Keyword research: a repeatable method in five steps

Keyword research works as a repeatable five step method: collect seeds from your own data and your sales conversations, expand them with a keyword tool, group the result into questions rather than phrases, qualify each question against four tests, then commit one page per question. The step people skip is the fourth, and it is the one that decides whether the work earns anything. Run the whole method in a day for a small site.

What has changed recently is where the research gets used, not how it is done. The list you build is still the plan for the year, and a question list travels further than a phrase list because it works on a results page, in a brief and in an assistant.

Step 1: seeds from reality

Start with sources that already contain real demand, in this order.

• Search Console. Your own performance report shows the exact queries that already bring you impressions, including wordings you never targeted. Filter for queries ranking between eight and twenty, because those are pages you already half own.
• Your sales team. Ask for the ten questions they answer every week. These are the highest value seeds on the list and they rarely appear in any tool.
• Support tickets. Questions asked after buying show you what the pre sale pages left unexplained.
• Competitor navigation. Read the menus and page titles of three competitors. Their page list is a summary of what their buyers ask.
• Where your buyers talk. The trade forum, the subreddit, the LinkedIn comments. Questions asked in public are phrased honestly.

Step 2: expand with a tool

Now take each seed into a keyword tool to find the wordings and the relative demand. Google Keyword Planner is free with a Google Ads account and reports volume in banded ranges for accounts with little or no spend, which is enough for ordering your list. Ahrefs Keywords Explorer, Semrush Keyword Magic Tool and Moz Keyword Explorer give larger suggestion sets and their own difficulty scores. Google autocomplete, the related searches at the foot of the results page and the People Also Ask boxes are free and show you live phrasing.

Pull everything into one sheet with the query, the volume figure, the source and a blank column for intent. Volume numbers from any tool are modelled estimates, so use them to rank one query against another rather than as a forecast of visits. Your own Search Console impressions are the accurate figure for queries you already appear on.

Step 3: group into questions, not phrases

Several wordings of the same question belong to one page. "Seo audit tool", "seo audit tools" and "best seo audit tool" are one page, not three. Collapse the sheet until each row is a question a person would actually ask out loud, and the rest of the planning becomes straightforward.

Then label the intent, because it decides the page type.

Intent	Looks like	Page that answers it
Informational	what is, how to, why does	A question page that answers in the first paragraph
Comparative	best, vs, alternatives, for small business	A comparison that names competitors accurately
Commercial	pricing, cost, agency, services, near me	A service or pricing page with real numbers or a real range
Transactional	buy, download, free trial, generator	The tool or the product page itself
Navigational	your brand plus a word	A page you already have, made easier to find

Step 4: qualify with four tests

This is the step that separates a plan from a spreadsheet. Put each question through all four.

Does it lead to money? Trace the path from this question to a sale in one sentence. If the sentence needs three steps, move the question down the list. High volume questions with no path to revenue are the most common way a content budget is spent without return.

Can you serve it honestly? You need a real answer, from your own work, data or product. Questions you would have to bluff are the ones that produce vague pages.

What does the current first page look like? Search the query and look at the format that ranks. Nine listicles means the format is settled. A mix of tools, forums and product pages means the intent is unresolved and there is room to do it better. This two minute check is worth more than any difficulty score.

Do you have the facts? Name the specific numbers, dates or examples the page needs, and who in your organisation supplies them. A question that passes the first three tests and fails this one goes on the list for next quarter, once you have gathered the facts.

Step 5: commit one page per question

Write the final list in priority order with the question as the working title, the intent, the page type, who supplies the facts and a review date. That document is the plan. Give each page one named number you will check on the review date, such as impressions for a question page or form submissions for a commercial one.

How to use difficulty scores

Each vendor calculates difficulty its own way, mostly from the links pointing at the pages currently ranking. Use the score for sorting a long list quickly, then make the real decision from the results page itself: who ranks, in what format, and whether you can produce something more specific than what is there. A moderate score on a query where the current results are thin is a better opportunity than a low score on a query with no path to revenue.

Why the long tail carries most of the wins

Specific questions are easier to answer completely and easier to answer better than anybody else. "Robots txt generator" is contested. "Robots txt rules for a staging subdomain" is a question you can own with one accurate page. Build the specific pages first, because they rank sooner, convert better and give you the material for the broader pages later.

What to do first

• Export three months of Search Console queries and pull every one ranking between eight and twenty.
• Ask your sales team for the ten questions they answer every week, and write them down verbatim.
• Check which of those ten already has a page. The gaps are your first briefs.
• Search your top five questions yourself and note the format that ranks for each.
• Name the facts each page needs and who supplies them, before any writing starts.

What this page covers, and what it leaves out

This page covers the method: sourcing, expanding, grouping, qualifying and committing to a page list. For writing the pages, for technical SEO and for local keyword work, use a dedicated guide for each. For current tool prices and limits, check the vendors directly, since both move.

Questions buyers ask about keyword research

Is keyword research still worth doing?

Yes, and the output is more durable than it used to be. Keyword tools tell you which wording real people use, and a question list built from that wording is the input for pages, briefs, FAQs and sales material alike.

What is the best free keyword research tool?

Google Search Console for your own queries, which no paid tool can replicate, and Google Keyword Planner for demand across a category, free with a Google Ads account. Add Google Trends for seasonality. That combination answers most questions a small site has.

How many keywords should I target per page?

One question per page, which naturally covers every wording of that question. A page answering six questions loosely tends to lose to six pages that each answer one tightly.

Why do two tools report different search volumes?

Each vendor models volume from its own data with its own method, and Keyword Planner reports banded ranges for low spend accounts. Pick one tool as your reference, use the figures to rank queries against each other, and take your own real numbers from Search Console.

Should I target high volume or low volume queries?

Work up from the specific to the broad. Specific questions rank sooner, convert better and prove the subject out, and the pages you build for them support the broader page when you write it.

Your keyword list, spoken aloud

Your keyword list is a list of questions, grouped one page to a question. The same questions get put to assistants, usually longer and chattier than the typed version, and the replies can name your competitors or you.

Take the top ten from the qualified list and ask them in ChatGPT, Gemini and Perplexity as a person would speak them. Note who is named and which pages are cited. AI Knows Us takes your list of questions and runs it across the assistants, then reports the names and sources, which makes the keyword work useful in a second place.