How does an SEO company do keyword research?
It works through a chain of steps: collecting seed terms that describe the business, expanding them into a large list of candidate searches with research tools, reading what searchers want from each one, judging which ones the site can realistically compete for, grouping searches that mean the same thing, and assigning each group to one page. The quality of the result depends less on the tools than on reading their numbers correctly, because several of the figures that tools report do not measure what they appear to measure.
Seed terms
The work starts with the plain words that describe what the business does and sells. They come from conversations with you, the language on your current site, the questions customers ask, and the terms your competitors use. For an existing site, the searches that already bring visitors, shown in Search Console, are added to the seed list. The seed list is a starting point, not a plan.
Expanding the list, and reading the numbers
From each seed, research tools produce related phrases, longer variations, and questions. Two numbers come with each candidate, and both need care.
In Google’s Keyword Planner, the search volume figure is an estimate, and Google’s guide to Keyword Planner metrics explains how it is calculated: “By default, the number of searches for the term (regardless of language) is averaged over a 12-month period.” An average across the year hides seasonal peaks, and a company planning content for a seasonal business should look at the monthly pattern, not the average alone.
Keyword Planner’s competition column is easy to misread, because it is about advertising, not about ranking in unpaid results. Google defines it as “The number of advertisers that showed on each keyword relative to all keywords across Google.” A term marked low competition there may still be very hard to rank for organically. The difficulty scores in third-party SEO tools are those tools’ own estimates of organic difficulty, each calculated differently.
What the searcher wants
Volume and difficulty do not say what a person typing the phrase is looking for. The reliable way to find out is to look at the results currently shown for it. If they are mostly product or service pages, people want to buy or hire; if they are guides, people want to learn; if they are lists of providers or a map, people want to compare or find someone nearby. That reading decides what kind of page could compete for the search.
What the site can realistically win
A high-volume search is not worth much if the pages that rank for it are far stronger than anything your site could offer soon. The company compares the pages that rank with what your site has, and sorts the candidates into terms to target now, terms that need more work first, and terms to leave. Searches where fewer strong pages compete are a sensible place for a newer site to start, and those are sometimes, though not always, the longer and more specific ones.
Grouping and mapping
Some phrases are simply different ways of asking the same question. Giving each its own page splits your effort and sets your pages against each other, so related phrases are grouped, for example by checking whether their results show largely the same pages. Each group is then assigned to one page on the site, either an existing page to improve or a new one to create.
Seeing how the research becomes a plan
Ask which tools the company uses and how it treats their numbers: whether it looks at seasonal patterns rather than only annual averages, how it judges organic difficulty, and how it reads intent. Ask to see how it groups phrases and which page each group is assigned to. The answers tell you whether the research will become a plan you can act on or a long spreadsheet of terms.