Keyword cannibalization: how to diagnose it and fix it
It is over-diagnosed. How to prove you actually have it using free data, and which of four fixes the evidence points to.
Entrovix AIAug 24, 2026 12 min read
Two of your pages mention the same phrase. That is not cannibalization. It is a website.
Cannibalization is a specific and much narrower condition: Google is visibly torn between two of your pages for the same query, and neither gets the benefit that one would. It is real, it is worth fixing, and it is diagnosed far more often than it occurs, usually by tools that flag any keyword overlap as a problem.
The short answer
You have cannibalization when the URL Google shows for a query keeps changing between two of your pages over weeks. That instability is the symptom that matters. Overlapping words in two titles is not, on its own, evidence of anything.
Proving it, with data you already have
Three checks, in order. The first is the one that settles it.
1. Search Console, filtered by query
Open the Performance report, filter by the exact query, then switch the view from Queries to Pages. You now see every URL of yours collecting impressions for that term. If two or more are picking up meaningful impressions, Google is genuinely choosing between them. If one page has 98 percent of them, there is nothing wrong.
2. The instability check
Still filtered by that query, look at the ranking URL over a longer window. A single page holding the position throughout is healthy, even if others get a trickle of impressions. A URL that swaps back and forth week to week is Google failing to settle, and that is the condition worth acting on.
3. The site: search, for the candidate list
Search site:yourdomain.com "the exact phrase" to see which pages are even in contention. Useful for finding pages you had forgotten exist, particularly old blog posts and tag or category archives, which are a frequent and invisible source of this.
The four fixes, and how to choose
Once it is confirmed, the decision is which of four actions the evidence supports. Choosing the wrong one is how a real problem becomes two problems.
| Fix | Choose it when | What you do |
|---|---|---|
| Consolidate | Both pages answer the same question and neither is complete | Merge the best of both into one page, redirect the loser to it |
| Differentiate | Both deserve to exist and serve genuinely different questions | Re-title and re-angle each so they stop competing, and fix the internal anchors |
| Redirect | One is obsolete and has nothing worth keeping | 301 it to the survivor. Do not delete it |
| De-optimise | You need both, and one is clearly secondary | Pull the contested term out of the secondary page's title, H1 and anchors, and point its links at the primary |
Consolidate: the default when in doubt
One strong page beats two partial ones reliably. Take the best material from both, publish it at the URL with the better history, and 301 the other one. The redirect matters: deleting the loser throws away whatever links and authority it had accumulated.
Differentiate: when both are legitimate
Common with a service page and a guide about the same subject. They should exist, and they should stop looking identical to a crawler. Re-title each around the question it actually answers, make the opening paragraph of each unambiguous about which reader it is for, and check that internal links to each use different anchor text.
The free test for whether two terms genuinely justify two pages is in how many keywords one page should target: search both and compare the first pages. If the results are mostly the same URLs, Google thinks it is one question.
De-optimise: the underused option
Often the right answer and rarely the one people reach for, because it feels like giving something up. You keep both pages, and you simply stop one of them from arguing. Remove the contested phrase from its title and H1, keep it in the body where it reads naturally, and change internal links pointing at it to describe what it actually covers.
Internal links are the lever people forget
Whichever page you choose to win should receive the internal links, with anchor text describing the destination. This is the clearest signal you control about which of your pages is the authoritative one, and on most sites it is contradicting the decision that has just been made everywhere else.
Check the older pages especially. A three-year-old post linking to the wrong page with the exact contested phrase as its anchor text is actively voting against you.
Things not to do
- Do not noindex the loser as a first move. It removes the page entirely rather than resolving the conflict, and you lose whatever else it was ranking for.
- Do not delete content without a redirect. That converts a ranking problem into a 404 and discards any links the page had.
- Do not canonicalise one page to the other unless they are genuinely near-duplicates. A canonical between two pages with different content is a signal Google may simply ignore.
- Do not act on a tool's cannibalization report without the Search Console check. Most of those reports flag keyword overlap, which is not the same condition.
- Do not restructure the whole site because two pages compete. Fix the two pages.
The instability is the diagnosis. If one page has held the position for months, two of your titles sharing a word is not a problem you have.
Where it comes from in the first place
Knowing the sources is how you stop recreating it.
- One thin page per keyword variation, which is the failure mode that keyword lists encourage.
- Tag and category archives left indexable, generating pages that compete with the posts they list.
- A location page per city that differs only by the city name.
- Old blog posts covering ground a newer service page now covers better.
- Filtered or parameter URLs producing near-identical crawlable pages.
The last two are as much a technical matter as a content one. Finding and fixing technical SEO issues covers parameter handling and duplicate URLs at the site level.
Questions people actually ask
How long until the fix shows?
Once both pages are recrawled, usually a few weeks for the signal to settle. A consolidation with a redirect tends to resolve faster than a de-optimisation, because the ambiguity is removed outright rather than reduced.
Is it cannibalization if both pages rank on page one?
Almost certainly not. Two results from one domain on the first page is a good outcome. Leave it alone.
Can my homepage cannibalise a service page?
Yes, and it is common. Homepages accumulate links and often out-rank the specific page that deserves the query. The usual fix is de-optimisation: stop the homepage from leading with that phrase, and make internal links to the service page use it instead.
Do tools that report cannibalization work?
They report keyword overlap, which is a candidate list rather than a diagnosis. Treat the output as pages worth checking in Search Console, not as a list of things to merge.
