How to Do Keyword Research in 2026: A 7-Step Method That Works

Quick quiz

How should you find the right keywords?

Question 1 of 425%

Where are you today?

I am starting and don't know what people searchI have ideas but need numbersI have keywords but no time to write

Knowing how to do keyword research decides whether your content brings customers or only visits. Most mistakes come from three shortcuts: targeting keywords too broad to rank for, writing one page per variant, and ignoring intent. Google's SEO Starter Guide puts it simply: think about the words a user would search for to find your content. This guide turns that advice into a seven-step method, with the tools and the decisions at each step.

What is keyword research?

Keyword research is the process of finding the searches your potential customers make, measuring their value (volume, intent, difficulty, business relevance) and deciding which page of your site should answer each one. It produces two things: a list of target queries and a keyword map that links each query group to a page. The full definition is in our glossary entry on keyword research.

How to do keyword research in 7 steps

1. List your seed topics

Write down your products and services, the problems they solve, the words customers use in calls and emails, and your competitors' category names. Each becomes a seed keyword: "accounting software", "invoice automation", "VAT for freelancers". Aim for 10 to 30 seeds.

2. Expand each seed into real queries

Use several sources, because each sees different queries:

• Google itself: autocomplete, People Also Ask and related searches (see Google Autocomplete for SEO).
• Google Keyword Planner: ideas and volume ranges from Google Ads data, as described in Google Ads Help.
• Search Console: the queries you already appear for in the Performance report, the best source for quick wins.
• Keyword tools: Semrush, Ahrefs and others for large lists and competitor keywords (compared in our ranking of keyword research tools).
• AI assistants: ask ChatGPT or Gemini how a buyer would phrase their need; these long questions become FAQ entries and GEO targets.

3. Check search intent

For each promising query, look at the current top results: guides, product pages, comparisons, tools, local results? The format that ranks shows what Google thinks the searcher wants. A query whose results are all product pages will not rank a blog post. Our glossary entry on search intent lists the main types.

4. Measure volume and difficulty

Volume tells you the potential; difficulty tells you the effort. A new site should favor long-tail keywords with modest volume and low difficulty, where it can rank within months, before attacking head terms. Our guides to keyword metrics and low-competition keywords explain how to read the numbers.

5. Group keywords into clusters

Queries that return the same results belong on the same page: "invoice template", "free invoice template" and "invoice template word" are one page, not three. Group by shared top results, then name each cluster by its main keyword. The free keyword clustering tool automates this step; see also keyword clustering.

6. Build the keyword map

Assign each cluster to one URL, existing or to create. Two clusters on the same URL means the page may be too broad; one cluster on two URLs means cannibalization. A simple map has five columns: cluster, main keyword, volume, intent, target URL. Our glossary entry on keyword mapping details the method.

7. Prioritize

Score each cluster on business value (how close it is to a sale), ranking chance and effort. Start with clusters where you already have impressions in Search Console (positions 8 to 20), then commercial clusters with low difficulty, then informational topics that support them.

How is keyword research changing with AI?

Buyers now ask assistants long, specific questions, and assistants split them into sub-queries before answering. Two consequences: long-tail and question keywords matter more, and each page should answer the related sub-questions in clear passages. Track a list of prompts linked to your clusters to see whether AI answers mention you (see our guide GEO vs SEO).

How often should you redo keyword research?

Do a full pass when you launch or change offers, then refresh every quarter: new queries appear, seasonality shifts and Search Console shows new opportunities. For the strategy around it, read our guide to keyword strategy.

Good keyword research ends with a short list of pages to create, each with one clear intent. Sorank, SEO and GEO software that runs on autopilot, runs this process for you: its keyword research finds and prioritizes keywords from your site and market, then it writes and publishes the matching articles. See the Sorank plans, with a 3-day free trial.