One Keyword Can Hide Four Different Customer Decisions

Google’s 2025 Search Quality Rater Guidelines separate searches by the need behind them: know, do, reach a website, or visit a place. These are instructions for human raters, not a secret ranking formula. For a business, however, they expose an expensive mistake: similar words can require different pages, evidence, and next steps.

Consider “commercial pipe leak.” One searcher wants a diagnosis. Another is comparing repair methods. A third needs a plumber before a bakery opens. A fourth may be trying to reach a contractor already hired. This is an illustrative example, not one company’s data. The words look similar; the decisions do not.

Search intent is the job a person expects search to help complete. Treat it as a content brief: what uncertainty must the page remove, what proof belongs here, and what should become easier next?

Start With the Decision After the Click

Start with your own reports. Google Search Console shows queries that produced impressions and clicks before a visit. Google Ads’ search terms report shows actual searches that triggered ads—and distinguishes a search term from an advertiser’s keyword. These records are traces of demand, not finished strategy.

A phrase can still contain mixed intent. Location, device, season, prior knowledge, and the result page change the likely task. Ask which decision appears most common and which valuable secondary route must remain easy to reach.

Intent also differs from funnel stage. A technical question may come from a procurement manager hours from purchase; a “download” query may carry little value. Classify the action before guessing the customer’s worth.

Use Four Buckets Without Worshipping Them

Know intent asks for a definition, specification, diagnosis, or method. Put the answer early, then provide evidence. Hiding it beneath a long company introduction turns curiosity into an exit.

Do intent asks for progress: calculate, configure, schedule, buy, or repair. Supply a usable action path and enough proof to take it safely. “Learn more” is weak when the customer is ready to act.

Website intent aims at a known destination such as a login, support page, return portal, or company. The job is navigation and reassurance. Clear naming and a direct route beat another persuasive essay.

Visit-in-person intent adds place. Hours, service area, availability, accessibility, parking, and local proof can determine whether attention becomes a visit. Mixed queries exist, so use these buckets to sharpen judgment—not replace it.

Let the Result Page Show Its Hand

Search the important queries in the market and location that matter. Record dominant result types. Explanations suggest a knowledge task; comparison pages suggest evaluation; maps suggest a visit; product or booking pages suggest action. The result page is imperfect evidence, but it shows what customers can already obtain.

AI search raises the bar for knowledge pages. Google says its AI features can surface relevant links without special optimization beyond established search practices. Nielsen Norman Group calls a need satisfied on the results page “good abandonment.” A basic definition may no longer earn a click; original demonstrations, local availability, calculators, first-hand evidence, or a better explanation can.

Do not manufacture a page for every wording. Google emphasizes helpful, reliable, people-first content, and its generative-AI guidance warns that producing many pages without added value may violate scaled-content rules. Cluster queries around distinct decisions; build only pages with a useful job.

Run a 30-Query Intent Audit

Export 30 to 50 important nonbrand queries from Search Console and paid search. Privacy thresholds can omit some low-volume terms from the Ads report, so treat the export as a sample. Put one query on each card. Have two people classify it independently as answer, compare, act, or arrive; record confidence and the current landing page.

Disagreements reveal ambiguous language that may need a hybrid page or clearer route. For each valuable cluster, write one page job. In the plumbing example: “Help a facilities manager decide whether a leak requires shutdown and request 24-hour response.” That beats “Rank for commercial plumbing.”

Inspect the page opening. Does it identify the customer, recognize the situation, provide proof, and offer the next action? If two intents need incompatible openings—an educational guide and emergency dispatch—separate them. One page cannot give every route equal priority.

Measure each cluster by its job for four to six weeks. Knowledge pages may earn qualified onward clicks or assisted leads. Action pages should produce calls, bookings, or purchases; visit pages may produce directions and appointments. Do not compare raw conversion rates as if every search began at the same decision.

Match the Page to the Moment

Rewrite the promise around the decision. Answer pages should name the exact question; comparison pages, the criteria; action pages, availability, price boundaries, or timing; visit pages, local proof. Nielsen Norman Group’s information-foraging research explains why: people judge whether a page is likely to contain enough relevant information for the effort required.

Carry the classification into the ad, result snippet, landing page, follow-up, and measurement plan. OrionPilot’s role here is organizational: preserve the customer decision across channels, then learn whether the promised route produced the intended behavior.

Traffic is only an arrival count. Intent fit determines whether the arrival feels like progress. A query is an unfinished decision. Give it the right page, and both the beginner seeking an answer and the buyer ready to act can move without fighting one generic message.