How keyword cannibalization happens
Cannibalization is what happens when two or more pages on the same site go after the same search intent. A blog post and a service page both try to answer a query about Google Ads management. Several location pages describe the same city in slightly different words. An old article was never retired when its replacement went live.
The search engine then has to pick one. It may pick the page you did not intend, it may alternate between them from week to week, or it may rank both weakly instead of one of them strongly. Meanwhile the links pointing at the topic, internal and external, are divided between the pages instead of concentrated on one, and so is the relevance the site has earned.
It is worth being precise about what this is not. Two pages mentioning the same phrase is not cannibalization. Two pages competing to be the answer to the same query is. A definition and a service page that share a name serve different intents and can sit side by side quite happily.
Why keyword cannibalization matters
The direct cost is a weaker position than the site has earned. The indirect costs are worse: the wrong page ranks and converts badly, reporting turns confusing because the ranking URL keeps changing, and link building becomes wasteful because nobody is certain which page to point at.
On bigger sites it multiplies. Location and industry pages built from a template are the usual source, because a template quietly invites near-identical copy.
Common mistakes with keyword cannibalization
The first is diagnosing it from titles. Two pages with similar titles may serve different intents, and two pages with quite different titles may compete directly. The test is the query: open Search Console, filter to it, and see whether more than one page collects impressions for it.
The second is reaching for a canonical tag as the default fix. Canonicalizing one page to another tells the engine to ignore it, which is right for duplicates and wrong for pages that answer different questions and both deserve to exist. The third is deleting the loser: if the older page holds links and history, merging its content into the winner and redirecting is almost always better than removing it.
How to act on it
Confirm the conflict before touching anything. Where one query returns impressions for several URLs over time, you have a real candidate; where one URL clearly dominates, you do not. Then decide which page should own the intent, normally the one with stronger links and the better fit for what the searcher actually wants.
Merge the useful parts of the others into it, redirect the retired URLs, and repoint the internal links so their anchors describe the surviving page. Then fix the cause rather than the symptom: record the winning page against that intent in your keyword map so nobody recreates the same conflict later. Recurring cannibalization on a larger site is usually structural, and it is one of the standard findings in an SEO audit.