How to decide what your site should be found for.

A list of terms with volumes beside them describes what people type. It does not say which of those people are worth reaching, which questions the business can genuinely answer better than a generic article, or which page is supposed to be the answer. Those are three separate decisions and a keyword tool makes none of them.

The consequence is visible on a lot of sites: forty pages, broad coverage, and nothing that is unmistakably the answer to anything. Each page is individually reasonable. Collectively they compete with each other for the same handful of intents while leaving the intents that would actually convert uncovered.

The alternative is to decide first what the site is for, in terms of questions rather than terms, and then to let the phrasing follow. The decision is a business one and it is not delegable to a tool.

The diagnostic: name the five questions the business most wants to be found answering. Then name the URL that answers each. If two questions resolve to the same URL, or one question to two, the architecture is the problem, not the content volume.

Terms are ambiguous; intents are not. The same phrase can be typed by somebody defining a concept and somebody ready to hire, and one page cannot serve both without serving neither well. Separating them is what makes it possible to say which page owns which question.

Sort every candidate question into one of these:

• UNDERSTAND

  Somebody is working out what a thing is or how it works. They are not buying and may never buy, but they are how a topic becomes associated with a business.

• DECIDE

  Somebody has a problem and is choosing between approaches. This is where comparison, cost and trade-off content belongs, and where most businesses are thinnest.

• IMPLEMENT

  Somebody has decided and wants to know how it is actually done. Useful, often high-effort, and the strongest evidence of competence a site can publish.

• HIRE

  Somebody wants the work done. This is the commercial page’s job and it is the one intent an article should not try to take.

Every intent worth owning gets exactly one page, and every page answers exactly one intent. That is the whole architecture. It is simple to state and uncomfortable to apply, because the corollary is that some existing pages have to be merged, retired or rescoped.

The discomfort is the point. Two pages aimed at one question do not double the chance of being found; they divide the evidence a search or retrieval system has about which one is authoritative, and they force the business to maintain two explanations that will drift apart. Where consolidation is the right answer, Google’s guidance on duplicate URLs describes how to signal which version is canonical.

The test for a collision is not phrasing overlap. It is whether a reader who arrived at either page would feel they had reached the right one. Two pages can share most of their vocabulary and answer genuinely different questions; two pages can use entirely different words and answer the same one.

A topic can be relevant, high-volume and still a bad use of the only publishing capacity the business has. These are the four disqualifiers worth applying before anything is written, because each one produces a page that exists and does nothing.

• NOTHING TO ADD

  The business has no first-hand view, no data and no distinct method. A competent summary of what everyone else says is the definition of a page not worth finding.

• THE WRONG AUDIENCE ENTIRELY

  The question is asked by people who will never be customers. Occasionally worth doing for reach; usually a distraction dressed as coverage.

• ALREADY OWNED INTERNALLY

  An existing page answers it. The correct action is to strengthen that page, not to publish beside it.

• UNMAINTAINABLE

  The answer changes quarterly and nobody owns keeping it current. A stale authoritative-sounding page is worse than no page.

The output of this work is not a publishing schedule. It is a map of intents to URLs, with the gaps visible, which is the artefact every later decision refers back to — what to write, what to merge, what to link, and what to measure.

1. 01

   List the questions, not the terms

   Decides

   What is somebody actually trying to find out?

   Output

   A question list in the business’s own language, before any tool is opened.

2. 02

   Sort each question by intent

   Decides

   Understand, decide, implement or hire?

   Output

   Questions grouped by stage rather than by topic.

3. 03

   Name the URL that owns each

   Decides

   Which page is meant to be the answer?

   Output

   An explicit owner per question — or a recorded gap.

4. 04

   Resolve the collisions

   Decides

   Where does more than one page claim the same question?

   Output

   A merge, a rescope or a retirement decision, with the canonical stated.

5. 05

   Apply the disqualifiers

   Decides

   Is there anything first-hand to say, and will it stay true?

   Output

   A shorter list that the business can genuinely maintain.

6. 06

   Order by commercial distance

   Decides

   Which gap is closest to the work the business wants?

   Output

   A sequence that starts near the money rather than at the top of the funnel.

Once the map exists, the tempting move is to fill every gap at once. The better move is to cover one intent group completely — understand, decide and implement for a single subject — before starting another.

A complete group gives a reader somewhere to go after the first page, gives the site internal links that describe genuine relationships, and gives search and retrieval systems repeated, consistent evidence about one subject rather than a single page each about twelve. Breadth is easy to produce and hard to be credited for.

It also makes the programme finishable. A map with three complete groups is a defensible position. A map with thirty half-covered ones is a backlog.

Intent maps go stale quietly. New pages get published without being added, so the map stops describing the site; then it stops being consulted; then the collisions it existed to prevent reappear.

The fix is procedural rather than clever: the map is updated as part of publishing, not after it, and a page that has no entry does not go live. That single rule is what separates a content architecture from a folder of articles.

It also gives the business a cheap review. Reading the map once a quarter and asking which entries are still the questions worth owning is a faster audit than reading the site.

An intent map settles coverage and ownership. It says nothing about whether the pages that fill it are worth finding, which is a separate discipline with its own failure modes — and the one that decides whether coverage turns into anything.

Nor does it settle how the pages relate to each other once they exist. A map is a plan for URLs; the links between them are what actually express the structure to a reader and to a machine.

1. A keyword list describes what people type; it does not decide which questions the business intends to be the answer to.
2. Sort questions by intent — understand, decide, implement, hire — because one page cannot serve two stages without serving neither.
3. One URL owns one intent. Two pages aimed at one question divide the evidence about which is authoritative and drift apart in maintenance.
4. A collision is when the same reader would be satisfied by either page; a cluster is when finishing one gives a reason to open the other.
5. Decline topics with nothing first-hand to add, the wrong audience, an existing internal owner, or an answer nobody will keep current.
6. Cover one intent group completely before starting another: breadth is easy to produce and hard to be credited for.
7. The map is updated as part of publishing, not after it — a page with no entry does not go live.