All resources SEO

Keyword Research from Scratch

Keyword research is two questions: what do people type, and which of those could you plausibly win.

8 min read Beginner Updated September 2026

Plenty of keyword research produces an impressive spreadsheet and no traffic, because it answers the first question and skips the second. This process keeps both in view, and it works with free tools.

Build a seed list from real language

Start with the words your customers already use, not the words your industry uses. Sales calls, support tickets, reviews of competing products and question threads are all better sources than a tool, because they contain phrasing you would never have guessed.

  • Search autocomplete: type your topic and read what the box offers.
  • The "People also ask" and "Related searches" blocks on the results page.
  • Your own site search log, which tells you what visitors could not find.
  • Competitor page titles, which are their best guess at demand.

Sort by intent before anything else

Two phrases with the same volume can be worth wildly different amounts, because the person typing them wants different things. Sort every candidate into one of four buckets and the content plan writes itself.

IntentThe searcher wantsYou publish
InformationalTo understand somethingA guide or explainer
CommercialTo compare optionsA comparison or review
TransactionalTo buy or sign upA product or course page
NavigationalA specific siteNothing; they already know where they are going
Let the results page tell you

Search the phrase and look at what already ranks. If the first page is all comparison articles, that is the intent, and a product page will not rank there however good it is.

Judge whether you can win it

Volume without a realistic chance is a wasted month. Look at the first page of results and ask three things: are these sites in your weight class, do the pages actually answer the question well, and is there an angle nobody has covered.

  • A first page of major brands and dictionaries means pick something narrower.
  • A first page of thin, dated pages is an opening, whatever the difficulty score says.
  • Longer, more specific phrases have less volume and far better odds. Ten of them beat one you never reach.

Group into pages, not keywords

Many phrases are the same question asked differently, and they should share one page. Group your list by the answer a searcher needs, and give each group a single target page. This is what stops you writing four thin articles that compete with each other.

One page: "what is on-page seo"
          "on page seo meaning"
          "on page seo explained"

Separate page: "on page seo checklist"   (different intent: a list, not a definition)

Turn it into a plan you will finish

End with a short table: target page, primary phrase, intent, the angle nobody else has taken, and a publish date. Ten rows you will actually write beat two hundred you will not.

Key takeaways

  • Collect language from customers and search results before opening a tool.
  • Classify by intent; the results page is the reliable evidence.
  • Read the current first page to judge whether you can realistically compete.
  • Group phrases into pages, so you never publish two articles that fight each other.

Ready to do this for real clients?

AI-Powered Digital Marketing & Freelancing covers SEO, paid ads and client acquisition, with live mentorship throughout.

Explore the course