Keyword cannibalization

Telling real cannibalization from harmless overlap

Most pages that share vocabulary are not cannibalizing anything. A definition page and a step-by-step guide can each repeat the same phrase forty times and never compete, because the people searching them want different things. Cannibalization is about intent, not words, and it only counts when both pages are trying to be the answer to the same question. These are the signals that it is actually happening:

• Positions swap. The URL ranking for a query keeps changing between two of your pages from week to week. The engine is undecided, and the wobble costs you the stable position either page might have held.
• Impressions split in Search Console. One query shows meaningful impressions against two or more of your URLs, with none of them strong.
• The wrong page ranks. A blog post outranks the product or landing page you actually wanted people to reach, so the traffic arrives at the version that cannot convert it.
• Both rank, both mid-page. Two of your URLs sit somewhere in positions six to fifteen for the same query, where one consolidated page could plausibly reach the top five.

The opposite case is worth naming too, because it causes needless rewrites: a content cluster is supposed to have many pages on one topic. Pages that share a subject but answer different questions are the intended design, not a defect. Check search intent before consolidating anything — if the two pages would satisfy different people, leave them alone.

Why it happens, and what it costs

Almost all cannibalization is accumulated rather than designed. A site publishes for a few years and someone writes a fresh take on a topic already covered; a landing page is built for a campaign and never retired; an idea gets a beginner version and later an advanced one that ends up covering the same ground. None of those decisions is wrong on the day it is made, which is exactly why the problem is only ever visible from above, looking at the whole site at once.

The cost is concentration. Ranking signals are earned per URL: links point at a specific page, engagement accrues to a specific page, and a topic's authority is built by one page being the obvious destination for it. Two pages splitting that inheritance each end up with half of it. Internal linking splits along with everything else — anchor text about a topic now points at two places, so neither becomes the site's answer for it.

It gets worst on templated sites. Programmatic SEO generates pages from a data set, so if two rows in that data set describe the same intent, you have generated your own competition automatically and at scale — and you will not notice by reading pages one at a time, because each page looks fine. First-party note from building AgentCeres — the AI Growth Officer at agentceres.com: we publish a small number of new pages every day, and the guard that keeps them from competing is applied before anything is written. Two pages shipped on the same day have to answer clearly different questions, and each proposed page is checked against the list of what already exists. It has cost us pages: a planned question about getting cited by ChatGPT and Perplexity was dropped because an existing page already answered it, and publishing the second would have produced two mid-ranking pages instead of one good one. The cheapest fix for cannibalization is the one made in the outline, before the page exists.

The fix, in order of preference

Three moves, least destructive first. Consolidation is irreversible in practice, so it should not be the first thing you reach for:

• Differentiate. Keep both pages and give one of them a genuinely different job — retitle it, rewrite the opening around the question it is better placed to answer, cut the overlapping middle, and link the two together so a reader and a crawler both see the distinction. This is the right move when both pages have real value to different people.
• Consolidate. Merge whatever is unique in the weaker page into the stronger one, then redirect the weaker URL to it with a 301. This is the right move when the second page exists mainly because someone forgot about the first. You lose a URL and gain a page stronger than either was alone.
• Demote. Keep the page for the people who need it but take it out of the competition, with a canonical tag pointing at the primary version — or noindex if it genuinely should not be in search at all. Use this for pages that must exist, such as a duplicate campaign landing page, but should never be the search result.