Keyword Research

Keyword research is the process of understanding what your audience searches for, and using that understanding to shape content, architecture, and priorities across an SEO strategy.

What is keyword research?

Keyword research is the practice of identifying the search queries people use when looking for information, products, or services relevant to your site. It answers two questions: what are people searching for, and what do they actually want when they search it?

Done well, keyword research produces more than a list of terms to target. It reveals the shape of your audience’s intent. That insight drives content decisions, site structure, internal linking, and prioritisation. Every other pillar of SEO depends on it.

Is keyword research still relevant with semantic search?

Yes, but not as the exercise it used to be. Google matches the meaning behind a query rather than the literal words, so a page that answers a topic thoroughly ranks for queries it never states verbatim. Its systems resolve queries to entities, the people, places, products and concepts a query is actually about, and evaluate whether a page covers that subject rather than whether it repeats a string.

That changes what the research is for. It is demand research: it maps what people want and how they ask for it, then assigns that demand to pages. It is not a list of strings to place in copy at a given density. The two practices produce different artefacts. One produces a content plan; the other produces keyword stuffing, which stopped working long before AI search arrived.

So the parts that survive are the parts about people: intent, phrasing, the questions behind the query, and which of them your site can credibly answer. The parts that do not survive are the mechanical ones, exact-match repetition, keyword density targets, and related-term checklists. See semantic search for how the matching works, and entity SEO for making a site legible to systems that reason in entities rather than strings.

The keyword spectrum

Keywords sit along a spectrum from broad to specific. The spectrum is useful for judging how much intent a query carries, not for deciding which strings to repeat: the further along it a query sits, the more precisely it states what the searcher wants.

Head terms (also called seed keywords) are short, one- or two-word queries with high search volume: “SEO”, “keyword research”, “local SEO”. They are rarely worth targeting directly; competition is fierce, intent is ambiguous, and a single page cannot satisfy the full range of searchers a head term attracts. Their primary use in keyword research is as starting points: entering a head term into a keyword tool generates the broader range of related queries to assess.

Mid-tail keywords add one or two words of qualification: “keyword research tools”, “local SEO for restaurants”. Lower volume, clearer intent, more achievable competition.

Long-tail keywords are specific multi-word queries with low individual volume but high collective importance. Most distinct queries are long-tail, and for most sites they are the main opportunity. See long-tail keyword strategy for how to capture them at scale.

Core keyword research elements

These break into three stages: building the list, evaluating and prioritising it, and organising it onto pages.

Building the list

• Building a keyword universe. Assembling the complete set of search queries relevant to your business from multiple discovery sources, before any filtering or prioritisation.
• Competitor gap analysis. Identifying queries your competitors rank for that you don’t.
• Long-tail keyword strategy. Specific, lower-volume queries that make up most distinct searches and tend to convert better.

Evaluating and prioritising

• Search intent. The underlying goal behind a query: informational, navigational, commercial, transactional. Matching content type to intent is the most important factor in whether a page can rank.
• Search volume explained. What monthly volume figures mean, where they come from, and where they mislead.
• Keyword difficulty. How tools estimate ranking difficulty, what those scores actually measure, and when to trust them.
• SERP analysis. Reading the results page to understand what Google believes satisfies the intent of a query.
• SERP features. The feature types that appear alongside organic results, how each affects CTR, and how to target featured snippets and AI Overview citations.

Organising and mapping

• Topic clusters and pillar pages. Grouping semantically related keywords around a central theme rather than chasing terms in isolation.
• Keyword mapping. Assigning target queries to specific pages to prevent cannibalisation and surface content gaps.
• Keyword cannibalisation. When multiple pages compete for the same query, how to detect it, and when to consolidate, differentiate, or remove.

Why does keyword research matter?

Without keyword research, SEO is guesswork. You can produce content that is technically excellent and well-linked, but if it doesn’t align with how your audience actually searches, it won’t be found.

Keyword research also determines where to invest. Some queries are too competitive for your current domain authority. Some have intent that doesn’t match your offering. Some have volumes too low to justify the effort. Good research filters for opportunity as well as relevance.

For sites targeting more than one country or language, this process must be repeated per market. International keyword research treats each language and country as a separate exercise: search volume, dominant phrasing, and user intent differ significantly across markets, even for the same underlying topic. International SEO mistakes often begin here: applying source-market terms without checking local usage. A multilingual content strategy built on locally researched keywords usually outperforms one built on translations of source-market terms.

Keyword research and AI search

AI engines retrieve passages that answer specific questions. Long-tail, question-shaped queries map more cleanly onto retrievable passages than head-term keywords, which is why long-tail strategy matters more, not less, in the AI search era. The shape of demand has also changed: queries that were once two- or three-word fragments are increasingly typed in full sentences, especially in conversational interfaces.

Keyword research and E-E-A-T

Keyword research shapes which topics you cover and how deeply. Sites that build in-depth, expert-level content around a focused set of topics signal topical authority more effectively than sites covering everything thinly. That concentration of depth is what earns E-E-A-T signals over time: it is harder to be the best source on ten focused topics than a mediocre source on a hundred.

What about LSI keywords?

LSI (Latent Semantic Indexing) keywords are sometimes promoted as a keyword tactic: identify “semantically related” terms and work them into your copy to trigger Google’s understanding of the topic. The term names a 1980s information retrieval technique, and no Google documentation uses it. It is one of the more durable SEO myths.

The argument against it is technical rather than a denial. Retrieval matches meaning through embeddings, which is a different method from the one LSI names, so a checklist of co-occurring terms forced into a page is not what makes the page legible. The underlying instinct is still sound: writing about a topic fully and naturally will include the related terms, and thorough content performs better than thin content. Write for the topic, not for a related-keyword checklist.