Keyword research is the process of finding out which phrases your customers type into Google, how often they search for them and which page of your website should answer each one. The result is a spreadsheet you use to plan your site structure, your content and your PPC campaigns – based on data instead of guesswork.
How to do keyword research in seven steps:
- 1Collect – gather phrases from Google Keyword Planner, Ahrefs, Search Console, autocomplete and your competitors.
- 2Clean – remove duplicates, typos and queries that clearly have nothing to do with your business.
- 3Intent – decide what the searcher wants: information, a comparison or a purchase.
- 4Relevance – score each keyword on a scale of 1–5 (1 = most important, 5 = delete).
- 5Categorise – sort keywords into a category → subcategory tree.
- 6Map – assign each keyword to exactly one landing page, with its current position and a proposed title.
- 7Prioritise – order the work by search volume, relevance and difficulty.
What keyword research is and why to do it before a website or campaign
Good keyword research answers three questions: what people in your market search for, how many of those searches there are, and which page will answer them. Google's own SEO Starter Guide advises you to think about the words a user might search for to find your content. Keyword analysis turns that into a repeatable process.
Why do it at the start rather than after launch? Menus, URLs and e-commerce categories are expensive to change once a site is live. If you find out that people search for “custom kitchens Prague” but your site only has a generic “Products” page, you end up rebuilding. Done upfront, the research tells you which pages the site needs before anyone starts designing it. In PPC, data saves you paying for clicks that never turn into enquiries.
It is not about repeating phrases: Google understands synonyms, and keyword stuffing is against its spam policies. The research tells you what to write about and where it belongs.
What keywords are and which types to distinguish
Keywords are the phrases people type into a search engine when they look for something – from a single word (“web design”) to a full sentence (“how much does a website cost for a small business”). In SEO and PPC the term stands for a group of queries with the same meaning: a page targets one main keyword and its variants, not every query separately.
In keyword research it pays to tell five types apart:
- Head terms – one or two words with high search volume, such as “web design”. Competition is strong and the searcher's intent is unclear.
- Long-tail keywords – longer, more specific queries, such as “web design for car repair shops”. Each has low volume, but together there are many of them and the searcher usually knows exactly what they want.
- Local keywords – queries with a town or region, such as “accountant Leeds”. For businesses serving a specific area they are among the most valuable. How to rank for them is covered in our guide to local SEO.
- Branded keywords – queries containing your company or product name, typed by people who already know you.
- Questions – “how”, “how much”, “what is” queries. They are a natural basis for articles and FAQs.
The type tells you what content answers the keyword and how quickly you can rank for it. A head term with thousands of searches is a long-term goal; a long-tail query with a few dozen searches can bring enquiries within weeks of publishing the page.
Step 1: Collect keywords from several sources
No tool has complete data, so we merge several sources into one spreadsheet:
- Google Keyword Planner – a free tool inside Google Ads that suggests keywords and shows average monthly searches. It requires a Google Ads account with billing details, but no running campaign. Without regular ad spend you usually see only ranges (e.g. 100–1K), not exact figures.
- Ahrefs – strong for competitor analysis: it shows which keywords competing sites rank for and estimates the difficulty of each keyword (Keyword Difficulty) on a scale of 0–100.
- Google Search Console – the only source of real data about your own site: queries, impressions, clicks and average position for the last 16 months. It shows keywords you appear for but get few clicks on.
- Autocomplete and related questions – real queries from Google's suggestions and “People also ask”, often long and specific.
- Local data – for Czech and Slovak markets we add Marketing Miner, which focuses on data for these countries.
- Competitors – the menus, categories and blogs of 3–5 competitors who actually rank at the top. How to choose those competitors and what else to check is covered in our competitor analysis guide.
What each tool does and how they compare is covered in our overview of SEO tools for research and audits.
Google Keyword Planner step by step
For Google keywords the most accessible data source is Google Keyword Planner. It is free, but it only works inside a Google Ads account:
- 1Create a Google Ads account and add billing details. If the setup wizard pushes you to create a campaign in Smart Mode, switch to Expert Mode – Keyword Planner is not available in Smart Mode. The campaign does not have to run or spend anything.
- 2Open Tools → Planning → Keyword Planner.
- 3Choose one of the two options. “Discover new keywords” suggests phrases based on words you enter or on a website URL, including a competitor's. “Get search volume and forecasts” shows volumes for a list of keywords you already have.
- 4Check the location and language so that the data match the market you are researching.
- 5Review the results. For each keyword you see average monthly searches, the level of competition between advertisers and an estimated top-of-page bid. Narrow the ideas with filters (for example by keyword text or competition) and with the “Refine keywords” side panel.
- 6Download the results and paste them into your research spreadsheet.
Two things to keep in mind: “competition” in Keyword Planner describes how many advertisers bid in Google Ads, not how hard the keyword is in SEO. And search volume is an average over the selected period, so for seasonal keywords look at the month-by-month trend too.
How to find keyword search volume
Search volume is the average number of searches for a phrase per month. No tool knows it exactly – each one estimates it from different data. It is worth knowing what each number means and comparing only figures from one source within a single piece of research.
Several tools provide keyword search statistics:
| Tool | Data source | What it shows | Access |
|---|---|---|---|
| Google Keyword Planner | Google searches | monthly volume, often only ranges without ad spend | free, Google Ads account |
| Ahrefs | its own keyword database | estimated volume and Keyword Difficulty | subscription |
| Google Trends | Google searches | relative interest 0–100, not the number of searches | free |
| Google Search Console | your website | impressions and clicks of your pages | free, verified site |
For the Czech market we also use Sklik's keyword tool, which shows volumes on Seznam, the Czech search engine, and Marketing Miner for Czech and Slovak estimates. Three rules help when reading the numbers:
- Search Console impressions are not search volume. They show how often your site appeared in results. If you rank 30th, impressions are low even when the phrase is searched often. Working with this data is covered in our Google Search Console guide.
- Google Trends shows a trend, not volume. A value of 100 means peak interest for the selected period and region, not 100 searches. It is useful for seasonality and for comparing two phrases.
- Low volume is no reason to drop a keyword. For specialised services, phrases with a few dozen searches a month can be very valuable – relevance in step 4 decides.
Step 2: Clean and merge
Before you score anything, do four things with the raw list:
- Remove duplicates from the merged exports.
- Merge variants with the same meaning: spelling variants, typos and word order usually belong to one phrase and one page.
- Drop unrelated meanings – queries with “keyword” often come from programmers looking for keywords in Java or C.
- Take search volume from one source so the numbers are comparable.
Step 3: Search intent
The same topic can have different intents. Someone searching “how to do keyword research” wants a guide. Someone searching “pellet boiler price” is close to buying. Intent decides which type of page answers the query:
| Intent | Example query | Page type |
|---|---|---|
| Informational | how to choose a boiler | article, guide |
| Commercial | best pellet boiler | comparison, buying guide |
| Transactional | pellet boiler price | category, product, service |
| Navigational | brand name | brand page |
Quick check: search the keyword and look at page one. If it is all articles, a product page will struggle there, and vice versa.
Step 4: Keyword analysis by relevance – scoring 1–5
This is the core of the work and what separates keyword analysis from a plain tool export. Each keyword gets a score based on its importance for your business:
| Group | Meaning | What to do |
|---|---|---|
| 1 | most relevant keywords | main pages and categories |
| 2 | supporting keywords | articles, glossary |
| 3 | less relevant but interesting | later, when capacity allows |
| 4 | unwanted keywords | negative keywords for PPC |
| 5 | does not belong | delete |
Why five groups and not yes/no? Group 2 feeds the blog, which reaches people earlier in their decision. Group 3 keeps interesting keywords from being deleted just because they are not a priority now. Group 4 has direct financial value – see below.
Do not score by search volume. A keyword with 50 monthly searches can be a 1 because the people searching are ready to order, while a keyword with 5,000 searches can be a 5.
Step 5: Categorise into categories and subcategories
Keywords in groups 1–3 are sorted into two levels: category I (main topic, e.g. “Websites”) and category II (subtopic, e.g. “Pricing” or “Redesign”). The site structure can be read almost directly from this tree: category I items become menu items, category II items become subpages or articles.
For an online store, this shows whether your categories match how people search – for example by material rather than by manufacturer.
Step 6: Keyword mapping and preventing cannibalization
Each keyword is assigned to exactly one landing page. If no suitable page exists, we note a proposal for a new one. We then add the current position (from Search Console or Ahrefs) and a proposed title for that page.
The rule that follows, and that we stick to when writing: a page's text targets only the keywords the research assigned to it. Even if the page for another keyword does not exist yet, we do not add that keyword elsewhere “for now”. Once the new page launches, two URLs would compete for the same query. This is keyword cannibalization: Google may alternate between the pages, and often neither ranks as well as one strong page would.
On an existing site, Search Console reveals cannibalization: filter by a query and check how many different pages collect impressions for it. If there are several, pick the main one and rewrite or merge the others.
Step 7: Prioritise – volume × relevance × difficulty
Finally, you need to know where to start. We combine three factors:
- Search volume – how many people search for the phrase each month.
- Relevance – group 1 comes before groups 2 and 3.
- Difficulty – how strong the pages ranking for the keyword are. Ahrefs expresses this as Keyword Difficulty; in practice you confirm it by looking at the search results.
The fastest results usually come from group 1 keywords where you already rank in positions 5–20. The page exists and often only needs a better title, more content or internal links. New pages for difficult keywords are a long-term investment.
Keyword research template
The output of our research is one spreadsheet (Google Sheets or Excel), one keyword per row, with seven columns: keyword, search volume (from one source), relevance 1–5, category I and II, landing page (URL or proposed new page), current position and proposed title. An example for a web design studio (volumes and positions are illustrative, not real data):
| Keyword | Volume | Relevance | Category I / II | Landing page |
|---|---|---|---|---|
| web design Brno | 300 | 1 | Websites / Local | new: Web design Brno |
| website cost | 500 | 2 | Websites / Pricing | new article: Website cost |
| website redesign | 150 | 3 | Websites / Redesign | Web design service |
| free website | 900 | 4 | – | negative keyword in PPC |
| Keyword | Position | Proposed title |
|---|---|---|
| web design Brno | 18 | Web Design in Brno for Small Businesses |
| website cost | – | How Much Does a Website Cost and Why |
| website redesign | 34 | Website Redesign: When It Pays Off |
| free website | – | – |
In practice this is one filterable sheet. Filtering by “landing page = X” gives the copywriter a brief: exactly these keywords and no others.
What to do with the output
- Site structure – categories I and II plus the proposed new pages give you the menu and URL map. When we build a website, this is the starting point for the sitemap.
- Content plan – group 2 keywords feed articles and glossary entries. Each article gets its own set of keywords; how to turn that into text that works for people and Google is covered in our article on SEO copywriting.
- Optimising existing pages – on pages with assigned keywords, update the title according to the proposal, then the headings and content. If you are not sure what technical state the site is in, start with a website SEO audit.
- Negative keywords for PPC – group 4 goes straight into the negative keyword list in Google Ads. Your ads then stop showing for queries like “free”, “tutorial” or “jobs” if they never lead to orders. We handle campaign setup and management as part of PPC campaign management.
Common mistakes
- Choosing by volume – the most searched keyword in a field tends to be generic and brings people who do not buy.
- One keyword on several pages – leads to cannibalization, see step 6.
- Research left in a drawer – if nobody uses the spreadsheet, the money was wasted.
- Doing it once – search volume and competition change. It makes sense to update the research when preparing a new section of the site, or once a year.
When it makes sense to bring in a specialist
You can do basic keyword research for a small site with one service yourself. A specialist makes sense when you are building a new website or online store with dozens of categories, when the site is losing traffic and you do not know why, or when you want the research to serve directly as the brief for site structure, copy and PPC. The time invested pays back because the site does not need rebuilding after launch.
Keyword research is the first step of our SEO services and of every new website we build. Over 9+ years and 50+ projects we have refined the method so the output goes straight into production. If you want to discuss your site, get in touch through our contact form.
Frequently asked questions
How do you do keyword research?
What are keywords?
Is Google Keyword Planner free?
How do I find the search volume of a keyword?
What is keyword mapping?
What is keyword cannibalization?
How often should keyword research be updated?
Sources

Written by
Peter GáboríkProject Manager – AI, Automation & SEO
Project manager and founder of Weboptim. Builds growth strategies around SEO, automation and AI; in online marketing since 2017.
Related service
Tips that save time – once a month
One e-mail a month: a new guide, the automation of the month and one number from practice. No spam.



