Keyword research is the step where you decide what to write about before you write anything. If you pick the wrong terms, even a well-written article can sit unread. If you pick terms that match what people actually search for, each article has a clear job. This guide explains how to collect keywords for research, sort them by intent, and decide which ones deserve a page on your site. It is written for founders, small business owners, and small marketing teams who publish on WordPress or a GitHub-based site and have limited time.
The short answer: start from your own business, not from a tool's list. Write down what you sell, who buys it, and the problems they try to solve. The rest of this article walks through each step with a reader-run check you can do today.
What a keyword research list is actually for
A keyword is a phrase a person types into a search engine. Keyword research is the work of finding those phrases, estimating what the searcher wants, and mapping each phrase to a page that can satisfy that want. The output is not a spreadsheet of impressive numbers. It is a short plan: which page answers which question, and in what order you write them.
For a small site, the most useful output has three columns: the phrase, the intent behind it, and the page it will live on. If a phrase has no clear page, leave it out for now. If two phrases would be answered by the same page, merge them.
Step 1: Start with your business facts
Most weak keyword lists come from starting with generic industry terms. A better starting point is a short, factual description of your business:
- What you sell, in plain words
- Who buys it and in what situation
- The questions your customers ask before they buy
- The problems that bring them to you
- Pages you already have, such as services, pricing, or a booking page
Write these down as seed topics. For example, a hypothetical dental clinic might list "teeth whitening," "dental implants cost," and "emergency dentist." A hypothetical accounting firm for freelancers might list "quarterly tax payments" and "invoice late fees." These seeds are not keywords yet. They are the raw material you expand in the next step.
Checking your own list against your site matters too. If your homepage describes a service you no longer offer, search terms about that service will bring visitors who cannot buy from you. Confirm the facts before you collect terms.
Step 2: Expand seed phrases with search suggestions
Search engines suggest completions as you type. Those suggestions reflect phrasings people have used, so they are a practical source of real wording. For each seed, try:
- The seed on its own
- The seed followed by a letter or a common modifier such as "cost," "how long," "vs," "near me," or "for beginners"
- A question form starting with "how," "what," "why," "when," or "can"
- The "related searches" block on the results page
Record every suggestion that describes a problem you can address. Ignore suggestions about unrelated brands or topics. This is manual work, but it costs nothing and it keeps you close to the language your readers use.
Your own search data is a second source. Those are often more reliable starting points than a guess, because they show what Google is already testing against your content.
Step 3: Label the intent behind each phrase
Two phrases can look similar and need different pages. "Dental implants" and "dental implants cost" may both be relevant to a clinic, but the second searcher is closer to a booking decision. Before you assign a page, label each phrase with one of these intents:
- Informational: the person wants to learn something, such as "how long does teeth whitening last"
- Comparison: the person is choosing between options, such as "Invisalign vs braces for adults"
- Transactional or local: the person is ready to buy, hire, or book, such as "emergency dentist open now"
Informational phrases usually suit a guide or explainer. Comparison phrases suit a page that sets out criteria and trade-offs. Transactional phrases usually belong on a service or booking page, not a blog post. Mixing these up is one of the most common reasons an article ranks poorly: the page answers a different question than the one typed into the search box.
Verify intent yourself. Search the phrase, open the top results, and note what they have in common: guides, product pages, directories, or forums. If the results are all guides and your planned page is a sales page, the mismatch is worth fixing before you write.
Step 4: Group related phrases into one page
Several phrases can share one page. "How long does teeth whitening last," "does teeth whitening damage enamel," and "teeth whitening results how long" are all answered by a single well-structured guide. Group phrases when a reader searching any of them would be satisfied by the same article. Split them into separate pages only when the answers differ enough that one page would become unfocused.
A simple test: write the one question the page answers, in a single sentence. If you need two sentences, you probably have two pages.
Step 5: Check whether your site can credibly answer the query
Not every phrase is worth targeting. Before you commit a page to a term, ask:
- Do I have accurate facts to answer this? If not, you need a source or an expert before you write.
- Does my site already have a page close to this topic? If yes, updating that page may beat creating a new one.
- Would a reader who finds this page be a realistic customer, or just a curious visitor?
This step matters most for businesses in regulated or technical fields, where a confident but wrong answer can cost trust. Treat each claim the page will make as something to verify against a primary source.
Step 6: Prioritize with evidence from your own analytics
There is no universal difficulty score that replaces your own data. Published keyword metric tools estimate volume and competition, but those estimates can differ widely between providers, and none of them can tell you how your specific site will perform. A more reliable approach is to test in your own analytics.
Reader-run check: choose five phrases from your list. Publish or update one page for each. Record the Search Console impressions and average position for each page at a fixed interval, such as every two weeks, for at least a few months. Compare the phrases that gained impressions against those that did not. Note which intents produced clicks. This is a conditional test, not a guarantee: results will depend on your site's authority, competition, and how well each page matches the search intent.
A practical prioritization rule that follows from this is to favour phrases where you already see some impressions, because Google is already showing your site for them, and where the page you would write matches the intent you labelled in Step 3.
Common mistakes when collecting keywords
- Chasing volume alone. A high-volume phrase with mismatched intent can bring visitors who never convert.
- Ignoring existing pages. Duplicate pages compete with each other. Check your site before adding a new one.
- Skipping the fact check. A keyword list is only useful if you can answer the question accurately.
- Treating the list as permanent. Search behavior and your business both change. Revisit the list when you add a service or notice a change in Search Console.
- Collecting too many terms. A list of 300 unsorted phrases is harder to act on than 20 grouped and labelled ones.
A simple workflow to repeat each quarter
- Update your business facts and the list of pages you have.
- Expand seeds with search suggestions and your Search Console queries.
- Label each phrase by intent and drop the ones you cannot answer accurately.
- Group phrases into page-sized topics.
- Assign each topic to a new page or an existing one to update.
- Record impressions and positions for the pages you published so you can compare results.
Used this way, keyword research becomes a repeatable editorial habit rather than a one-time project. The output is a short, verified list of topics your site is positioned to answer.
Where automation fits
Manual keyword research is practical for a small list. As the list grows, the work of keeping facts current, drafting, checking claims, and publishing on a schedule can take more time than the research itself. Some teams use software for part of this process.
You do not need software to apply the steps above.
If you want to try this workflow with a tool that handles the research-to-publishing steps, you can Try Pentra on the free tier and compare its output against your own keyword list.