Competitor keyword research is usually described as a way to steal traffic. That is the least useful thing it does. Before a rebuild, its real job is to show you which page shapes win which terms in your market, so the new site is built in the shapes that rank rather than the shapes a designer likes. It also tells you where nobody is good, which is where a small firm gets in.

I ran this on KOODOS in September 2026 — twelve competitor domains, 4,550 ranking rows, 3,852 unique keywords, for a total tool spend of $1.88.1 The numbers below are from that run. The method works at any scale; the cost is mostly time.

Step one: choose the right competitors

Not the firms you compete with for clients. The sites that rank for the terms you want. These are often different, and the difference is the point. Take the head term for each of your services — from the demand research — search it, and note every domain on page one that is a business rather than a directory or a tool. Do the same for the AI assistants: ask ChatGPT or Perplexity the question a buyer would ask and note who gets cited. Eight to fifteen domains is plenty.

In my case the Google set and the AI set barely overlapped. Four of the twelve domains I looked at came only from AI citations, and they turned out to hold some of the most useful pages.

Step two: pull what each one ranks for

Ahrefs, Semrush and DataForSEO will each give you every keyword a domain ranks for, with volume, position and the ranking URL. Set a floor — I used 50 UK searches a month — and export the lot for each domain. Then merge them into one sheet and remove:

  • Brand and navigation terms (their name, “login”, “careers”).
  • Tool and software terms, if you sell a service.
  • Jobs, locations you do not serve, and anything outside what you sell.
  • Near-duplicates: “seo audit service” and “seo audit services” are one keyword for planning purposes.

What remains is the market’s vocabulary, with a rough map of who owns which part of it.

Step three: look at the pages, not the keywords

This is the step most guides skip and the one that matters most for a rebuild. Group the keywords by the URL that ranks for them. You will find that a handful of pages on each competitor site do most of the work. In my set, one agency’s cost guide accounted for 136 ranking keywords on its own; another’s maintenance service page carried 49. One domain with 1,207 available keywords had 165 of them on its homepage.

From those pages, note the shape that wins each kind of term:

  • Service pages win commercial terms: “seo audit services”, “website redesign services”, “ai seo agency”. Five different agency service pages ranked for the first of those.
  • Detailed guides win question terms: cost, how much, checklist, how to. Thin FAQ pages do not.
  • Comparison pages win platform-choice terms: X vs Y.
  • Location pages win local terms. Local blog posts do not.

When the rebuild plan has a page in the wrong shape for its term, it will not rank however well it is written. This step is how you find out before building it.

Step four: score the gaps

Now find the terms worth building for. For each keyword, a simple score works:

score = volume × (1 − difficulty ÷ 100) × intent weight × fit

Intent weight: 1.0 for commercial terms, 0.7 for informational, 0 for navigational. Fit: 1.0 if the term is squarely what you sell, 0.5 if adjacent, 0 if not. Sort by score. The top of the list is your page plan, and it will surprise you. Mine put “website redesign services” at difficulty 0 with 1,300 searches a month above “seo agency” at 6,600 searches and difficulty 42, because the second is not winnable for a small domain and the first is.

Step five: turn it into pages

The output of all this is short. For each service, and for the guides section:

  • The target term and its shape.
  • The competitor URL to beat, with what it does well and what it lacks.
  • The sections the page needs: pricing, migration, comparison, FAQ, whatever the modifiers said.
  • The gaps nobody has covered, ranked, as the guides backlog.

That is a build brief. It is also the thing a template cannot give you, which is why the pillar guide puts competitor research second, straight after demand, and before anyone chooses a design.

If you would rather not

This is a day’s work for someone who does it regularly and a week for someone who does not. It is one of the five parts of GroundWork, at £250, and the sample Demand Map shows what the output looks like when it is written for a decision rather than as a spreadsheet.

  1. KOODOS competitor gap analysis, DataForSEO Labs ranked_keywords, United Kingdom, 17 September 2026. Twelve domains, keywords at 50 UK searches a month or more. ↩