Keyword Research
Finding the words your audience searches with, the intent behind them and the pages that can answer them.
Reviewed by Alexander Yarovenko · Updated: 2026-10-04
Keyword research is the work of finding out which words and phrases your audience actually types into a search box, what they want behind each phrase, and which of those phrases your site can answer with a useful page. Its output is a list of topics mapped to pages, not a list of numbers. Volume and “difficulty” figures come from tools and advertising data, not from Google’s ranking systems.
What it is and where the boundary lies
Google’s SEO starter guide asks you to expect your readers’ search terms: people who know a topic well and newcomers often use different words for the same thing, such as “charcuterie” and “cheese board”. Keyword research is the systematic version of that advice. It ends where writing begins: Google says that anticipating these differences could produce positive effects, which is not a promise of rankings, and nothing in the documentation rewards repeating a phrase.
How it differs from similar terms
| Concept | Question it answers | Typical data source |
|---|---|---|
| Keyword research | what does the audience search for and what do we cover | Search Console, Trends, Keyword Planner, SERP review |
| Search intent | what the person wants behind one query | reading the results page |
| Keyword clustering | which queries one page can serve together | comparing results for different queries |
| Content gap analysis | which topics are missing on your site | your page list against the researched topics |
| Position tracking | where pages appear for chosen queries | Search Console or a rank tracker |
Why it matters
A page can be well written and still answer a question nobody asks in those words. Research shows the vocabulary of the audience, the questions around a topic and the seasons when interest rises. It also protects against the opposite mistake: building pages for a phrase that is popular but that your site cannot serve better than the results already shown.
Where the data comes from
Each source measures something different, so none of them is “the truth”.
- Search Console, Performance report: the queries dimension groups your data by the query users typed. These are real queries that already showed your site, but only for pages you already have, and results depend on the time, place, device and history of the searcher.
- Google Trends: a largely unfiltered, anonymized sample of real searches. Each point is divided by total searches for its place and period and scaled 0 to 100, so it compares relative interest and does not give counts.
- Keyword Planner: a Google Ads tool that suggests keywords and estimates monthly searches and the cost of showing an ad. It needs an account with billing information, cannot show very rare or sensitive keywords, and describes advertising, not organic ranking.
- The results page: the only place to see what Google considers an answer to a query today: page types, formats, “People also ask” questions.
How to apply it
- Write the topics of the business in the customer’s words: products, problems, questions, comparisons.
- Pull real queries from Search Console and note phrases that differ from your page titles.
- Expand the list with suggestions, related searches and the “People also ask” box; keep a source column for each phrase.
- Open the results for each important phrase and record the dominant page type and intent.
- Group phrases that one page can answer and give every group one target page, which avoids cannibalization.
- Prioritise by value to the business and fit with your site, treating tool volume as a rough size of demand.
- Re-check after publication in Search Console and return to the list at least when the audience or the season changes.
Practical example
A shop selling boards for serving food first lists its topics, then compares real queries with the vocabulary on its pages:
| Audience phrase | Intent seen in results | Decision |
|---|---|---|
| cheese board | buying and ideas | category page plus a guide |
| charcuterie | recipes and how to assemble | one guide, shared with the category |
| how to clean a wooden board | care instructions | separate help article |
The result is three target pages with an explicit reason for each, not one page stuffed with all the phrases.
Common mistakes
- Treating tool volume as Google data: estimates differ between tools and are not ranking information.
- Choosing by “difficulty” alone: it is a score invented by a tool; open the results and judge them yourself.
- Ignoring intent: a phrase with a product grid in the results will not be served by an essay.
- Using only Keyword Planner: it is built for ads and hides rare and sensitive keywords.
- One page per phrase: near-identical pages compete with each other; group by intent.
- Researching once: vocabulary and seasons change; compare the list with Search Console again.
How to validate the result
Before publishing, check that every target page has one stated intent and no other page claims the same one. After publishing, open the Performance report for those pages and see which queries really appear, then compare them with the plan. Wait for a full reporting period before judging: Google notes that results are specific to time, place, device and history, so a single manual search proves little.
More questions
Do I need a paid tool?
Not to start. Search Console and Google Trends are free, and reading the results page costs nothing. Paid tools add speed and competitor data.
Does keyword density matter?
The documentation describes writing for readers with the words they use, not hitting a density. Repeating a phrase is not a recommended technique.
Is search volume the traffic I will get?
No. Tool volume is an estimate of searches, while your visits depend on position, appearance in the results and demand in your market.
Next practical step
Work through the lesson What a keyword actually is, then export the queries of one section of your site from Search Console and compare them with its page titles.
Related concepts
- Search intent — what stands behind a query.
- Keyword clustering — grouping queries by page.
- Long-tail keywords — specific, rarer phrases.
- Google Search Console — the source of your real queries.
- Keyword Planner — the advertising tool and its limits.