A keyword research method is the process of deciding what your website should write about, based on what your customers actually type into Google. The deliverable is not a keyword list. It is a schedule of pages, where every page owns one search intent, so Google has a reason to recommend it.
What keyword research really is, in one paragraph
- Keyword research means finding queries that have real demand and match your business, using your customers’ own search language (see the Google Search Central SEO Starter Guide).
- Identify search intent first, then decide whether the answer should be a service page, a case study or an article.
- Search volume validates the size of demand. It does not replace knowing your customer.
- The output has four linked parts: keyword group → page → content → schedule.
Step 1: Start from seed keywords and customer language
Seeds do not have to be precise. Their job is to draw a boundary. Three sources work best:
- Customer language: pull the exact words clients use in support chats, enquiry emails and sales notes, not your internal product names.
- Service inventory: list the services, industries and regions you actually want, to build 10 to 30 seeds.
- Existing data: the Queries report in Search Console shows which searches already surface your site. It is free and closest to reality.
Step 2: Let search intent decide the page type
The same topic can need different pages depending on intent. Google’s Search Quality Rater Guidelines group queries roughly into “know”, “do”, “website” and “visit in person”; the working shorthand most teams use is informational, navigational, commercial investigation and transactional.
- Informational (“what is a keyword research method”) → articles and guides
- Commercial investigation (“best SEO agency in Hong Kong”) → service pages, comparison pages, case studies
- Transactional (“website design quote”) → service and pricing pages
- Navigational (brand terms) → homepage and about page
Getting intent wrong wastes budget in two ways: an article that attracts readers who never enquire, or a service page that never ranks because nobody searching that way is ready to buy.
Step 3: Validate demand with tools, not intuition
- Google Keyword Planner: free, shows volume ranges and related terms
- Google Search Console: the queries your own site already appears for
- Google autocomplete and People Also Ask: closest to how people phrase things
- Ahrefs / Semrush: keyword difficulty and the terms competitors rank for
Two data points matter, because they are often misread:
- The long tail is the majority. Backlinko’s analysis of 306 million keywords found that 91.8% were long-tail terms of three words or more, and 94.74% of keywords get fewer than 10 searches a month. A term with ten monthly searches is not useless; it accumulates as part of a content group.
- New demand keeps arriving. Ranktracker’s 2025 figures suggest roughly 15% of Google searches are queries never searched before, which is exactly why historical volume misses emerging demand.
Step 4: Cluster keywords and map them to pages
Group keywords that share one intent, then give each group one page. The test is simple: if two keywords on the same page would look odd to a reader, split them; if they answer the same question, merge them.
- One group = one page = one primary intent plus close variants
- Put the primary keyword in the title, the opening paragraph and at least one subheading, then let the rest appear naturally
- Never let two pages chase the same term, or Google chooses for you
Step 5: Prioritise
Not every keyword deserves to be first. Score each group on four dimensions and write one sentence per line:
| Dimension | Question to answer |
| Distance to conversion | How far is the searcher from enquiring or buying? |
| Business value | Is this the kind of work we actually want? |
| Competition | Who ranks today, and do we have data or cases to compete? |
| Available resources | Can we publish a better answer within two to four weeks? |
The four most common mistakes
| Common mistake | What works instead |
| Delivering a keyword list with no pages attached | Assign one page per keyword group, with an owner and a date |
| Treating brand terms as the goal | Brand terms already rank; spend the effort on words newcomers search |
| Creating one page per keyword | Merge synonyms. One deep page beats ten thin ones |
| Doing it once and never revisiting | Review quarterly, and after new services or ranking shifts |
How AI search changes the method
With AI Overviews now common, keyword research has one more goal: being cited, not just ranked.
- Intent still governs everything, because AI answers are assembled around the same intent.
- The Princeton and IIT Delhi study (Aggarwal et al., KDD 2024) tested nine optimisation tactics and found that adding verifiable citations and specific statistics lifted visibility in AI-generated answers by roughly 30–40%.
- In practice: write the paragraph that owns your keyword so it can stand alone. State the conclusion first, then the reasoning and the numbers.
FAQ
How often should keyword research be repeated?
Quarterly is a sensible baseline, plus ad hoc when you launch a service, enter a market, or see a page’s rankings move. Doing it regularly matters more than doing it exhaustively once.
Can we do this without paid tools?
Yes. Search Console, Google autocomplete and People Also Ask cover most of what a small or B2B business needs. Paid tools mainly save time and reveal competitor keywords.
Are more keywords better?
No. Keywords are an input; a page that answers a customer question is the output. Ten well-mapped groups beat five hundred terms with no page behind them.
Next step
If you already have a keyword list but cannot decide what to do first, or you suspect several pages are competing for the same term, start with a website review: we compare search intent against your existing page structure and point out the gaps worth closing first. Further reading: What is SEO? A practical Google SEO guide for Hong Kong businesses, and B2B content marketing: connecting keywords to enquiries.
Written and reviewed by the Jionghui web team. Last updated: 22 September 2026.