Start with the search, not the keyword

A query is a job, not a string of words

The idea is older than modern search. In 2002 Andrei Broder proposed that web queries fall into three broad classes: navigational (reach a particular site), informational (find information) and transactional (do something on the web, such as buy or download).^[1] Search engines have refined the thinking since. Google’s Search Quality Rater Guidelines, the manual it gives its human evaluators, describe intent with five labels: Know, Know Simple, Do, Website and Visit-in-person.^[2] Raters use the guidelines to judge result quality. They are not a ranking formula, but they show what Google’s own evaluators are asked to look for.

Two details in the guidelines are worth borrowing. First, a Know Simple query is one where most people would agree on the right answer and it fits in one or two sentences or a short list, such as a height or opening hours. Most queries are not like that: broad, ambiguous, controversial or opinion-based queries need a fuller page.^[2] Second, the guidelines acknowledge that some queries carry several intents at once, and that the user’s location and locale can change how a query should be read.^[2]

Read the results page before you write

Rater guidance is a useful vocabulary, but the best evidence about intent for your query is the page of results that exists today. Search the query yourself and record what you see: the mix of page types (guides, service or product pages, directories, videos, maps), the typical depth, the questions those pages answer and the questions they leave open. If the first page is almost entirely service pages and your plan is a 3,000-word explainer, the evidence says your explainer meets a different need from the one most searchers have.

Treat the results as a hypothesis and not a template. They vary by location, language and device, so check from the market you serve and on a phone as well as a desktop. And the aim is to understand the job, not to imitate whoever ranks first. If every result is a thin list, the opening may be a page that does the job properly.

Give each page one job

A content map turns intent into a plan: one primary page per job, and other pages that link to it. This example is illustrative, for a fictional physiotherapy clinic.

When two pages chase the same job, they compete with each other and neither reads as the answer. When a query mixes intents, lead with the dominant one and give the other a visible route, such as a short paragraph and a link, instead of forcing one page to be everything.

Decide whether a new page is justified

Before you create a page, ask four questions. Does it serve a different job from every existing page? Would the best answer take a different format, such as a table, a calculator or a checklist? Can you add something existing pages cannot, such as first-hand evidence or a tool? Is it written for a different audience? If every answer is no, extend the page you have.

That finding explains why one strong page usually collects its wording variants on its own, so a page per variant adds duplication and no reach. It is one vendor’s analysis of its own dataset, so read it as a pattern and not a promise. Google also draws a line here. Its spam policies describe doorway abuse as pages created to rank for specific, similar queries that lead people to intermediate pages less useful than the final destination, including pages aimed at individual cities that funnel users to one page.^[4] A page for each town you serve is fine when each carries real local information. Without it, one page covering the area is better.

Google’s people-first guidance asks the same question from another side: was the content created primarily to help people, or primarily to attract search visits?^[5] A page that exists only because a keyword exists fails that test.

Give the reader a next step

The page should end where the job ends. A Know page points to the deeper guide, tool or comparison. A comparison points to the decision. A Do page makes the action easy, with a short form and the price or timing information someone needs before committing. Give each page one primary next step and design the others as secondary. The UX & UI Playbook covers the mechanics in Calls to action and Lead forms.

The next step is also your best measure of whether you read the intent correctly. Search Console tells you which queries reach the page. Your analytics tell you whether those visitors take the step. A page with many impressions and no next-step clicks is probably answering a different question from the one people asked.

A practical review exercise

Choose the ten queries that bring, or should bring, enquiries. Search each one on a phone from your market and fill in a row. Do this before you commission any new page.

Apply it to your website

• Every target query is labelled with the job behind it and checked against today’s results from the market you serve.
• Each page has one primary job and one primary next step.
• No two pages compete for the same job. Overlapping pages are merged or clearly differentiated.
• A new page exists only because its intent, format, audience or evidence is different from what you already have.
• Location pages carry real local information, or they are replaced by one page that covers the area.
• You revisit the intent map when rankings, enquiries or the results page change.