Definition
Cannibalisation (keyword cannibalization) is the situation in which two or more pages on the same website compete for the same search query, undermining each other's ranking, clicks and authority. Google has to choose which page to show, often alternates between them, and gives neither the position a single strong page could have reached.
Cannibalisation almost never happens on purpose. A blog writes 'what is AI SEO' in 2023 and 'AI SEO explained' in 2025; a webshop has a category page and a blog article about the same product; a service page and an ad landing page cover the same intent; or a site has separate pages for 'SEO Amsterdam' and 'SEO agency Amsterdam'. To the writer these are different pages; to Google it is the same question.
The symptoms are recognisable in Search Console. Filter on a keyword and look at which pages are shown for it: if you see two or more URLs alternating, with a position that jumps up and down and a low CTR, that is cannibalisation. A rank tracker that records the ranking URL per keyword shows the same thing as a changing URL over time.
Not every overlap is a problem. Two pages that both rank for a broad keyword but serve different intents (an informational article and a product page) can coexist. It becomes a problem when the intent is identical, the pages alternate in position and neither reaches the top three. The test is: would a visitor with this query be better served by one page?
You fix it with a decision per case. Merge: fold the weaker page into the stronger one and add a 301 redirect. Differentiate: rewrite both pages so they genuinely cover a different intent, with a different title, H1 and focus. Canonicalise: if both pages must exist for a good reason, point one to the other. Deindex: set the redundant page to noindex. Traze checks every new page for overlap with existing clusters and proposes a merge instead of writing a new page.