How to Do Keyword Research (Step by Step)

Keyword research is the process of finding the actual words and phrases your audience types into search engines, then deciding which of them to target based on relevance, demand, difficulty and intent. Done well, it tells you what to write, which pages to build, and where the realistic opportunities are — so you create content people are actually searching for rather than guessing. The method is a clear sequence: brainstorm seed terms, expand them with tools, evaluate each by volume, difficulty and intent, group related terms into clusters, and map each cluster to a page. This guide walks that sequence step by step, with a worked example and the free and paid tools for each stage.

It connects closely to what is search intent and how to optimize for it, because intent is the factor that decides what kind of page each keyword needs.

Why keyword research matters

Without research, content strategy is guesswork. You might write a beautifully crafted article on a topic nobody searches for, or chase a term so competitive your site has no chance of ranking, or build a page that answers a question the searcher was not asking. Keyword research replaces that guesswork with evidence: it reveals real demand (what people search and how often), real competition (how hard each term is to win), and real intent (what searchers expect to find). Armed with that, you invest your content effort where it can actually pay off — on terms that are relevant, achievable and aligned with what your audience wants.

It also underpins everything downstream. The keywords you choose shape your title tags and meta descriptions, your page structure, and your internal linking. Get the research right and the rest of your SEO has a solid target to aim at.

Step 1: Brainstorm seed terms

Start with seed terms — the broad, core topics that describe your business, products or expertise. If you run a website-analysis tool, seeds might be "technical SEO," "website speed," "tech stack detection," "AI search readiness." If you sell running shoes, seeds might be "running shoes," "trail shoes," "marathon training." These are usually short, high-level, and obvious; their value is as starting points for expansion, not as final targets.

To generate seeds, think about: the products and services you offer, the problems you solve, the questions customers ask you, and the language your audience uses (which is not always your internal jargon). Talk to sales and support, read customer emails, and look at the topics competitors cover. You are aiming for a list of perhaps ten to thirty seed topics that capture the breadth of what you do.

Step 2: Expand the seeds with tools

Seed terms are too broad and too competitive to target directly, so the next step is to expand each one into the many specific phrases people actually search. This is where tools earn their keep.

Google Keyword Planner (free with a Google Ads account) takes a seed and returns related keywords with search-volume ranges. Google autocomplete — the suggestions that drop down as you type a query — shows real, popular phrasings; type your seed and note the completions. The People Also Ask boxes and related searches at the foot of the results page reveal the questions and adjacent terms searchers use. Google Search Console is a particular goldmine: its Performance report shows the queries your site already gets impressions for, including terms you never deliberately targeted, which often points to easy wins.

Paid platforms scale this dramatically. Ahrefs, Semrush and similar tools take a seed and return thousands of related keywords, each with estimated volume, a difficulty score, and often the intent and the pages currently ranking. They also do competitor keyword analysis — showing which terms a rival ranks for, so you can find gaps. You do not need a paid tool to do good research, but they save enormous time at scale.

The output of this step is a long, messy list — hundreds or thousands of candidate keywords. That is fine; the next steps refine it.

Step 3: Evaluate by volume, difficulty and intent

Now you judge each candidate against three criteria.

Search volume estimates how many times a keyword is searched per month. It tells you the size of the prize. Beware of two traps: volume figures are estimates and ranges, not exact counts; and high volume often means high competition, so a big number is not automatically a good target.

Keyword difficulty estimates how hard it is to rank on page one, usually derived from the strength (authority, links, content quality) of the pages already ranking. A new or low-authority site cannot realistically win a high-difficulty head term, so difficulty must be weighed against your own site's authority. Choose terms that are winnable for you, not just terms that are popular.

Search intent is what the searcher actually wants — information, a specific site, to compare options, or to buy. It determines what kind of page can rank and whether the traffic is useful to you. A term with healthy volume and low difficulty is still the wrong target if its intent does not match what you offer. The four intent types and how to read them are covered fully in what is search intent and how to optimize for it.

The art is balancing the three. The sweet spot is usually moderate volume, low-to-moderate difficulty, and intent that matches your page — which, for most sites, points squarely at long-tail keywords.

Step 4: Embrace the long tail

Long-tail keywords are longer, more specific phrases — "best running shoes for flat feet" instead of "running shoes," or "how to check if a website uses Cloudflare" instead of "Cloudflare." They have three properties that make them the realistic path to traffic for most sites:

• Lower competition. Far fewer sites target the specific phrase, so a smaller, newer site can rank.
• Higher intent. A specific query usually signals a searcher who knows what they want, which often means they are closer to acting.
• Collective volume. Individually long-tail terms have low volume, but there are vast numbers of them, and together they account for the majority of all searches.

The strategic implication: rather than fighting giants for a handful of high-volume head terms, target many long-tail terms clustered around your topics. You will rank more easily, attract more qualified visitors, and accumulate meaningful total traffic. Head terms remain aspirational targets you may earn as your authority grows, but the long tail is where most sites should focus first.

Step 5: Group keywords into clusters

A long keyword list contains many terms that are really the same question phrased differently. "How to do keyword research," "keyword research steps," "keyword research guide" and "how to find keywords for SEO" all want the same thing — and they should be answered by one page, not four thin ones competing with each other.

So the next step is clustering: grouping keywords by the underlying intent and topic so that each group can be satisfied by a single comprehensive page. A cluster has a primary keyword (the main target, often the highest-volume term in the group) and secondary keywords (the variations and related questions the page should also naturally cover). Modern SEO favours this approach because Google understands topics and synonyms, so one strong page targeting a cluster outperforms many fragmented pages. Paid tools can cluster automatically; manually, you sort the list and group terms that a single searcher would consider interchangeable.

Step 6: Map clusters to pages

Finally, give every cluster a home. For each cluster, decide whether an existing page should target it (optimise that page) or a new page is needed (create it). This keyword-to-page mapping is the bridge from research to action. It prevents two common problems: keyword cannibalisation (multiple pages competing for the same term, splitting signals) and orphan keywords (terms you want to rank for with no page targeting them). The map becomes your content plan: a list of pages to create or improve, each with a primary keyword, its secondary keywords, and the intent it must satisfy.

A worked example

Suppose you run a website-intelligence tool and your seed term is "technical SEO." Here is the research condensed into a table — expansion, evaluation, and mapping.

Keyword	Approx. volume	Difficulty	Intent	Cluster / page
technical seo	High	High	Informational	Pillar: "What is technical SEO"
what is technical seo	Medium	Medium	Informational	Pillar: "What is technical SEO"
technical seo audit	Medium	Medium	Informational / commercial	Pillar (audit section) or dedicated guide
technical seo checklist	Medium	Low-Med	Informational	Same pillar (checklist section)
how to fix crawl errors	Low	Low	Informational	Long-tail support article
best technical seo tools	Low-Med	Medium	Commercial	Separate tools comparison page
technical seo for shopify	Low	Low	Informational	Niche long-tail article

A few lessons fall out of the table. The head term "technical seo" is high-difficulty, so you target it via a strong pillar page rather than expecting to win it quickly. Several terms ("what is technical seo," "technical seo checklist") share intent and cluster onto that one pillar as sections. "Best technical seo tools" has different (commercial) intent and needs its own page. And the long-tail terms ("how to fix crawl errors," "technical seo for shopify") are low-difficulty, winnable support articles that funnel authority to the pillar. That mix — one ambitious pillar plus a cluster of achievable long-tail supports — is a typical, effective structure.

Free vs paid tools: what you actually need

You can run thorough keyword research with free tools alone: Google Keyword Planner for volume and ideas, Search Console for the queries you already appear for, and autocomplete plus People Also Ask for real phrasing and questions. For many small sites that is genuinely enough.

Paid tools — Ahrefs, Semrush, and others — add three things worth paying for as you scale: breadth (thousands of keyword ideas per seed), difficulty scores (so you can prioritise winnable terms quickly), and competitor intelligence (seeing exactly which terms rivals rank for and where the gaps are). If keyword research is a regular, central part of your work, a paid tool pays for itself in time saved; if it is occasional, the free stack is fine. Either way, treat all volume and difficulty numbers as estimates — directionally useful, not precise.

How keyword research connects to GEO

The way people query AI answer engines is conversational and question-shaped — "what's the best way to…", "how do I…", "why does…". That makes the question-style long-tail keywords you uncover in research doubly valuable: they map directly onto the prompts users give ChatGPT, Perplexity and Google AI Overviews. Structuring content around real questions, with crisp answers, helps you rank in classic search and improves your odds of being cited in AI answers — see how to get cited by AI search engines. So the People Also Ask questions and long-tail queries you collect are not just SEO targets; they are a preview of how AI users phrase their needs.

Common keyword research mistakes

A few errors recur. Chasing volume alone, targeting popular head terms a site has no authority to win. Ignoring intent, building the wrong page type for the keyword and never ranking. Targeting one keyword per page in isolation rather than clustering related terms, which fragments effort. Keyword cannibalisation, where several pages fight for the same term. Treating estimates as facts, over-trusting precise-looking volume numbers. And never revisiting, treating research as one-off when search behaviour, competition and your own authority all change over time.

A keyword research checklist

• Brainstorm seed terms from products, problems and customer language.
• Expand each seed with Keyword Planner, autocomplete, People Also Ask and Search Console.
• Pull the queries you already get impressions for from Search Console.
• Evaluate candidates by volume, difficulty and intent.
• Favour winnable long-tail terms over unwinnable head terms.
• Cluster related keywords that share intent into single-page groups.
• Map each cluster to an existing or new page; avoid cannibalisation.
• Note the intent for each cluster so you build the right page type.
• Revisit research periodically as competition and authority shift.

Where to start

If you are starting from scratch, do this: open Google Search Console and list the queries your site already gets impressions for but does not yet rank well on — those are your fastest opportunities, because Google already considers you relevant. Pick three or four of them, expand each with autocomplete and People Also Ask to find the related long-tail questions, and cluster them. Then either improve the page that nearly ranks or create one that fully answers the cluster, matching the page type to the intent. That single loop — mine Search Console, expand, cluster, map to a page — is keyword research in miniature, and repeating it steadily builds a content plan grounded in real demand.

Go deeper

• Match the page to the searcher: what is search intent and how to optimize for it.
• Turn keywords into content: how to write SEO-friendly content that ranks.
• Write the click-earning snippet: how to write a title tag and meta description.
• Win AI citations too: how to get cited by AI search engines.

Curious what a site already ranks and is built for? Analyse any URL with StackOptic — one report covering SEO, tech stack, performance and AI-readiness, free, no sign-up.