Keyword research used to mean one thing: find the phrases with the most searches and write pages for them. That no longer works. People now search in full questions, get answers from AI summaries before they ever click, and expect a page to solve their exact problem. Good keyword research in 2026 is less about chasing volume and more about understanding what people need, grouping those needs into topics, and matching each one to the right page. This guide walks through the whole process.

Key Takeaways

• Start with seed keywords from your customers, your site, and your competitors — then expand them into topic clusters, not a flat list.
• Check search intent and the live results page before committing to any keyword; the SERP tells you what Google thinks the searcher wants.
• Prioritise by business value — relevance, intent, and likely enquiries or sales — rather than by search volume alone.
• Cover the AI search layer: clear entities, direct answers to real questions, and content structured so AI assistants can cite it.
• Map one primary keyword and intent to each page, keep a simple keyword map, and review real query data in Google Search Console.

What Is Keyword Research — and Why Does It Work Differently in 2026?

Keyword research is the process of finding the words and questions your customers actually use when they search, then deciding which of those searches your website should answer. In 2026 the emphasis has shifted from raw search volume to topics and intent, because people search in full questions and often get an AI-generated answer before they ever click.

The practical change is in the question you ask. Instead of “which single phrase should this page target?”, you now ask “what is this person trying to get done, and what would a complete, satisfying answer look like?” Google’s own SEO starter guide makes the same point: think about the words users might use, but build content for people first. Treat keyword research as market research that happens to use a search box, and the rest of the process falls into place.

Where Do You Find Seed Keywords to Start From?

Seed keywords are the short, core terms that describe what you sell — the starting points you later expand into a full list. The best sources are close to home: the words customers use on calls and in emails, the services on your own website, the queries already appearing in Google Search Console, competitor pages, and Google’s own autocomplete suggestions.

Work through them in order. First, write down your services in plain customer language — “emergency plumber” rather than “rapid-response plumbing solutions”. Second, open Search Console and look at the queries your site already appears for; these are proven seeds, because real people already associate them with you. Third, check a few competitors: their ranking pages usually reveal seeds you have missed. Finally, type each seed into Google and note the autocomplete suggestions and related searches. Tools such as Google Keyword Planner, Ahrefs, or Semrush can expand and organise this list, but the seeds should come from your market, not a tool’s database.

How Do You Group Keywords Into Topic Clusters?

A topic cluster is a group of related keywords organised around one central theme: a main pillar page covers the broad topic, and supporting pages cover the narrower questions and subtopics around it, all linked together. Grouping keywords this way mirrors how people explore a subject — and how search engines judge whether a site genuinely covers a topic.

To build a cluster, sort your expanded list by shared intent and shared subject, not by similar wording. “How much does local SEO cost”, “local SEO pricing” and “is local SEO worth it” belong together because one page can answer all three well — our own guide to what local SEO costs was built exactly that way. A practical cluster has one pillar topic and a handful of supporting subtopics. If two proposed pages would give the reader the same answer, they are one page, not two — a rule that also protects you from cannibalisation later.

Why Should You Analyse the SERP Before Choosing a Keyword?

The results page is the clearest evidence of what Google believes a searcher wants, because it shows the pages and features Google already rewards for that query. If the results are dominated by product pages, a blog post will struggle however well it is written; if they are guides and comparisons, a sales page is simply the wrong format.

Do this for every serious candidate. Note the types of pages ranking (guides, tools, product pages, local listings), the features present (AI Overviews, featured snippets, video, “People also ask” boxes), and the angle the top pages take. Then look for the gap: a question they answer poorly, an audience they ignore, or advice that is out of date. That gap is your opportunity. If there is no gap and the results are locked up by huge sites serving a different intent, move on — SERP analysis is as much about the keywords you reject as the ones you choose.

What Are the Main Types of Search Intent?

Search intent is the goal behind a query, and it falls into four main types: informational, where the person wants to learn something; navigational, where they want a specific site; commercial, where they are comparing options; and transactional, where they are ready to buy, book, or enquire. Matching your page to that intent matters more than matching exact words.

Learn to read the signals. Words like “how” and “guide” usually mark informational intent; “best”, “vs” and “review” signal commercial research; “buy”, “price”, “near me” and a service plus a location point to transactional intent. One keyword can also carry mixed intent — “keyword research” is mostly informational, but many searchers also want a tool. When intent is mixed, satisfy the dominant intent first and serve the secondary one further down the page. If you are unsure, the SERP check from the previous step settles it: the formats Google ranks are the intent it has detected.

Why Do Long-Tail and Question Keywords Matter More Now?

Long-tail keywords are longer, more specific phrases — often full questions — that individually attract fewer searches but describe a much clearer need. They matter more now because voice search and AI assistants have trained people to ask complete questions, and because a specific query usually comes from someone who knows exactly what they want and is closer to acting.

Build questions into every cluster deliberately. Mine the “People also ask” boxes on your target SERPs, customer emails, sales calls, and Search Console queries for the exact phrasing people use. Then answer each worthwhile question directly and early on the relevant page — a clear, self-contained answer in the first lines under the heading. That habit wins featured snippets, satisfies readers who skim, and, as the next section explains, makes your content usable by AI assistants too. Depth on specific questions also compounds: ten precise answers usually outperform one vague page chasing a single broad term.

How Has AI Search Changed Keyword Research?

AI search adds a second audience to your research: the AI assistants and AI-generated summaries that read your content and decide whether to cite it. People increasingly ask tools like ChatGPT and Perplexity full questions instead of typing short keywords, so research now also covers entities and the real questions people ask assistants.

Practically, three things change. First, keep your entity information consistent everywhere: your business name, services, locations, and descriptions should say the same thing on your site and across the web, so an AI system can describe you accurately. Second, structure content so it can be quoted: short, direct answers under clear question headings, definitions stated plainly, and claims you can stand behind. Third, track the questions customers ask in conversation — those are increasingly the exact prompts typed into AI tools. We cover the citation side of this in depth in our guide to getting your business cited by ChatGPT and Perplexity. None of this replaces traditional research; it layers on top, because AI answers are still built from content search engines can find and understand.

How Should You Prioritise Keywords by Business Value?

Prioritise keywords by business value: how relevant the searcher is to what you sell, how close their intent is to enquiring or buying, and how realistic it is to compete. Search volume belongs in the decision, but last — a low-volume keyword from people ready to spend beats a popular one from people who will never buy.

A simple scoring pass keeps this honest. For each candidate keyword or cluster, rate relevance to your services, intent strength, and whether you can genuinely make the page the best answer available. Start with high-relevance, high-intent terms — usually service-plus-location and problem-plus-solution phrases — then add the informational cluster content that supports them. Research also meets budget here: understanding what your wider SEO work costs and returns helps you decide how many clusters you can build properly instead of spreading thin content across all of them. If you would rather hand the whole process over, it is a core part of our SEO services.

How Do You Map Keywords to Pages Without Cannibalisation?

Keyword mapping means assigning each cluster — and one primary keyword within it — to a single page, so every page has a distinct job. Cannibalisation is the opposite problem: two or more of your pages target the same keyword and intent, compete with each other, and force Google to guess which one you actually want to rank.

Keep a simple keyword map: a spreadsheet with one row per page, listing its URL, primary keyword, intent, and the supporting questions it answers. Before creating any new page, check the map — if an existing page already owns that intent, improve that page instead. Cannibalisation often creeps in through well-meaning blog posts that repeat a service page’s topic, or near-identical location pages. The fix is rarely deletion: merge overlapping pages into the strongest one, redirect the spare URL, and differentiate pages you keep by intent — one informational, one transactional, never two doing the same job. Review the map whenever you add content, and sanity-check it in Search Console, where two URLs earning impressions for one query is the classic warning sign.

How Do You Measure Whether Your Keyword Research Is Working?

Measure keyword research with Google Search Console, because it shows the actual queries your pages appear for — real data about your site, not estimates about the market. Its Performance report lists impressions, clicks, click-through rate, and average position per query and page, and the gaps it reveals are where your next round of research should start.

Work through Search Console roughly monthly. Following Google’s own Performance report guidance, compare periods, filter by page, and read the queries tab for wording you did not plan for — those unplanned queries are next month’s research, straight from your audience. When impressions grow for a cluster and clicks follow, the research was right; when a page attracts entirely the wrong queries, its intent mapping needs revisiting. Also watch branded versus non-branded queries, and whether your question-led content is being picked up in AI answers — which increasingly shows up as referral visits and branded searches rather than classic rankings.

Keyword research in 2026 is a loop, not a one-off task: listen to customers, group their needs into topics, check what the results page rewards, publish the best answer you can, and let real query data refine the next round. Sites that work this way stop guessing — and Google’s guidance points the same way, towards helpful, people-first content built around genuine needs.

Frequently Asked Questions

How long does keyword research take?

For a small website, a solid first pass — seeds, expansion, clustering, SERP checks, and a keyword map — usually takes a few focused hours spread over a couple of days. The ongoing part is lighter: a short monthly Search Console review. Businesses with many services or locations should expect longer, because each service deserves its own cluster.

Can you do keyword research without paid tools?

Yes. Search Console shows the real queries your site already appears for, Google’s autocomplete and “People also ask” boxes reveal what people search around any seed term, and Keyword Planner is available with a Google Ads account. Paid tools make expansion and competitor analysis faster, but the core work — intent, clustering, and SERP checks — can be done entirely with free sources.

How many keywords should one page target?

One primary keyword and one clear intent per page, supported by the related questions and variations that naturally belong to the same topic. A page that tries to rank for several unrelated keywords usually ends up answering none of them well. Think in clusters instead: one pillar page for the broad topic, and supporting pages for each distinct subtopic or question that deserves its own answer.

What is the difference between a keyword and a topic cluster?

A keyword is a single search phrase; a topic cluster is the organised group of related keywords and questions around one theme, mapped to a pillar page and supporting pages. Targeting a cluster helps you cover a subject completely, gives internal links a logical structure, and reduces the risk of your own pages competing with each other.

Does keyword research still matter with AI search?

Yes — it matters differently. AI assistants and AI-generated summaries still draw on web content, and the questions people ask those assistants are keyword research in a new form. Research now also covers entities and conversational questions: defining your business and services clearly, answering real questions directly, and keeping information consistent so AI systems can cite you accurately alongside traditional rankings.

How often should you repeat keyword research?

Treat it as a loop rather than a one-off project. Do a light review monthly in Search Console — new queries, rising impressions, pages attracting the wrong searches — and a deeper refresh whenever you launch a service, enter a new location, or notice competitors changing what they publish. A list written once and never revisited quietly goes stale.