Site architecture
Keyword mapping, so every search has one page and every page has one search.
A keyword map is the document that says which page on your site answers which search. It sounds administrative. It is the single decision that stops a site competing with itself, and it decides where every hour of writing goes.
What a keyword map is
A keyword map is a table. Each row is a page, existing or planned. Next to it sit the main search that page is meant to win, the closely related searches it should also answer, the page it links up to, and whether it exists yet. That is all. What makes it valuable is the rule behind it: no search appears on two rows.
Most sites are built without one. Pages get written because they seemed a good idea at the time, and over a few years three or four of them end up chasing the same search. Google then has to guess which one you mean, and often guesses the page least able to take a booking. A keyword map prevents that before it starts, and on an existing site it shows exactly where it has already happened.
How I build one
Start from what you sell
Your services, in order of what they earn, and the places you genuinely serve. These become the top rows, the pages the rest of the map exists to support.
Collect the real searches
From Search Console, the searches your site already appears for. From keyword research, the ones it does not yet. Questions customers ask on the phone count too; they are usually searches as well.
Group searches by what the person wants
Searches that want the same answer belong to one page, even if the words differ: end of tenancy cleaning and move out cleaning are one page. Searches that want different things never share: what an end of tenancy clean includes is a different page from booking one.
Give each group one address
An existing page if one fits, or a new page if not. The address is written down, along with the parent it links up to.
Check the map against the site
Every existing page is placed on the map. Pages with no row are either given one or considered for merging. Rows with two existing pages are the cannibalisation to fix first.
Order the new pages
The pages that support the money pages most, in the places you do the most work, go first. The map becomes the writing plan.
Intent decides the grouping
The hardest part of keyword mapping is deciding which searches belong together. The test is not whether the words overlap but whether the person wants the same thing. A search for a family lawyer in Henderson and a search for how much a divorce costs in Nevada overlap in subject but not in need: one is ready to hire, the other is still weighing options. They need different pages, linked to each other.
The quickest check is to look at what Google already shows. If the results for two searches are mostly the same pages, Google treats them as the same need, and they can share a page. If the results are different, they should be separate.
A small map, worked for a cleaning firm
End of tenancy cleaning, the service page
Main search end of tenancy cleaning, with move out cleaning and end of lease cleaning on the same page because they want the same thing. Parent is the services hub. This page carries the price and the booking form.
What an end of tenancy clean includes
A question page for people still deciding, answering the checklist landlords and letting agents use. Parent is the end of tenancy service page, linked with those exact words.
How much end of tenancy cleaning costs
A pricing page answering the cost question honestly, with what changes the price. Parent is the service page.
End of tenancy cleaning in Stockport
A town page about the work in Stockport specifically: the areas covered, the kind of homes and lets there, how quickly a team arrives. Parent is the service page; it links across to neighbouring towns.
What the map rules out
A blog post called the ultimate guide to end of tenancy cleaning, because it would compete with both the service page and the checklist page. The map shows there is no row for it to occupy.
Where a map changed the outcome
For a cloud and DevOps training academy, twenty six pieces were written in its first seven months, each mapped to a real search and linked into the part of the site it belonged to. With forty keywords on page one, the map also showed something unusual: new posts were adding fresh keywords at position fifty and dragging the average down. Publishing paused for a month while off-site work caught up. Across those seven months the tracked set moved from single digits to more than forty page one keywords, and from an average position of 21 to 10.
For a Las Vegas injury firm, mapping each accident type to its own page beneath one practice hub, and each surrounding community to its own hub, took the firm from 9 top-three rankings to 44 in the Ahrefs rank tracker by July 2026.
What a good map prevents
Pages competing with each other
One of the commonest reasons a site with plenty of content still ranks poorly, in my experience. With one search per row, it cannot happen by accident.
Writing that lands nowhere
Every new page knows its search and its parent before it is written, so it strengthens something rather than floating on its own.
The wrong page ranking
Service searches are assigned to service pages, so articles do not take the searches that should land on a page with a price and a booking form.
Repeating yourself next year
A new writer or agency can see what already exists and what each page is for, instead of writing the same article again.
Mapping an existing site
On a site that already has hundreds of pages, the map is built backwards as well as forwards. Each existing page is matched to the search it is actually collecting in Search Console, which is often not the one it was written for. The overlaps that appear become decisions: which page leads, which merges into it, which narrows to a different question. Nothing is deleted just to tidy the map; pages that earn links or traffic are merged and redirected so what they earned passes on.
Mapping services across many towns
Businesses that serve many towns face a choice for every service and place: a separate page for each combination, or one town page covering all services there. The map decides it by search, not by habit. Where people search a specific service in a specific town in meaningful numbers, that combination earns its own row. Where they search the town and the trade in general, one town page covering all services is stronger than several thin ones.
Either way the rule holds: one search, one page, and every page linked up to the service or hub it belongs to.
What you receive
The map itself, as a sheet you own: page, address, main search, related searches, parent, status and date due. A list of the overlaps found on your current site with a decision for each. And the order in which new pages should be written. It is the document every later piece of work is checked against, and it is written so anyone you hire next can use it.
It is kept in your portal alongside the work log, so you can see which rows have become live pages and which are still to come.
Questions about keyword mapping
Is keyword mapping only for big sites
No. A ten-page site benefits too, because even there two pages can quietly compete. The map is just shorter.
How many keywords should one page target
One main search, plus the closely related searches that want the same answer. The number of words is not the point; the shared intent is.
Do I need to rewrite pages to fit the map
Sometimes. A page that currently collects the wrong search may need its title, headings and content sharpened toward its assigned search.
How often should the map change
It grows as new searches appear in Search Console and as the business adds services or places. The rule, one search per page, never changes.
Can I use a tool to make the map
Tools help collect searches and spot overlaps. The grouping by intent and the decisions about existing pages need a person who understands the business.
Want to see where your pages overlap?
Send me the site. I will map what you have and show you which searches are split across more than one page.