Definition
Search intent (also called user intent or query intent) is the goal a person is trying to reach when they type a query or ask an AI assistant a question. The same words can carry different goals, and both search engines and answer engines judge a page by how well it serves the goal, not by how many query words it contains.
The idea predates modern SEO. In 2002, Andrei Broder of AltaVista sorted web queries into informational, navigational and transactional. Marketers later split out commercial investigation, which gives the four types most SEO tools use today:
- Informational: learning something ('how does a heat pump work').
- Navigational: reaching a known site or page ('Notion login').
- Commercial: comparing options before a decision ('best heat pumps for cold climates').
- Transactional: completing an action ('buy Daikin Altherma 3').
Google uses its own labels. Its Search Quality Rater Guidelines, last updated September 11, 2025, ask raters to classify queries as Know (and Know Simple), Do, Website and Visit-in-Person, and to consider queries with several likely intents. Raters judge 'Needs Met' against that intent, so an excellent page can still fail if it answers the wrong question.
Conversational AI widens the picture. An OpenAI and NBER study of ChatGPT usage, published in September 2025, classified messages as Asking (49%), Doing (40%) and Expressing (11%) (How People Use ChatGPT). Doing, where the user wants the model to produce something such as an email or a plan, barely exists in classic web search. Prompts are also longer: Google said in May 2026 that the average AI Mode query is three times the length of a traditional search query. Long prompts often state constraints ('for a 10-person team, under $20 a seat, works with HubSpot') that make intent explicit, and they also carry latent intent, the unstated needs behind the question.
AI search engines act on intent in two places. First, they decide whether and how to search: a navigational prompt may get a single link, while a commercial prompt triggers query fan-out into several searches for reviews, pricing and alternatives. Second, they pick sources whose format fits the job: comparison pages and listicles for 'best X for Y' prompts, step-by-step guides for 'how to' prompts, official pages for navigational ones.
For GEO, intent is the organizing principle for both tracking and content. Group the prompts you monitor by intent, because visibility usually differs sharply between informational questions (where brands are cited as sources) and commercial ones (where brands are recommended by name). Then check that each target prompt has a page built for its intent. A product page will rarely be cited for an informational question, and a glossary-style explainer will rarely be recommended for a buying question. Search intent classification covers how queries are labeled at scale; our guide to finding the right queries to track shows how to build an intent-balanced prompt set. Related: prompt research and natural language queries.
Examples of Search Intent
- 'Mailchimp pricing' is navigational and commercial at once: Google leads with Mailchimp's own pricing page, while an AI assistant may answer with the plan prices and then compare them with alternatives, because the prompt suggests the user is weighing cost.
- A heat pump installer ranks well for 'heat pump cost' with a sales landing page but is never cited in AI Overviews for it. Adding a cost breakdown with regional price ranges and installation factors matches the informational intent and starts earning citations.
- A prompt like 'write a cold email to a CFO about our expense software' is a Doing request in OpenAI's taxonomy. No web page can rank for it, but the model may still name tools in the draft, so brand mentions matter even for non-search intents.
- A B2B team segments its 300 tracked prompts by intent and finds 40% visibility on informational prompts but 6% on commercial ones, which tells them their blog is cited while competitors win the recommendations.
