Some of the most frustrating SEO problems are not caused by competitors. They are caused by your own pages. Keyword cannibalisation is when two or more URLs on your site target the same query, splitting the ranking signal that should have concentrated on one, so both underperform. On large sites it is rampant and rarely diagnosed, and I have watched it quietly hold down pages that were doing everything else right.

What Cannibalisation Actually Does

Search engines want to show one best result from your site for a given query. When you give them two pages that both look like the answer, they hedge: they split link signals, relevance, and click data across the two, rank whichever looks marginally better at any moment, and swap them in and out. Neither page accumulates the authority it needs to break through, so both sit stranded, usually together on page two, trading places while a competitor with a single strong page owns the top.

On a large marketplace I worked on across four markets, this pattern was everywhere: a blog post and a category page both chasing the same term, or two near-identical category URLs generated by filters, each suppressing the other. The fix was rarely new content. It was deciding which URL should own the term and making every signal point at it.

How to Diagnose It

How to Fix It

Once you have found it, the fix is a decision followed by consolidation. Pick the URL that should own the term, usually the one with the most authority, the best intent match, or the most commercial value, and then make everything point at it.

Done across a site, this is some of the highest-return technical work available, because it takes pages you already have and lets them finally rank. No new content, no new links, just the removal of the self-inflicted competition that was capping them.

Why Large Sites Cannibalise Themselves

Small sites rarely have a serious cannibalisation problem, because they do not have enough pages to compete with themselves. Large sites cannibalise constantly, and usually not on purpose. Faceted navigation spins up multiple near-identical category URLs for the same set of products. A blog post written for traffic ends up targeting the exact term a category page needs. Two teams, content and merchandising, each create a page for the same topic without knowing the other exists. At scale, self-competition is the default state, not the exception, which is why diagnosing it has to be a routine audit rather than a one-off.

This is also why cannibalisation and index bloat travel together. The same filter and parameter URLs that bloat the index are frequently the ones competing with your clean category pages for the same terms. Fixing one often fixes the other: canonicalising or pruning the duplicate URLs both trims the index and ends the cannibalisation, concentrating the signal on the single page you actually want ranked. On a large site, treating them as one problem rather than two is usually the faster path to both fixes.

Step by Step: Finding Cannibalization in Search Console

You do not need paid tools to catch most cannibalization. Search Console shows it plainly once you know the pattern to look for. Open the Performance report, click into a query that matters to you, and switch to the Pages tab. One healthy query shows one dominant page earning nearly all the impressions. A cannibalised query shows two or more of your pages splitting them, and if you check the same query across different date ranges, you will often see the ranking URL flip between them. That flip is the signature: Google keeps changing its mind about which of your pages to show, which means neither is winning.

Run this check on your ten most important queries first, then on any query where your position has been stuck or sliding. Note every query where two or more URLs share the impressions. That list is your cannibalization worklist, ordered by how much the query matters to the business. The deeper walkthrough, with the exact clicks and filters, is in the dedicated guide to identifying cannibalization in Search Console, and if you prefer working in a rank tracker or crawler, there is a separate method using SEO tools to find cannibalization at scale.

The Decision Framework: Merge, Redirect, Differentiate, or Canonical

Every cannibalization case resolves through one of four moves, and choosing the right one is most of the skill. Picking the wrong move is how people fix a split and lose traffic doing it, so it is worth being deliberate.

A special case worth naming: when a blog post starts outranking or suppressing the commercial page for a buying query, that is content cannibalization, and it has its own playbook covered in the guide to blog versus money page cannibalization. The short version is that you keep both but re-point the blog post at the informational intent and channel its authority to the money page through internal links.

A Real Pattern: What Fixing It Actually Looks Like

On a large marketplace I worked on, one commercial head term was being contested by three URLs at once: the category page, a filtered variant of the same category, and an old blog post written before the category existed. Rankings for the term had been stuck between positions twelve and twenty for months, with the ranking URL rotating between the three. The fix took an afternoon: canonical on the filtered variant to the category, the blog post retargeted to a genuinely informational angle with a prominent link down to the category, and internal anchors across the site unified so they all said the same thing and pointed to the same place.

Within weeks the category page became the only URL Google showed for the term, and it climbed to a stable position it had never reached while the signal was split three ways. Nothing new was written. No links were built. The site simply stopped competing with itself, and the authority that was already there started pulling in one direction. That is what makes cannibalization such a satisfying problem: the fix is usually subtraction and alignment, not new investment.

Preventing It: The Habit That Keeps You Clean

The Six Kinds of Cannibalisation, and Why the Difference Matters

Most guides treat cannibalisation as one problem with one fix. In practice there are six distinct varieties, they look nearly identical in a report, and applying the wrong remedy to the wrong type is how people lose traffic while trying to fix it. Before touching anything, identify which of these you are actually looking at.

When It Is Not Cannibalisation

This section barely exists in most guides and it prevents more damage than any other part of this article. Not every query showing two of your URLs is a problem, and the reflex to consolidate on sight causes real losses.

A query can legitimately surface two of your pages when the query itself carries more than one intent. Someone searching a broad product term might want to browse a category or read a comparison, and Google may reasonably show both. If your two pages serve those two intents cleanly, and one of them holds a stable majority of impressions, you have coverage rather than conflict. Merging them would remove an answer rather than strengthen one.

Branded queries are another frequent false alarm. Your homepage, your about page and a service page all appearing for your company name is normal and healthy. Nobody is being harmed and no signal is being wasted, because you already own that result set entirely.

The third case is temporary testing. Google routinely trials a different URL for a query for a short period, especially after you publish or update something. A single week showing an unexpected page is not a pattern. This is precisely why the diagnosis above insists on checking across date ranges: persistence over time is what separates a real split from normal churn.

The Decision Tree, In Four Questions

Once you have a confirmed case, the choice between merging, redirecting, differentiating and canonicalising can be reduced to four questions asked in order. Working through them in this sequence prevents the two most common mistakes: merging pages that should have stayed separate, and spending hours salvaging content from a page that had nothing worth saving.

The first question is about intent, because if the intents genuinely differ, nothing else applies and the answer is to sharpen the targeting rather than consolidate. The second is about value, and it saves the most time: if the weaker page has no traffic, no links and nothing unique, redirect it and move on rather than conducting an archaeology project on its content. The third handles the e-commerce and utility case where both URLs must remain reachable for users even though only one should rank. And if none of those apply, you have the standard case: merge the best of both into the stronger URL and redirect the weaker one.

Which Cases to Fix First

A site-wide audit typically surfaces more cases than anyone has time to work through, and treating them as an undifferentiated list is how the project stalls in month two. Prioritise on three factors together rather than fixing whatever appears at the top of the export.

The first is commercial value: does the contested query sit in front of buyers, or is it an informational term that would never have converted anyway. The second is proximity to the threshold: a split holding two pages at positions eleven and fourteen is worth far more than one holding two pages at sixty and seventy, because consolidating the first can produce a page-one result almost immediately while the second still leaves you invisible. The third is effort: pure duplicates and obvious redirect cases are quick, while genuine merges involving content decisions take real time.

Work the intersection of high commercial value and near-threshold position first, because that is where consolidation converts directly into traffic. On most sites that is a surprisingly short list, often ten to twenty cases, and clearing it delivers the majority of the available gain.

What Happens After You Fix It

Expectations matter here, because the pattern after consolidation is not a smooth climb and people panic at the wrong moment. In the first week or two after a merge and redirect, rankings frequently wobble and can dip below where either page sat before. This is normal: Google is re-evaluating a changed URL, processing the redirect, and reassessing the consolidated page. Reverting at this point, which people do, converts a temporary dip into a permanent mess.

From roughly weeks two to six the consolidated page typically settles and begins climbing past where either individual page ranked, because it now carries the combined relevance signals and internal links of both. On queries where the split was severe, the improvement can be dramatic. On queries where the second page was barely present, the change is modest, which is exactly why prioritisation matters.

Two things speed this up. Updating internal links to point at the surviving URL rather than relying solely on redirects, because a redirect chain is weaker than a direct link and your own linking is the clearest signal you control. And requesting indexing for the consolidated page so the change is processed sooner rather than whenever the crawler next arrives.

Cannibalisation in the AI Answer Era

One newer consideration worth naming, because it changes the stakes slightly. AI answer engines and generative results select sources to cite, and they favour clarity: a single authoritative page that comprehensively answers a question is easier to quote confidently than two partial pages that each cover half of it and contradict each other in emphasis.

That means cannibalisation now costs you twice. It splits your ranking signal in traditional search, and it makes you a weaker candidate for citation in generated answers, where being the single clear source matters more than appearing twice. Consolidation, which was always good practice, has become slightly more valuable than it was, and the sites carrying dozens of half-answers on the same topic are quietly disadvantaged in a channel they may not be measuring at all.