When two or more of your pages compete for the same search term, splitting their ranking signals.
Why two of your own pages end up fighting
The "cannibalization" label is vivid but a little misleading. Google does not refuse to rank a site twice for one query, and having several pages mention the same phrase is perfectly normal. The problem is narrower: it happens when two or more pages are each built to be the answer for the same query and the same underlying intent, so Google has to choose between them. Because neither page is the clear, single best match, the choice is often unstable and rarely lands on the page you would have picked.
The damage is mostly indirect. Links, internal anchors, and topical relevance that should all reinforce one strong URL get spread across two weaker ones. Google may swap which page it shows from week to week, your click-through suffers when the wrong or thinner page surfaces, and a page you have quietly retired can outrank the one you are actively promoting. None of this triggers a penalty — it is simply self-inflicted dilution, which is exactly why it is easy to miss and worth catching during keyword research rather than after publishing.
Genuine cannibalization is about overlapping intent, not overlapping words. "Running shoes for flat feet" and "best running shoes for flat feet 2026" look like distinct phrases, but a searcher wants the same thing from both, so two pages targeting them compete. By contrast, "what is keyword difficulty" (someone learning) and "keyword difficulty checker" (someone wanting a tool) share a word yet serve different goals, and separate pages for each is correct. Reading the intent behind a query — by looking at what actually ranks on the search engine results page — is how you tell the two situations apart.
A worked example, and the mistake to avoid
Imagine a blog that publishes "How to choose a standing desk" in January and, eight months later, a fresh "Best standing desks" buying guide. Both target buyers comparing desks; both chase the same commercial query. Google now alternates between them, each picks up a slice of the backlinks, and neither cracks the top five. Merge them into one authoritative guide — keep the stronger URL, fold in the best of the other, and 301-redirect the loser — and the consolidated page often climbs precisely because every signal now points at a single target.
The most common mistake is diagnosing cannibalization from keyword strings alone and then "fixing" pages that were never in conflict. Two URLs ranking for an identical phrase are only a problem when they answer the same intent; if one is a tutorial and one is a product page, they should coexist. Confirm a real clash before merging by checking, for each suspect query, whether the URL Google ranks keeps flip-flopping and whether both pages would leave the searcher equally satisfied. Only then is consolidation the answer rather than two deliberately differentiated pages.
For research, the practical takeaway is to plan intent before you plan content. Map your target terms to intent up front, group the ones that share a goal so a single page owns each group, and reserve separate pages for genuinely separate needs — the discipline covered in keyword clustering. Doing this at the planning stage, as part of how to do keyword research, prevents the overlap from being created in the first place, which is far cheaper than untangling redirects and lost rankings later.