Build a keyword map from customer questions

Collect language before opening a tool

The best source of keywords is the people who already want what you sell. Read the last 30 enquiries, call notes, quote requests and reviews. Note the exact phrases customers use to describe the problem, the options they were weighing and the doubts that held them back. Add the queries from your own site search and the ones Search Console already shows for your site. Customers rarely use the vocabulary on your price list, and the gap between the two is where pages get missed.

Google’s starter guide makes the same point: users who know a topic well may search with different words from someone new to it, and its example is “charcuterie” against “cheese board.” It adds that you do not need to anticipate every variation, because Google’s language-matching systems can relate a page to many queries without the exact terms.^[1] The practical lesson is to record the phrases your customers really use, and to stop chasing every synonym.

Expand and qualify with the tools you have

Search Console is the one tool that shows queries your site already appears for. Its Performance report gives clicks, impressions and average position, where position is that of the topmost result from your site for the query.^[2] Look for queries with many impressions and few clicks, and for pages that appear for a query they were not written for. Remember what the report omits: it leaves out rare queries to protect privacy and stores top rows rather than every row, so it understates the long tail.^[2]

Google Ads’ Keyword Planner is the next step. Give it seed words or your own URL and it returns keyword ideas with estimated volumes. Two cautions apply. It was built to plan advertising, so treat its numbers as estimates and not as measured demand. And keywords with very low volume, or considered sensitive, are not discoverable or forecastable.^[3] A phrase that returns no data may still be what your customers type. Finally, read the results page for the query itself: the questions, related searches and headings on pages that already rank show how the topic is usually broken down.

Read those numbers as a warning about how skewed tool data is, not as an instruction. Most distinct phrases are rare, yet a few head terms hold most of the measured volume, so any list sorted by volume leads you to the most contested terms. Volume also says nothing about value. A query with 30 searches a month from people ready to book can be worth more than one with 3,000 from people who want a definition. The study defines long-tail by volume (1 to 100 searches) and covers US data from one tool vendor, so treat it as indicative.^[4]

Group by the job and by the evidence

Do not group by lookalike wording. Group by whether one page could satisfy everyone who typed the phrases. “Sports massage benefits,” “what does sports massage do” and “is sports massage worth it” are one job: understanding the treatment. “Sports massage price” and “book sports massage” are another: acting. A practical check is to search two phrases and compare the results. If the same pages answer both, they probably belong on one page. If the results differ, the jobs probably differ. That test is an editorial heuristic, not a rule from any search engine.

One page can serve many phrases, and publishing a page per variant produces duplicates. Ahrefs found that the average page ranking first also appears in the top ten for nearly 1,000 other keywords, with a median of about 400.^[6] The aim is therefore a short list of strong pages, each with a primary phrase, which is the customers’ most natural wording, and a set of secondary phrases it should address naturally.

Keep every phrase in plain, readable text. Google’s spam policies list keyword stuffing, meaning filling a page with keywords or numbers in an unnatural way, as a violation.^[7] A primary phrase belongs in the title and the first heading because it describes the page, and the rest should appear because the explanation needs them.

Prioritize a small working set

A keyword map can hold fifty rows. The plan for the next quarter should hold five to ten. Score each group against four questions and choose the groups that clear all four.

Be careful with “keyword difficulty” and similar scores. They are vendors’ own estimates, and Google states that no third-party tool has access to its internal ranking or AI systems.^[8] Use them to compare options, then check the results page by eye. For each chosen group, write a brief using the content brief chapter, and record the page, the publishing date and the baseline impressions. Google’s starter guide suggests waiting a few weeks before judging a change, so schedule the first review for then.^[9]

Apply it to your website

• You have a language log built from at least 30 real enquiries, calls or reviews, in the customers’ own words.
• Search Console queries and impressions are in the map, and you know what the report leaves out.
• Tool volumes are treated as estimates, and phrases with no reported volume are kept if customers use them.
• Phrases are grouped by the job a single page could satisfy, and each group has one primary page.
• The next quarter’s plan holds five to ten pages that score well on value, evidence, feasibility and readiness.
• Every chosen page has a brief, a baseline measurement and a review date.