the work is on your own site or on someone else's.