Keyword Research: A Step-by-Step Process From Seed to Page

Keyword research is the process of finding the words people actually type into search engines, working out what they want when they type them, and deciding which page on your site should answer each group of searches. Done properly, it ends with a page-by-page plan, not a spreadsheet of terms, and every step below is aimed at that output.

This guide walks through the full process in order: seeds, expansion, intent, volume and difficulty, clustering, mapping and prioritisation. Two narrower topics already have their own guides on this blog, long-tail keyword discovery and competitor keyword analysis, so here they appear only as steps inside the larger workflow.

How to do keyword research step by step

The workflow has seven steps. Each one narrows the list and adds a decision, so skipping a step usually shows up later as a page that ranks for nothing or two pages that compete with each other.

1. Write down seed topics. List the products, services and problems your business actually handles, in the words a customer would use rather than internal product names.
2. Expand the seeds. Pull variations from Google autocomplete, the "People also ask" box, related searches at the bottom of the results page, Google Keyword Planner and the queries already reported in Google Search Console.
3. Classify intent. Label each term as informational, commercial investigation, transactional or navigational by looking at what Google currently ranks for it.
4. Judge volume and difficulty. Treat both as relative signals for comparing terms against each other, not as precise forecasts.
5. Cluster. Group terms that one page could satisfy, usually because the same URLs rank for them.
6. Map clusters to pages. Assign each cluster to an existing URL or to a new page you plan to create, with one primary term per page.
7. Prioritise. Order the work by business value, the gap between your current position and page one, and the effort needed to build or fix the page.

Step 1: start with seeds, not tools

A seed is a short phrase that describes a topic you can honestly serve, such as "accounting software", "office cleaning" or "visa for remote workers". Good seeds come from sales calls, customer emails, support tickets and the service list on your own site. Starting with a tool first tends to produce terms that are popular but irrelevant, which wastes the rest of the process.

Write seeds the way customers phrase them. A company might call its product a "workforce management platform" while customers search for "staff scheduling app". Both can be seeds, but the second is the one that will produce useful expansions.

Step 2: expand with free sources first

Every seed should be expanded into a longer list of real queries. The free sources are often enough to start:

• Google autocomplete: type the seed and note the suggestions, then add letters of the alphabet or question words such as "how", "what", "best" and "cost" after it.
• People also ask: each question you click open loads more questions, which is a quick way to collect the informational side of a topic.
• Related searches: the list at the bottom of the results page shows nearby phrasings and adjacent topics.
• Google Keyword Planner: inside Google Ads, "Discover new keywords" returns related terms with search volume. Accounts with little or no ad spend often see volume as broad ranges rather than exact averages, which is still usable for comparing terms.
• Google Search Console: the Performance report lists the queries your site already appears for, with impressions, clicks and average position. Queries with many impressions and a poor average position are some of the most valuable finds, because Google already associates your site with them.

Paid SEO tools add larger databases, difficulty scores and the ability to see which terms other domains rank for. That last feature is the basis of competitor keyword analysis, which works best as one input to this step rather than the whole method. Long-tail phrases, the longer and more specific queries, come out of this step naturally if you keep the question words and modifiers instead of trimming them.

Step 3: classify each term by intent

Search intent is the goal behind a query. It decides what kind of page can rank, so it matters more than volume. The reliable way to judge it is to search the term yourself and look at what Google already shows, because the results page reflects what Google has learned searchers want.

What search intent means and how to read it

Most frameworks use four intent types. The labels matter less than the page type each one implies, so the table below pairs them.

Read the results page for three signals. First, the page types: if the top results are all guides, a service page will struggle, and the reverse is also true. Second, the features: a map pack suggests local intent, shopping results suggest product intent, and a large "People also ask" block suggests the topic carries many open questions. Third, the angle: if most results are beginner explainers, an advanced piece may still rank but needs to cover the basics too.

Some queries have mixed intent, and Google shows a blend of guides and service pages. In that case, decide which half you want to win and build for it, rather than trying to make one page do both jobs badly.

Search volume vs keyword difficulty as concepts

Search volume is an estimate of how often a term is searched in a period, usually shown as a monthly average. Keyword difficulty is a score that SEO tools calculate to estimate how hard it is to reach the first page, typically based on the strength of the pages and backlink profiles already ranking. Both are estimates, and they are best used to compare terms with each other.

A few things are worth knowing about each number:

• Volume hides seasonality. A monthly average can blend a peak month and several quiet ones. Keyword Planner shows the monthly trend, which is worth checking for anything tied to holidays, tax deadlines or school terms.
• Volume counts the term, not the topic. A page that ranks well usually picks up traffic from many related phrasings, so a cluster's combined demand is more meaningful than one term's figure.
• Difficulty scores differ between tools. Each tool uses its own formula and its own backlink data, so a score from one tool cannot be compared with a score from another.
• Difficulty is relative to your site. A term that is hard for a new domain may be reachable for an established site with strong topical coverage. Your own Search Console data, showing where you already appear, is a better guide than any score.

A practical way to combine them: filter out terms with clearly no demand, then favour terms where intent matches what you sell, the difficulty looks reachable for your site, and the volume is enough to matter. Low-volume terms with strong buying intent often bring more enquiries than high-volume terms with informational intent.

How to cluster keywords

In practice, keyword clustering groups terms that a single page can rank for. Without it, you end up writing one page per phrasing, and those pages compete with each other for the same results, which is known as keyword cannibalisation.

There are two common ways to cluster:

• By results overlap. Search two terms and compare the top ten results. If several of the same URLs rank for both, Google treats them as the same topic and one page can target both. Many SEO tools automate this comparison across a whole list.
• By meaning. Group terms that share the same core topic and intent by reading them. This is faster for small lists but less reliable, because two phrases that look similar can have different intent.

As a hypothetical example, a bakery might find "birthday cake delivery", "order birthday cake online" and "same day birthday cake" all return many of the same local shops, so they form one cluster for one ordering page. "How to decorate a birthday cake" looks related but returns tutorials and videos, so it belongs to a separate informational cluster and a different page.

Each cluster gets one primary term, normally the one with the clearest intent and meaningful volume, plus secondary terms that appear naturally in headings and body copy.

How to map keywords to pages

Keyword mapping assigns each cluster to exactly one URL. The output is a simple sheet with columns for the cluster, its primary term, its intent, the target URL and whether that URL exists or needs to be created. This blog has a separate piece on keyword mapping as a discipline; the short version of the rules is below.

• One primary cluster per page. If two pages target the same cluster, merge them or change the focus of one.
• Match page type to intent. Transactional clusters go to service or product pages, informational clusters go to guides.
• Prefer improving an existing URL. A page that already has impressions for the cluster in Search Console is usually a faster win than a new page.
• Link the cluster pages together. Guides should link to the service page that fits the same topic, so informational traffic has a path to the commercial page.

Before mapping, it helps to know what is already on the site and how it performs. A technical and content review such as an SEO audit shows which pages are indexed, which ones overlap and which ones already have traction, so the map is built on the real site rather than an assumed one.

Step 7: prioritise what to build first

A finished map usually contains more work than a team can do in a month. Score each row on three things: business value (does this cluster lead to revenue), opportunity (how close is the current ranking, if any) and effort (a small rewrite or a new page with supporting content). Pages that score high on value and opportunity and low on effort go first. Hypothetically, a service page sitting at position 12 for a transactional cluster is often a better first task than a new guide for a high-volume informational term with no existing coverage.

Common keyword research mistakes

• Chasing volume alone. A large number attached to the wrong intent brings visitors who do not convert.
• Skipping the results page check. Intent assumed from the words alone is often wrong. Searching the term takes seconds.
• One page per phrasing. This creates thin, overlapping pages and splits ranking signals.
• Ignoring Search Console. Your own query data is first-party and free, and it shows terms where you are already close.
• Treating research as a one-off. Language, products and results pages change. Reviewing the map every quarter keeps it accurate.

Thai-language keyword research notes

Thai keyword research needs some adjustments, because the language and the way people search in Thailand behave differently from English.

• No spaces between words. Thai is written without spaces between words, so the same query can be typed with or without spaces, and tools may split it differently. Check variants with and without spaces before deciding that a term has no volume.
• Mixed script. Many Thai searches mix English and Thai, for example an English product or platform name followed by a Thai question phrase such as "คืออะไร" (what is it) or "ราคา" (price). Research both the English term, the Thai transliteration and the mixed form, since they can have different results pages.
• Transliteration variants. An English loanword can be spelled in Thai script in more than one way. Each spelling may have its own demand, and the results can differ.
• Question phrasing. Informational Thai queries often end in phrases like "คืออะไร", "ทำยังไง" or "ดีไหม", which map neatly to informational and commercial investigation intent.
• Set the right location and language. In Keyword Planner, set the location to Thailand and the language to Thai when researching Thai terms, otherwise the figures describe a different audience.

For bilingual sites, research each language separately rather than translating an English keyword list. The Thai version of a page should target what Thai searchers type, which is not always a direct translation of the English primary term. Teams working on the Thai market specifically can see how this fits into a wider programme on the SEO Thailand page.

What to measure after you publish

Keyword research is only as good as what it produces. After the mapped pages are live, use the Search Console Performance report filtered by page to check three things: whether the page appears for the queries in its cluster, whether its average position is moving toward page one, and whether it is picking up queries you did not plan for. New queries that keep appearing are a signal to expand the page or create a new one, which feeds the next round of research. Pair this with conversion data in your analytics so that rankings are judged by enquiries and sales rather than by traffic alone.

Content production is where most keyword plans stall, because the map produces a long list of pages to write or rewrite. A structured content marketing process with briefs built from each cluster keeps the writing tied to the research.

Frequently Asked Questions

How long does keyword research take?

For a small site with a handful of services, a first round of keyword research can usually be done in a few days, while large sites with many products take longer because clustering and mapping scale with the number of pages. The first pass is the slowest; quarterly updates are much quicker because the seeds and map already exist.

Is Google Keyword Planner enough for keyword research?

Google Keyword Planner is enough to get started, especially combined with Search Console, autocomplete and the results page itself. Its limits are that volume may show as ranges for low-spend accounts and that it does not estimate organic difficulty, which is where paid SEO tools add value.

Should I target high-volume or low-volume keywords?

You should target the keywords whose intent matches what you offer and whose difficulty your site can realistically handle, whatever their volume. Many sites get more enquiries from low-volume transactional terms than from high-volume informational ones, and a well-built page usually ranks for its whole cluster rather than a single term.

What is the difference between keyword research and keyword mapping?

Keyword research finds and evaluates the terms, while keyword mapping is the later step that assigns each group of terms to one page on your site. Research without mapping tends to produce lists that never turn into pages, and mapping without research is guesswork.

How often should keyword research be updated?

Keyword research should be reviewed at least every quarter and whenever you launch a new product or service. Search Console data makes the review fast, because it shows new queries your pages are already appearing for.

Turning keyword research into rankings

The process above works for any site, but it takes time to run well and to keep current. If you would rather have the research, mapping and page plan handled as part of an ongoing programme, Relevant Audience's SEO services cover that work from audit through to content and reporting.