Keyword map: how to build, clean up, and assign it to pages

What a keyword map is and why a site needs one

A keyword map answers three practical questions. What do people search for in your subject area? Which page on your site should answer each group of queries? Which pages are missing, and which ones are redundant or duplicate each other?

Without a keyword map, pages often appear on a whim: for an idea, for a single query, for a request from the sales team. The result is several similar pages on one topic, while noticeable demand is left without a matching page.

A finished keyword map is the basis for several decisions:

• structure — which sections and levels of nesting the site needs;
• page types — where you need an article, where a catalog or product card, where a page describing an offer;
• content — which pages to write from scratch and which to expand;
• internal linking — which pages are related in meaning and should link to each other;
• priorities — which groups of queries to start with.

Keyword map, cluster, landing page — terms

• Search query — the phrase a user types into search.
• Search volume — how many times a query was entered over a period, according to a search engine or statistics tool. In Yandex Wordstat this is “Number of queries” — the total count of queries containing the given phrase.
• Intent — what a person wants to get from a query: to learn something, to choose, to buy, to find a specific site.
• Cluster — a group of queries with one intent that a single page can answer.
• Landing (target) page — the URL on the site assigned to a cluster.
• Cannibalization — a situation where two or more pages on a site compete for the same intent and get in each other’s way.

Step 1. Scope: which queries actually relate to the business

Collection doesn’t start with a statistics tool — it starts with deciding what falls within the site’s subject area. Before you collect anything, write down:

• which products, areas, or topics you actually offer;
• which regions you work in;
• who you serve: individual customers, companies, specialists;
• what you don’t do — and don’t plan to;
• whether you need informational topics for a blog, and which ones.

Scope is a working filter. Every query you collect is later checked against one question: “Can a page on our site honestly answer this?”

Why a keyword map grows out of control without scope

Statistics tools and autocomplete suggestions return everything that contains the word you searched for. The word “repair” shows up in queries about apartments, phones, and shoes alike. Without criteria, any similar-looking query seems useful.

The consequences show up later. Topics you have no offer for end up in the map. Pages get proposed for them that won’t bring in customers. Important groups get lost among thousands of rows, and setting priorities becomes impossible.

Seed queries: from scope to a first list

Seed queries (basic, “marker” queries) are short phrases people use to name your products, areas, and topics. They’re entry points for expansion, not a finished keyword map.

Where to find them:

• the site’s structure and product range — names of sections and categories;
• wording customers use in inquiries, correspondence, calls, and support questions;
• queries the site already shows up for in search (covered in step 2);
• synonyms and professional terms used in your industry.

Write seed queries in the customer’s own words, not internal names. An internal product name may never show up in search at all.

Step 2. Query sources

At this step, every seed query gets expanded: longer phrasings, synonyms, questions, and adjacent topics are found. It’s more reliable to pull queries from several sources, because each one shows its own slice of demand.

Your own site’s data is a source on its own. The Performance report in Google Search Console shows which search queries your site appears for. In Yandex Webmaster (Yandex’s counterpart of Google Search Console), the “Search query group management” tool shows which queries your pages appear for and which pages show up for those queries. This data comes in handy again at step 5, when you go looking for cannibalization.

Wordstat and search volume: what it means and what it doesn’t

Wordstat is a tool that shows statistics on search queries sent to Yandex. It lets you see how a query’s popularity has changed over time and build top-query lists built around the words you need.

Two tabs are useful for building a keyword map:

• “Top queries” — data for the past month on popular queries that contain the given phrase, and on queries similar to it;
• “Dynamics” — how the number of queries changed month by month; it reveals seasonality.

Results can be filtered by region and device type, and downloaded as CSV.

The key thing to know about the number: “Number of queries” is the total count of queries that contain the given phrase. The volume for “garden furniture” includes both “buy garden furniture” and “rattan garden furniture.” That’s why you can’t add up the volumes of nested phrases — the same queries would get counted more than once.

Operators help refine the volume. They can be combined with each other:

Operator	What it does	Example
" "	Fixes the exact word count: matches queries that contain the phrase with no extra words	"garden furniture"
!	Fixes the word’s exact grammatical form (number, case, tense) — a Russian-morphology detail	garden !furniture
[ ]	Fixes word order; all word forms and stop words are still counted	[furniture for the garden]
+	Fixes stop words — function words, pronouns, and other words with no extra meaning of their own	furniture +for the garden
( ) and |	Group words for complex queries	(garden|patio) furniture
-	Excludes words from the query	garden furniture -rattan

What volume doesn’t mean. It’s not a traffic forecast: the number of queries isn’t the number of visits to a specific page. High volume doesn’t make a query right for your site, and low volume doesn’t make it junk. The decision about a query depends on scope and intent; volume just helps set priorities.

Wordstat shows demand on Yandex. For Google, similar data comes from Keyword Planner in Google Ads: it helps you find new keywords and shows the number of searches per keyword per month.

Queries that reveal competitors

Search competitors are sites that appear in the results for your queries. They don’t always match your business competitors: the results can include aggregators, directories, and media outlets.

Their queries are useful in two ways:

• as a source of phrasings — some SEO analytics tools show which queries a competitor’s pages are visible for, surfacing phrasings your seed queries missed;
• as a hint about structure — which of a competitor’s pages rank shows how they split demand between sections and page types.

You can’t copy someone else’s keyword map wholesale. It carries the competitor’s brand, their product range, their regions, and their mistakes. Every query you find this way goes through your own scope filter, just like any other.

Suggestions, similar queries, questions

Yandex’s search suggestions are variants of the most popular queries that start with the same letters as your query. They’re generated automatically from user query statistics. Google’s suggestions are built from real user queries, taking into account language, location, trending topics, and search history.

This has a practical consequence: suggestions depend on the person viewing them. Yandex’s suggestions can be personalized, and Google factors in search history. Collect them signed out and with the right region set, and check the volume for anything you find separately.

Two more sources:

• similar queries in Wordstat — on the “Top queries” tab;
• questions — from suggestions, from headings and subheadings of pages that rank, from customer inquiries. Questions often turn into informational clusters for a blog.

Step 3. Cleanup: irrelevant, other brands’ branded, and junk queries

After expansion, the list inevitably has a lot of excess. Cleanup means checking every query against the scope from step 1.

What to remove:

• Irrelevant queries. A different meaning of the word, a product or area you don’t have, a different region, a different audience.
• Other brands’ branded queries. Names of competing companies and their products. These queries are usually navigational: the person is looking for a specific company, not yours. The exception is brands you actually sell or work with.
• Junk. Meaningless word combinations, queries with an obviously unrelated task — such as “download,” “essay,” “job vacancies” — when they have nothing to do with your business.
• Duplicates. The same phrase with a different word order, in different word forms, with a typo, or transliterated. Keep one main wording and treat the rest as its variants.

Don’t delete borderline queries right away. Move them to a separate “pending decision” list — for example, adjacent topics that might be useful for the blog later. This keeps the keyword map clean without losing ideas.

Keyword map cleanup checklist:

1. Every query has been checked against the business’s scope: products, areas, region, audience.
2. Queries where the keyword has a different meaning have been removed.
3. Other brands have been removed, except those you work with.
4. Queries with an unrelated task have been removed: downloads, job vacancies, school assignments, and similar.
5. Duplicates have been consolidated into one main wording.
6. Volume hasn’t been summed across nested phrases.
7. Borderline queries have been moved to a separate list instead of being mixed into the keyword map.

Step 4. Grouping

Grouping (clustering) combines the queries that one page can answer. Here it’s enough to understand the principle and check the result by hand.

By search results and by meaning

There are two main approaches:

• By meaning. Queries are combined if they look like synonyms or the same topic. The method is fast but subjective: similar words don’t guarantee the same user task.
• By search results. A common approach: queries land in the same group if a set number of URLs overlap in the top results for them. The logic is simple: if a search engine shows the same pages for two queries, one page can probably answer both.

Grouping tools save time, but their output is a hypothesis. Check borderline groups by hand: open the results for several queries in the group and see whether the page types match. If similar words show different pages in the results, those are most likely different clusters. Choosing the parameters and checking the groups is covered in a separate article on how query clustering by search results works.

Group intent: informational, commercial, navigational, mixed

Intent is determined for the whole group, not for a single query. It decides what type of page is needed:

• informational — the person wants to learn or figure something out. An article, guide, or reference piece fits;
• commercial — the person is choosing what to buy or who to approach. A catalog, product card, or a page describing an offer fits;
• navigational — the person is looking for a specific site or section, such as a brand or a contact page. The brand’s own page answers this query;
• mixed — the results contain both articles and commercial pages. The decision is based on which page type dominates the top results and which is closer to your site.

The most reliable way to determine intent is to look at the results. If a group’s top results are articles, a commercial page is unlikely to satisfy it, and vice versa. How to read the results and which page type to choose for each type of intent is covered in a separate article on intent types and reading search results.

Step 5. Assigning pages

This is the main step: this is where the list of queries turns into decisions. For each group, fill in a row of the map:

Group	Intent	Target page	Decision	Comment
Main wording and variants	info / commercial / nav / mixed	existing URL or “new”	create / update / merge / don’t create	why, and what to do

Existing page or new: the “create / update / merge / don’t create” decisions

First compare the group with what already exists on the site. A new page is the last option, not the first.

Situation	Decision	Sign
A suitable page exists but doesn’t fully answer the group: subtopics are missing, information is outdated	Update	The existing page’s type matches what’s in the results, but its content covers only part of the group
No page answers the group’s intent, the topic is within the business’s scope, demand is confirmed	Create	The results for the group show dedicated pages on exactly this topic, and your site doesn’t have it even as a section of another page
Two or more pages answer the same intent and split the queries between them	Merge	Different pages on your site show up for the group’s queries; their content largely overlaps
The group is outside scope, the intent isn’t yours, or it’s just a variant wording of an already-assigned group	Don’t create	The query isn’t about your business — it’s excluded; or another page already covers it as a subtopic — then the query is added to that page’s cluster

Merging affects existing pages: content gets moved over, and the redundant URL redirects to the main one. This decision is made separately and carefully, not automatically as a byproduct of grouping.

How to avoid building a page for a single query

The mere existence of a query isn’t a reason for a new page on its own. If a query differs from an already-assigned group only in word order, word form, or a minor qualifier, a separate page will almost certainly duplicate an existing one.

Search engine documentation has provisions that cover pages like this:

• Google’s spam policies describe doorway pages — pages created specifically to rank for particular search queries. The examples given include pages that are nearly identical, lack a clear hierarchy, and read more like search results than real pages.
• Google’s guidance on helpful content suggests rethinking your approach if you’re producing a lot of material on different topics in the hope that some of it will rank.
• Yandex may leave a page out of search results if it has little chance of being useful to users — for example, if it duplicates pages the crawler already knows about.

The test question: will the new page have content that no other page on the site has, and will it answer a distinct user task? If not, add the query to the cluster of the existing page. A separate article covers how to use queries in page copy.

Cannibalization: two pages for one intent

Cannibalization happens when several pages on a site compete for the same group of queries. The typical cause is pages created without a shared map: an article and a catalog section on the same topic, two similar articles, a page built around a synonym.

How to spot it:

• in the query-to-page map, one group is assigned to two URLs, or to none at all;
• Yandex Webmaster data shows different pages of the site appearing for the same query;
• two pages share the same topic, headings, and most of their content.

How to prevent it: the “one intent, one URL” rule. Before creating a page, check whether an existing page already covers the group. If an overlap turns up, the fix is to update one page, merge the pages, or separate their intents.

How we make decisions about pages is described on the page about how we work with keyword research and site structure. The next step after the map is deciding how to connect pages with internal links so the links mirror the logic of the keyword map.

How a keyword map turns into a site structure and content plan

A query-to-page map is already a draft structure. Groups that share a theme become sections, and narrower ones become pages within them. This way, the site’s hierarchy reflects how people search, not just how the company is organized.

The same map produces a content plan. Every row with a “create” or “update” decision is a separate task. Its group of queries, intent, and page type are already known — that’s the basis for a brief.

Several criteria determine the order of tasks:

• how important the group is to the business — whether it leads to your core offer;
• how much demand the group has according to the statistics;
• how closely the page type matches what the search results expect;
• how much effort it takes — improving an existing page or building a new one.

Volume here is one criterion, not the only one. A small but precise commercial group can matter more than a large informational one that isn’t tied to your offer.

How to keep a keyword map up to date

A keyword map goes stale along with the site and demand. There’s no universal update schedule, but there are events after which the map is worth revisiting:

• new products, areas, or regions have appeared;
• the site’s structure is changing: a new section appears, or the site moves to a new domain;
• Webmaster and Search Console data shows queries that aren’t in the map;
• the results for an important group have changed — for example, commercial pages replaced articles in the top results;
• demand is seasonal — its changes show up on the “Dynamics” tab in Wordstat.

When updating, don’t start from scratch. Add the new queries, run them through the same scope, cleanup, and grouping steps, and check that the new group doesn’t break the “one intent, one URL” rule. Record decisions you’ve already made along with the reasons behind them, so that months later it’s still clear why a group is assigned to that particular page.

Common mistakes

• A keyword map as just a list of keywords. A spreadsheet without assigned pages and decisions doesn’t help either structure or content.
• Collecting without scope and copying someone else’s keyword map. Topics, brands, and regions that have nothing to do with your business end up in the map, and pages get built for them.
• Treating volume as a ready-made answer. The volumes of nested phrases get added together, and the number of queries gets mistaken for a traffic forecast.
• Blindly trusting automatic grouping. A tool’s output is a hypothesis; borderline groups should be checked against the search results.
• A page for every query. New pages get created for variant wordings without checking existing pages first — that’s how cannibalization starts.
• Treating it as a one-off project. The keyword map gets built once and never updated after the site or demand changes.

Checklist: the keyword map is ready to use

1. The business’s scope is written down: what’s in, what’s out, regions, audience.
2. Seed queries are phrased in customers’ own words.
3. Queries are collected from several sources, including your own site’s data.
4. The map has been cleaned up using the cleanup checklist; borderline queries are in a separate list.
5. Groups have been checked against the search results, and each has a defined intent.
6. Every group has a target page or a “don’t create” decision.
7. Each group has a recorded decision — create, update, merge, or don’t create — with a reason.
8. No group is assigned to two pages.
9. “Create” and “update” rows have been turned into prioritized tasks.
10. It’s clear when, and after which events, the map gets revisited.

Frequently asked questions

How is a keyword map different from a list of keywords?

A list of keywords is raw material. A keyword map appears once every group of queries has an intent, a target page, and a decision about what to do with it.

How do you build a keyword map for a site?

Define the business’s scope, collect queries from several sources, clean them up, and group them. Then assign the groups to pages — the result of that work is the query-to-page map.

How many queries should a keyword map have?

There’s no universal number. The size of the map is set by the business’s scope and the number of distinct intents, not by a goal of racking up rows. A map with a small number of precise groups is more useful than a huge list with no decisions attached.

Do you need a separate page for every query?

No. One page answers a group of queries that share an intent. A separate page is needed only when a group has its own distinct user task and the results for it show dedicated pages. Variant wordings get added to the existing cluster.