Search intent

Two people type "robots.txt". One wants to know what it is. The other wants to edit theirs before a launch. Same words, different jobs. Search intent is the job, and every search engine and AI assistant tries to work it out before it decides what to show.

What are the types of search intent?

The marketing shorthand has four:

• Informational

  The person wants to

  Learn or understand

  Example

  what is share of voice in AI search

• Navigational

  The person wants to

  Reach a specific site or page

  Example

  Attensira pricing

• Commercial

  The person wants to

  Compare options before buying

  Example

  best AI search visibility tools

• Transactional

  The person wants to

  Act now: buy, sign up, download

  Example

  start Attensira free trial

Google's own rater guidelines, the instructions it gives the people who evaluate its results, use different names: Know (with Know Simple for one-fact answers), Do, Website and Visit-in-Person. They also say what matters most in practice: many queries have more than one likely intent, and a good result serves the dominant one without failing the others.

How does intent change an AI answer?

More than it changes a results page, because an AI answer is one response, not ten options. The engine decides what kind of answer the intent calls for, and that decides which pages it reads and which brands it names.

Here is our own tracking. Attensira asks ChatGPT and Perplexity 61 questions about its category every day and files them into topics, most of them by what the buyer is trying to do. In the 30 days to 4 October 2026, read over the Attensira MCP, no answer in any topic that does not ask about us by name mentioned Attensira. Across seven of those topics, Attensira was named in none of the answers. Three patterns, each about intent:

• Commercial questions are answered with products, built from roundups, review sites and competitors' pages. That is where we are absent, and it is the work that matters most to us.
• Pricing questions read our pages but name others. About one answer in five, 43 of 217, links attensira.com as evidence while naming other brands. The intent is "what does this cost", and our page is a good source for the number, not yet a reason to choose us.
• How-to questions are answered with steps. Products appear less often, and the engines lean on documentation and guides; attensira.com was linked in 3 of 663 GEO education answers. The intent is to do a task, and a product is at most a footnote to it.

We publish our zeros because a tool that only shows the flattering rows cannot show you where the work is.

How do you find the intent behind a query?

Read what already wins it:

1. On a results page, look at the types of result: guides, product pages, comparison articles, a map. The engine has already guessed the intent.
2. In AI answers, read several answers and their citations. Do they name products? Which kinds of page do they cite? Do they answer in place or send the reader on?
3. Group your own questions by intent, not by keyword, and read results per group. A blended number hides that you win how-to questions and lose buying ones.

What do people get wrong about search intent?

• Writing one page for every intent. A page that explains, compares and sells at once is the best answer to none of them.
• Chasing volume over intent. A large informational query rarely produces buyers; a small commercial one might.
• Assuming the engine shares your classification. Check what it shows, then decide.
• Measuring visibility across all intents at once. Your share of voice on how-to questions says nothing about buying questions.

How does Attensira use intent?

When you onboard, the crew learns your business and picks the questions your buyers ask AI, grouped by what those buyers are trying to do. Each group is reported on its own, with the answers, the pages they cite and the sample size, so you see which intents you win and which you lose. The work follows: a page that answers a how-to question directly, a placement on the roundup that decides a buying question. In the default Draft mode, each piece waits for your approval. See prompt tracking for how questions are chosen and grouped.