How to build a keyword strategy: goals, seed keywords, intent, long tail, clusters, prioritisation, one keyword per page and adapting to AI prompts.

In short
a keyword strategy is the plan that decides which searches your site will try to win, with which page and in what order. You build it in seven steps: set business goals, collect seed keywords, identify the intent behind each query, expand to the long tail, group keywords into topic clusters, prioritise them by value and difficulty, then assign one main keyword per page in an editorial plan. In 2026 the same plan must also cover the conversational questions people ask ChatGPT, Gemini and Perplexity.
Want to go further? Ask
Without a keyword strategy, content gets written on instinct: some articles compete with each other, others target searches nobody makes, and the pages that could sell are missing. A strategy turns keyword research into decisions. It also changes how you read the data: a keyword is the visible trace of a need, and your job is to answer that need better than anyone.
Start from the business, not from search volume. Write down two or three goals (leads for a service, sales of a product range, sign-ups for a tool, local visits) and the audiences behind them. Every keyword you keep must serve one of these goals. A query with high volume and no link to your offer is a distraction. For each goal, also note the page that converts (pricing, contact form, product, booking), so that every informational article can lead readers to it.
| Intent | Example query | Page that fits |
|---|---|---|
| Informational | how to choose a CRM | Guide or blog article |
| Commercial investigation | best CRM for small business | Comparison, ranking, use case page |
| Transactional | CRM pricing, CRM free trial | Pricing page, product page |
| Navigational | [brand] login | Home, login or brand page |
| Local | accountant near me | Google Business Profile, local page |
Check intent by searching the query: the format Google already rewards (guides, lists, product pages, map) tells you which page can rank.
A long-tail keyword is a longer, more precise query with lower volume. Taken together, these queries represent a large share of searches, they are easier to rank for and they often convert better because the need is explicit. Expand each seed with question forms (how, why, which, can I), modifiers (best, cheap, for beginners, 2026, near me) and the People Also Ask questions.
Group keywords that share the same intent and would be answered by the same page: they belong together. Then group pages that share a theme into a topic cluster: one pillar page on the broad subject, several detailed pages around it, all linked to each other. A simple test: if the top 10 results for two keywords are mostly the same URLs, one page can target both.
| Criterion | Question to ask | Weight |
|---|---|---|
| Business value | Does a visitor from this query become a customer? | High |
| Search volume | How many searches per month, including variants? | Medium |
| Difficulty | Can your site realistically reach the top 5 against the current results? | High |
| Current position | Do you already rank between 5 and 20? | High (quick win) |
| Effort | New page, update or simple internal links? | Medium |
| AI visibility | Is the question also asked in ChatGPT, Gemini or Perplexity? | Medium |
Score each cluster on these criteria and start with those that combine high value, reachable difficulty and existing positions. Pages already ranking on page 2 are often the fastest gains.
When two pages target the same query, Google has to choose, and often ranks neither well: this is keyword cannibalization. Assign one main keyword (and its close variants) to each URL in a keyword map, a simple table with the keyword, the intent, the target URL and the status. Before creating a page, check the map; if a page already covers the intent, improve it instead.
Volumes from third-party tools are estimates: compare several sources and trust Search Console for your own site.
People ask AI assistants full questions with context ("which CRM should a 10-person agency choose if it uses Gmail?") rather than short keywords. These prompts resemble the long tail, with more detail. Add them to your research: note the questions customers ask, test them in ChatGPT, Gemini and Perplexity, and see which sources are cited. Pages that answer these questions directly, with clear sections and sourced facts, can rank on Google and be cited by the AI at the same time.
A keyword strategy decides which needs you will answer, with which page and in which order, so that every article serves a business goal. To automate the research part, Sorank analyses your site and finds your keywords automatically, then writes and publishes the matching articles on autopilot.
One main keyword, plus its close variants and the related questions that share the same intent. A single well-built page can rank for dozens or hundreds of queries if they all express the same need. Targeting unrelated keywords on one page dilutes relevance and usually ranks for none of them.
Keyword research collects data: queries, volumes, difficulty, intent. Keyword strategy turns that data into decisions: which keywords to target, which page answers each one, how they are grouped and in what order you publish. Research is the input; strategy is the plan you execute and measure.
Review it every quarter and after any major change, such as a new product, a redesign or a big Google update. Check which pages gained or lost positions, which new queries appear in Search Console and which topics competitors started covering. A strategy that never changes slowly drifts away from what people search for.
THEY TRIED SORANK BEFORE YOU
