How to Find the Words People Actually Search — Keyword Research Before You Write
The result first
Before writing an article, check the words people actually type into the search box — and let those words, not your own phrasing, lead the title. The check itself is simple: gather candidate words, look at how much each is searched and what the searcher wants, then choose. And treat every number honestly: an estimate stays an estimate, and a word you could not measure stays “unconfirmed.”
I am Kurashiki Bouzu (kurashiki-bouzu), born in 1971, a former funeral director now living in an old house in Kurashiki, Okayama, working with my AI butler Moka (Muse). Today’s topic is keyword research — the homework done before an article is written.
A “keyword” here means the exact words a reader types into the search box. This article explains how to find them, and how to use them.
Writers and searchers use different words
People looking for the same thing use different words depending on how much they already know about the topic. A writer tends to write in the writer’s own phrasing — but what lands in the search box is the searcher’s phrasing. Keyword research starts from the searcher’s side.
The material is of two kinds, and the difference matters. One is the record: which words already show your site in the results, and which get clicked. That is measured. The other is the forecast: how much a word you want to aim at is being searched. That is estimated.
The four numbers in Search Console
The record lives in Google Search Console, Google’s free tool for seeing how your site appears and gets clicked in Google search results. It looks at the search-results side — before the visit — not at what visitors do after they arrive.
For each query (the words actually typed), four numbers line up:
- Impressions: how often your link appeared in the results. This counts chances to be seen — not how often the article was read to the end
- Clicks: only the clicks that leave Google for an outside page are counted. Clicks that stay inside Google are not
- CTR (click-through rate): clicks divided by impressions — how often you were chosen out of the chances you had. It grades your title and description in the results, not just your rank
- Average position: roughly where your link appeared. The figure is an average of the topmost position of your link, so do not read it as an absolute — watch how it moves, especially sudden changes
One more caution: the newest data is still provisional and can change within hours. Numbers are a guide, not a mood.
Search intent — what the searcher wants to get done
Behind every query sits a purpose. Search intent is generally sorted into four kinds:
- Informational — to know, to understand, to learn how
- Navigational — to reach a specific site or page
- Commercial investigation — to compare and narrow down candidates before acting
- Transactional — to act: apply, buy, ask
The same words with a different intent need a different article. So candidates are sorted by intent, not by spelling.
The three steps of choosing a word
- Gather candidates. Start from a seed word and widen the list. The queries your site already appears for (in Search Console) are seeds too — demand and contact with your site are already measured there
- Check volume and intent. For each candidate, check how much it is searched (search volume) and what the searcher wants. Google Keyword Planner, the Google Ads tool, can find new candidates and show monthly search volume
- Choose. Weigh the readers you want to reach against the article you can actually write, and pick the word to aim at
Once chosen, the main word goes in the title and the opening. Different phrasings of the same question belong in the same article, received naturally in headings and text — you split articles where the question itself differs, not where the wording differs. One page, one intent. And never stuff the same word in over and over: it tires readers and breaks the rules.
Numbers, honestly — estimates are estimates
Search volume is an estimate, not an absolute answer; it varies by tool and conditions. In Keyword Planner, an account that is not running ads may only get a range such as “1,000–10,000.” A monthly average is an average of the last 12 months, so seasonal words also need their trend, not just the average.
So the house rule is: never dress a number up.
- Keep a figure in the shape you got it — exact number or range
- A word you could not measure is written down as “unconfirmed,” honestly
- When no exact figure is available, line up indirect signs — for example, major sites continuing to publish on the same word — labeled as indirect signs, with no claims about size
This thinking is built into our draft basics (decided 2026-11-13): every draft carries (1) a check of other blogs for the same angle, (2) the keyword search situation — the main words’ search counts and how they are searched for, or indirect signs recorded honestly when exact figures cannot be obtained — and (3) sources for any prices or fees. Item (2) is where keyword research lives.
Our own case — a core phrase replaced after checking the counts
This blog has replaced a phrase for exactly this reason. When we checked the words for our articles’ core phrasing, “AI導入 初心者” (“AI introduction, beginners”) turned out to be barely searched — 10 to 20 times a month or fewer. “Introduction” was the maker’s word; it was not what people typed.
So the core became “AI 使い方 初心者” (“how to use AI, beginners,” about 1,300 searches a month), with “素人” (“amateur / layperson”) swapped in as the alternative — decided on 2026-11-10. The searcher’s word, not the maker’s word, leads.
There is an exception: where a word’s very meaning is the point — as in the title of an already published article — we do not swap it just for search convenience. Numbers are one measuring stick, not a reason to break a word’s meaning.
What I do, and what Moka does
My part
- Hand over the purpose: which readers the article should reach
- Make the final decision on the target word and the title
Moka’s part
- Gather candidate words and check their search counts and how they are searched for
- Hand over the candidates together with the reason for each
- Record honestly: “unconfirmed” for what could not be measured, “indirect signs” for what is only a sign
An article’s reach is largely decided before the writing starts, in the choice of words. Start from the searcher’s words. Doing that obvious thing carefully is how keywords are researched in this house.
- November 21, 2026 ― English draft (result-first rewrite of the Japanese article). Draft created by reorganizing knowledge that had been kept as a draft into its current form.(5,298 characters)
- November 21, 2026 ― Published (all 7 languages at once, on the user's instruction on 2026-11-21 to publish the unpublished articles in every language). Japanese original: ledger No. 1818.(5,298 characters)
