Keyword Cannibalization: When Your Own Pages Compete for One Search

What Keyword Cannibalization Is

Keyword cannibalization is an SEO term, not a Google one, for a site whose pages compete with each other for the same search. Two articles answer the same question, or a blog post and a service page both chase the same commercial query, and search engines are left to choose between pages that each have part of what one strong page would have.

The overlap of keywords is not the problem in itself. Large sites rank several pages for related queries all the time, and two results from the same site can be a good outcome when each serves a different need: a guide that explains a treatment, and the page that books it. Cannibalization is the case where the pages serve the same intent, so internal links, external links and engagement are split between them, and neither page is the clear answer.

That makes it an intent problem first and a technical problem second. The test that decides every case below is the one from reading search intent from the results page: if two pages would be built for the same results page, one of them should not exist in its current form.

The Symptoms

• The URL that ranks for a query changes from week to week, and neither version holds a position.
• A page ranks for a query it was never written for, such as a blog post appearing for a service query that the service page should own.
• Rankings for a topic plateau mid-page after a new article on the same subject is published.
• Two pages have similar titles and headings, and the team cannot say which one is the main page for the topic.

None of these proves cannibalization on its own. Rankings move for many reasons, and a page ranking for an unexpected query is often a sign of breadth, not conflict. The symptoms are a reason to look at the data.

Confirm It in Search Console

Search Console’s Performance report shows which of a site’s pages Google showed for a query, which is the evidence a diagnosis needs. Work through it query by query for the topics that matter to the business.

1. Open the Performance report, set the date range to the last three months, and switch on impressions, clicks and average position.
2. On the Queries tab, click the query you are investigating. The report filters to that query.
3. Switch to the Pages tab. Every page listed earned impressions for the query in the date range, with its own clicks and position.
4. To see whether the pages alternate, filter to one page at a time and look at the Dates tab or the chart: alternation shows as one page’s impressions dropping while the other’s rise.

Read the positions the way they are calculated

Search Console defines position as “the topmost position occupied by a link to your property or page in search results”. When a site appears at positions 2, 4 and 6 for one query, the site-level figure records 2. The report also counts differently by tab: “Data grouped by Queries, Countries, Devices, or Dates is aggregated by property. Data grouped by Pages or Search appearance is aggregated by page.” The consequence for diagnosis is that the site-wide view can look stable while the pages underneath swap places, so the comparison has to be made on the Pages tab, one page at a time.

Why It Happens on Client Sites

Cannibalization is rarely one bad decision. It accumulates from ordinary publishing:

• Content calendars built from keyword lists. Variants such as teeth whitening cost and how much is teeth whitening become separate posts, though one results page answers both.
• Blog posts that drift commercial. A post written to explain a service gains pricing and a booking call to action over the years, until it competes with the service page it was meant to support.
• Redesigns that keep the old pages. A new service page launches beside the old one instead of replacing it, often with both still linked from somewhere. The redirect map in the website redesign SEO process exists to prevent exactly this.
• Archive and variant pages. Tag and category archives, filtered product listings and city-name copies of one service page can all produce near-duplicates of the page that should rank.
• Several writers, no shared record. Without a list of which page owns which topic, each new brief starts from scratch.

Four Fixes, and When Each Applies

Start by choosing the page that should win. It is usually the one with the stronger history: more external links, more conversions, the better-established URL, and the format that matches the results page. Then apply the fix that matches the case: merge when the pages do the same job, refocus when both deserve to exist, and point internal links at the winner either way.

1. Merge and redirect

When two pages serve the same intent, combine them. Move the useful material from the weaker page into the stronger one, so the merged page is better than either was, then redirect the weaker URL permanently with a 301. Google’s documentation says that with a permanent redirect, “the indexing pipeline uses the redirect as a signal that the redirect target should be canonical”. Update internal links to point at the final URL rather than through the redirect, and remove the old URL from the sitemap.

2. Refocus one page on a different intent

When both pages deserve to exist, make them different. The usual split is a guide that answers the informational question and a service page that handles the decision. Rewrite the title, the main heading and the opening of the page that is moving, so each page states its own purpose, and remove the duplicated sections from one of them.

3. Point internal links at the page that should win

Internal links are the clearest statement a site makes about which page is its answer. Links using the query’s wording should point at the chosen page, and links from the supporting page should send readers onward to it. Google’s link guidance asks for descriptive anchor text, which in practice means the anchor names the topic of the page it leads to, not “click here”.

4. Use a canonical only for genuine duplicates

The canonical link is meant for “duplicate or very similar pages”: print versions, parameter URLs, the same product in two categories. Google treats it as “a strong signal that the specified URL should become canonical”, not a command, and it decides what counts as a duplicate based on similarity of content. Pointing the canonical of one substantially different article at another is likely to be ignored, and it leaves both pages live to keep competing. If two pages differ enough to need separate content, they need fix 2, not a canonical.

What Not to Do

• Delete a page without a redirect. Its links and any traffic it still earns are lost, and visitors arriving from old links land on an error.
• Noindex the weaker page and leave it linked. A noindex removes a page from results, but it does not tell Google which page should take its place, so whatever the page has earned stays with a URL that no longer ranks. Reserve noindex for pages users need but searchers do not, such as internal search results.
• Fix every overlap. Two pages ranking for a query is only a problem when they serve the same intent and neither is winning. A guide and a service page that both appear on page one are an asset.
• Judge the fix after a week. Merged pages and redirects take time to be recrawled and reprocessed. Compare the chosen page’s impressions and position over a full month against the same period before the change.

Prevent It With a Query-to-Page Map

The durable fix is a record of which page owns which query. It can be a spreadsheet: one row per primary query, with the intent, the owning URL, the page type and the supporting pages that link to it. Before any new brief is approved, the target query is checked against the map. If a page already owns it, the work becomes an update to that page, not a new one.

Review the map quarterly against Search Console. Filter the Performance report to the site’s most valuable queries and confirm that the page receiving impressions is the page the map says should. A mismatch is either a new conflict or a sign that the map is out of date, and both are worth an hour.