Keyword research before you build

A light keyword pass before a build is worth doing, because it informs two decisions that are expensive to change afterwards: which pages exist, and what each one is called. A full competitive programme is not a build task and belongs after launch.

The distinction matters because “keyword research” is sold as both, and the second is an ongoing engagement with a monthly fee while the first is an afternoon.

What a keyword’s search volume number actually is

A search volume figure is a modelled estimate of the searches a keyword gets in a defined place over a defined period. It is not a count. Different tools return different figures for the same phrase, and figures are usually rounded to a scale that hides real differences.

Three other columns usually appear beside it, and each answers a narrower question than its name suggests.

Column	What it measures	What it does not
Volume	Times a phrase is typed	Whether the people typing it buy anything
Difficulty	A model’s estimate of competition	Whether you can rank, given no history
Cost per click (CPC)	What advertisers pay	Anything about organic results
Competition	Advertiser density in the ad auction	Difficulty in organic results

Difficulty scores deserve particular caution. In the dataset gathered for this site, the phrases web design brisbane and website development brisbane return identical Google Ads figures — the same volume, the same cost per click, the same competition index — while their modelled organic difficulty scores differ by a wide margin. Two phrases that the advertising system treats as one market cannot plausibly have very different organic difficulty. The scores are model output, and treating them as measurements is a mistake this project made and corrected.

What keyword research before a build is actually for

Naming. Whether to call a page by the trade term or the customer term, when the two differ. This is a real use and it is covered from the structural side on what to call your pages.

Structure. Whether two services are searched separately, which is evidence that they are two pages rather than one. That decision is on splitting the services page.

Scope reality. Finding out early that the phrase the business hoped to own measures a difficulty of 81 is better than finding out after paying for content aimed at it.

Nothing else. It does not tell you what the business should sell, and a business that reshapes its offering around a volume figure has let a model make a commercial decision.

Where search volume, keyword difficulty and intent mislead

High volume, wrong intent. A phrase can be searched heavily by people who will never buy. In the Australian data gathered for this project, page speed test measures a thousand a month at the highest difficulty on the scale. Almost everybody typing it wants a free tool, not a supplier.

High volume, wrong page shape. The results for a phrase reveal what a search engine has decided the query wants. Several commercial-sounding phrases in this project’s research returned pages of galleries, listicles and templates rather than suppliers, which means a service page aimed at them competes against a page shape it is not. That is why this site records a SERP check separately from a volume figure, and why a page can be right about demand and wrong about the page.

Low volume, high value. A phrase measuring ten a month with clear purchase intent can be worth more than one measuring a thousand. Most Australian suburb-plus-service phrases sit here.

Zero measured volume, real question. This is the one that changes how a site is planned. what is a website brief returns no measurable Google volume in the Australian data; website brief returns thirty a month. Every business commissioning a website faces the first question. Nobody types it. A page can be worth writing because a buyer meets the problem, not because a tool counted it.

Two different keyword volume measurements that are not interchangeable

Google volume and AI-assistant prompt volume are separate figures from separate sources, and a phrase can be strong in one and absent from the other. In this project’s data several phrases carry real assistant volume and no Google volume at all, which is the signature of a question people ask an assistant rather than a search box.

They must never be summed, and an assistant figure must never be presented as a search figure. There is a further trap: one term in this project’s assistant dataset returned a figure an order of magnitude above everything else and is flagged as possibly not Australian. No plan here depends on it and it is not quoted as an Australian number anywhere on this site. The wider treatment is on AI search and a small business website.

How to research keywords yourself, the light version

1. Write down the ten phrases you believe describe what you sell.
2. Search each one and read the first page of results. Not the numbers — the pages. Are they suppliers, galleries, directories, or government sites? That tells you what the phrase actually wants.
3. Note which phrases return your competitors and which return something else entirely.
4. Check whether your own words appear anywhere in what customers write to you. Your enquiry inbox is better evidence than any tool because it is your actual audience.
5. Only then look at volume, and only to break ties between two candidate namings.

Step 2 is the one professionals do and amateurs skip, and it costs nothing.

What this site does with its own keyword and volume numbers

Every figure published here traces to a dated file in research/, gathered in Australia, and pages that have no measured demand say so rather than implying otherwise. Several pages in this section state plainly that their subject returns no measurable volume. The method is on how this site measures things, and the pricing dataset’s rules are on how the pricing figures are gathered.

What to do next: keyword research before you write the sitemap, not to guess whether you’ll rank

Do the five steps above before you write your sitemap, not after, and bring the results to the brief. The project brief generator has a place for them. If a supplier offers keyword research as a line item in a build quote, ask which of the two things it is — the afternoon or the programme — and what the deliverable is.