Keyword research: the process in 6 steps

What is keyword research?

Keyword research, also keyword analysis, is the systematic search for the terms your potential customers actually type into search engines, plus classifying them by relevance and intent. The result is a prioritized list you align your pages and content to. It stands at the start of almost every SEO effort, because without it you optimize on a hunch.

A keyword is not just a single word. It is any phrase or question someone uses to look for information, a product, or a service. The real value comes not from the word itself but from the question behind it: what does this person really want right now, and is my page the best answer to it?

Why search intent matters more than search volume

The most common beginner mistake is staring only at search volume. A keyword with 10,000 searches a month looks tempting but is worthless if the people behind it are only browsing and never buy. A keyword with 100 searches and clear buying intent brings you more real inquiries.

Behind every search sits an intent, the search intent. Google and the AI systems are now extremely good at recognizing this intent and showing exactly the pages that match it. That is why intent, not volume, decides whether a page ranks and whether the traffic converts. Classically there are three basic types:

Keyword research in 6 steps

There are many methods, but at its core every clean keyword research runs through the same six steps. Start broad, gather as many ideas as possible, then filter consistently by intent, feasibility, and value. The mistake is almost never in the gathering but in sorting out too early.

Step 1: gather seed keywords

Seed keywords are your starting terms, the rough topic areas of your business. HubSpot calls them topic buckets. Gather five to ten of them, without any tool, from your head and from real customer contact. Ask your sales and service team what words customers use to describe their problem. That language is what you want to hit, not your internal jargon.

Step 2: expand keywords with tools

Now you turn a few seeds into hundreds of ideas. The Google Keyword Planner delivers search volume and CPC, free with a Google Ads account. Google Suggest (autocomplete) and the "related searches" at the bottom of the page show real user phrasing. Google Trends reveals seasonal patterns. For questions, AnswerThePublic and AlsoAsked are strong, delivering the W-questions around a topic. Going deeper, you use suites like Sistrix, Semrush, or Ahrefs, which additionally provide keyword difficulty and competitor data. Do not forget autosuggest from YouTube, Amazon, and Bing, where people search differently than on Google.

Step 3: classify search intent

Go through your list and tag each keyword as informational, transactional, or navigational. The fastest test: enter the keyword in Google and see what ranks. If it is guides and definitions, the intent is informational. If it is product and offer pages, it is transactional. The search results themselves reveal Google’s verdict on intent, you do not have to guess. This step later decides what kind of page you even build.

Step 4: assess volume and difficulty

Now the numbers come in. Two values count: monthly search volume (how often it is searched) and keyword difficulty (how strong the competition is). The ideal combination is relevant, good volume, and low difficulty, but that is rare. For new or small websites there is a clear rule of thumb: start with the easy keywords, often long-tail, and work your way up to the hard ones. A first place on a niche keyword beats position 40 on a contested head term.

Step 5: cluster keywords into topics

Google today no longer thinks in single keywords but in topics. So you group keywords that share the same intent and the same topic into clusters, even when the wording is completely different. "b2b website cost", "what does a company website cost", and "price web design agency" belong in one cluster because the same question sits behind them. A cluster later becomes one page. This avoids several of your pages competing against each other for the same keyword, so-called keyword cannibalization.

Step 6: map keywords onto pages

The final step turns research into a plan. Each cluster gets exactly one target page, and the cluster’s intent determines the page type. Transactional clusters become service or offer pages. Informational clusters become guide or knowledge articles. The result is a keyword map: a table that assigns each cluster a URL, a main keyword, and a page type. This map is the blueprint for your entire website structure.

Short-tail vs. long-tail: why the niche wins

Short-tail keywords are short and broad, like "website" or "lawyer". They have high volume but brutal competition and unclear intent. Long-tail keywords are longer and more specific, like "get a b2b website built for mechanical engineering". They have less volume but less competition, clearer intent, and much higher conversion.

The numbers favor long-tail: around 70% of all searches are long-tail, only about 2.8% are a single word, and the vast majority of all Google searches have at least four words. Whoever reaches a person with "criminal defense lawyer munich" wins a more likely customer than with "lawyer". In B2B especially, where audiences are small and inquiries valuable, long-tail is almost always the better lever.

Our approach: take the search results apart

Tools deliver numbers, but the most honest source sits right in front of you: the search results page itself. Before we build a page, we google the target keyword and take apart what already ranks. Which page types are on top, guides or offers? Which subtopics do all top hits cover, which are missing? Which questions does Google ask in "people also ask"? That shows us the actual intent and the bar, instead of guessing it.

This reverse-engineering is exactly the core of our work. We do not write a page without first understanding why the current winners win. That is the difference between a page optimized on a hunch and one that deliberately fills the gap the competition leaves open.

Keyword research for AI search (GEO)

AI search changes research fundamentally but does not replace it. Three shifts matter for you. First: prompts are long and full-sentence. A query to ChatGPT averages around 23 words, a Google search around 4. People ask whole questions like "which crm for a 30-person b2b agency with hubspot integration under 150 euros per user". That is long-tail in its purest form, only more extreme.

Second: there is no longer historical volume for everything. Around 15% of daily searches are brand new queries with no history, even more in AI search. Pure volume tools run empty here. You have to think in topics and questions, not in single keywords with a volume label.

Third: complement your research with real customer questions and the prompts people use to describe your problem. Ask the AI systems themselves which questions come up around your topic, and cover them with clear answers. How you become visible in AI answers with this is explained in our article on generative engine optimization.

Common mistakes

Most wasted work in keyword research goes back to a few recurring mistakes. Knowing them saves you months aimed at the wrong target.

Only looking at volume

High search volume is tempting but worthless without matching intent. A keyword with lots of volume that only brings browsers fills your stats, not your pipeline. Always check what the people behind the number really want before committing to a keyword.

Ignoring search intent

If your page type does not match the intent, you will not rank, no matter how good the text is. Optimizing an offer page for an informational keyword where Google only shows guides is wasted effort. Always let the actual search results decide what kind of page you build.

Keywords that are too broad

Beginners love to optimize for the shortest, biggest keywords, exactly the ones with the fiercest competition. For a new website that is the surest way to never become visible. Go granular: more specific keywords bring more relevant traffic and can actually be won in the first place.

Your first steps

The five steps above are the lean version of the process you can start with right away: gather, expand, sort by intent, cluster, map onto pages. Start with the easy long-tail keywords, win first rankings there, and work your way up to the contested terms.

Research is not a one-off project. Review your keywords every few months, because search behavior, competition, and the AI systems change fast. This is exactly where our work starts: we build B2B websites whose structure emerges from a clean keyword map from the start, instead of gluing keywords on afterwards. That way every page matches exactly one real search intent.

Frequently asked questions