Keyword-to-Page Mapping
A keyword map is a table matching a task to a single URL. Without one, the decision is made not by your team but by indexing: when pages look alike, Google clusters them and picks the one it judges most complete and useful. This lesson covers that documented mechanism and how to build a map so the choice stays yours.
It rests on the canonicalization documentation and on the ways to specify a canonical URL.
What happens without a map
The documentation states the mechanism directly: if Google finds multiple pages that seem to be the same, or whose primary content is very similar, it clusters them together. Then Google chooses the page that, based on the signals collected, is objectively the most complete and useful for search users, and marks it as the canonical.
That has a measurable consequence for crawling: the canonical page is crawled most regularly, while duplicates are crawled less frequently. Similar pages are not "penalised" — they simply get less attention, and the page shown in the results is not necessarily the one you promoted.
Your signals are a preference, not a command
The ways to express a preference are listed in order of influence: a redirect is a strong signal that the redirect target should become canonical; a rel="canonical" annotation is a strong signal; inclusion in a sitemap is a weak signal. The same page notes that these methods stack and become more effective combined, yet none of them is required.
And the canonicalization overview notes that Google's choice can differ from the site's preferences. So a keyword map is not a declaration in the markup but a decision about content: pages have to differ in substance, not only by a tag.
How the map gets built
- Tasks, not queries. Group queries by task, as in the lesson on search intent.
- One target URL per task. The primary query and the secondary ones live on the same row.
- Check what exists. Before creating a page, look at the queries the site already appears for: the performance report and the pages report. Updating an existing page is often cheaper and safer than adding one.
- Difference in substance. For every pair of neighbouring rows, state how the pages differ for a reader. If you cannot state it, it is one row.
- Relationships. The map also sets internal links: where neighbouring pages lead and with what anchors. That is part of the architecture, and a link only works when it is an a element with an href attribute.
An example map
| Page | Task | Primary query | How it differs |
|---|---|---|---|
| /laptops/ | Choose and buy from the range | buy a laptop | A catalogue with filters and prices |
| /laptops/gaming/ | Choose for gaming | gaming laptop | A selection by hardware requirements |
| /blog/how-to-choose-a-laptop/ | Understand the criteria | how to choose a laptop | Explains the specs, sells nothing |
The "how it differs" column is the important one. If it cannot be filled in, the rows merge.
Finding conflicts in your own data
- One page per query. Watch which page earns impressions for the target query. If it changes from day to day, Google is choosing among your pages: that is what the cannibalization check shows.
- Unintended pages. If a query surfaces a page you did not plan, the map has drifted from reality.
- Excluded URLs. Indexing states and the canonicals Google selected are in URL auditing.
- Empty slots. Tasks with no page are found through coverage gaps, and the completeness of one page through page analysis.
- Relationships. Who links to the target URL and with what anchors is internal link analysis.
Dealing with duplicates you already have
- Pick one target URL — the one with more impressions and the more useful content.
- Bring the best of the others into it, rather than gluing texts together.
- Set up redirects or rel="canonical" — both are strong signals, and they can be combined.
- Update internal links to the target URL; links to the old addresses do not need to stay.
- Allow time for crawling: surplus addresses are crawled less often, so changes are not instant, and crawling is budgeted per host.
Mistakes
- A page for every phrasing of a query.
- A difference that exists only in the canonical tag, not in the content.
- Creating a new page without looking at what already appears.
- Believing rel="canonical" guarantees the choice: it is a strong signal, but the outcome can differ from the site's preference.
- A map with no "how it differs" column.
What comes next
The map is the plan seen page by page; how it enters the plan and gets checked against outcomes is in the lesson on the keyword plan. The next lesson is about linking the rows of the map into clusters.