Keyword cannibalisation is a decision you already made
Two of your pages are competing for one query. That is not a bug that crept in. It is the visible result of two briefs written months apart by people who could not see each other’s work.
What cannibalisation actually is
The standard definition is that two or more pages compete for the same query, so neither ranks as well as one page would. True, and not very useful, because it describes the symptom and says nothing about how you got there.
The useful framing is that cannibalisation is a planning failure that becomes visible in search results months later. Nobody sets out to publish two pages about the same thing. It happens because the second brief was written without knowing the first page existed, or because two keywords looked different in a spreadsheet and turned out to mean the same thing to a search engine.
That matters because it changes the fix. If you treat it as a technical fault you go looking for a technical remedy: canonicals, redirects, noindex. If you treat it as a planning failure you fix the brief process, and the problem stops recurring.
This is the clearest example of why an SEO audit produces a list that regrows. Merge the duplicates and you have fixed today’s instances. Leave the briefing process alone and the same defect reappears with the next quarter’s content.
Two pages, one intent
Two pages, related intent
One page, many keywords
How to tell whether you have it
The diagnosis people usually reach for is a site search: look for pages targeting similar keywords, eyeball the list, merge the ones that look alike. It finds the obvious cases and misses the expensive ones, because similarity of wording is a poor proxy for similarity of intent.
Two better signals, both from data you already have.
- URL flipping in Search Console. Filter to a single query and look at the pages tab over ninety days. If the ranking URL changes more than once, the engine is not confident which page answers the query. That uncertainty costs you position, and it is the clearest signal there is.
- Impressions without clicks, split across two URLs. One query, two pages, each picking up impressions at position eleven to eighteen. Combined they would often clear position eight. This is the case that costs real traffic and is invisible if you only look at your best-performing pages.
The tell that you do NOT have a problem: two pages ranking for the same query at positions three and seven, both getting clicks, with different titles that promise different things. That is two useful answers, and search results have room for both.
| Pattern | Ranking URL | Verdict |
|---|---|---|
| One query, URL changes weekly | Unstable | Cannibalisation. Merge or differentiate. |
| Two URLs, both stuck at 11-18 | Split | Cannibalisation. Usually merge. |
| Two URLs at 3 and 7, both clicked | Stable | Healthy. Leave alone. |
| One URL, forty query variations | Stable | Working as intended. |
Merging is not the only fix, and often not the right one
The reflex answer to cannibalisation is to merge: pick the stronger page, fold the weaker one into it, redirect. Sometimes correct. Frequently it destroys something worth keeping.
There are four options, and the right one depends on why the pages overlap rather than on how much they overlap.
The consolidation case is the one worth being careful about, because it is irreversible in practice. A redirect can be undone; the rankings the redirected page held often cannot be rebuilt.
Catching it before it exists
Everything above is remediation. The version that actually compounds is catching the collision at planning time, when it costs one line in a brief rather than a merge, a redirect and a recovery period.
Two checks, both cheap.
- Cluster the keyword list before assigning pages. Two keywords with the same intent and most of their words in common are one page. The plural pair is the classic miss: “crm software” and “crm softwares” look like two rows in a spreadsheet and are one query to a search engine.
- Write the target query into the brief, and check it against every published page first. If the answer is already covered, the brief becomes an update rather than a new URL. This is the single change that stops the problem recurring.
The reason this rarely happens is not that people disagree with it. It is that nothing in the normal content workflow forces the check, so it depends on somebody remembering. Put it in the brief template and it stops depending on memory.
Where this sits in a wider plan is a sequence of decisions rather than a one-off cleanup: cluster the keywords, assign pages, write the link rules into the brief, then publish. Doing it in that order is what stops the problem existing.
The version that scales: template families
Everything so far assumes a handful of hand-written pages. The version that does real damage is structural, and it comes from generating pages from a template.
A location page for every city, a comparison page for every competitor pair, an integration page for every tool. Each is defensible on its own. Together they produce hundreds of pages whose differences are a noun and a number, competing with each other for queries that differ by the same noun.
The failure is not that the pages are generated. It is that the keyword list they were generated from was never checked for collisions first, so the template faithfully reproduces one collision several hundred times.
| Scale | How it arises | What it costs | Fix |
|---|---|---|---|
| 2 pages | Two briefs, months apart | One query, unstable ranking | Merge or differentiate |
| 20 pages | A cluster planned without clustering | A pillar that cannot win its own topic | Re-plan the cluster, merge the overlaps |
| 200+ pages | A template run over an unchecked keyword list | Crawl waste, thin-content risk, self-competition at scale | Fix the source list, then regenerate |
The reason the largest case is the cheapest to prevent and the most expensive to fix: at 200 pages you are not merging, you are regenerating, and every URL you drop was earning something you have to account for.
There is a test that catches this before generation. Take the list the template will run over, and ask whether any two rows would produce pages a reader could not tell apart from their titles alone. If the answer is yes, the list is the problem, not the template.
What it actually costs you
The case for fixing cannibalisation is usually made with a diagram and a shrug. It is worth being more specific, because the cost is not evenly distributed and some instances genuinely are not worth touching.
The damage concentrates in three places, and only the first is about position.
- Position, through instability. An engine switching between two URLs never lets either accumulate the click and engagement history that a single page would. This is the compounding cost: not that you rank at eleven instead of eight today, but that neither page is building the record that would move it.
- Crawl efficiency, at scale. On a forty-page site this is irrelevant. On a site with programmatic families it is not, because near-duplicate pages consume the same crawl allowance as useful ones and give nothing back.
- Your own link equity, split. Every internal link pointing at the weaker page is a vote you cast for the page you did not want to win. This is the part people miss, because it looks like an external problem and is entirely self-inflicted.
The honest version of the third point: I have watched a page sit at position fifteen for a year while forty internal links pointed at a thinner page covering the same ground. Nothing external was wrong. The site was voting against itself.
If you do merge, the mechanics matter
Deciding to consolidate is the easy half. The execution is where the traffic is usually lost, and it is lost in predictable ways.
Redirecting before the content moves
Keeping only the stronger page’s content
Leaving the internal links pointing at the old URL
Merging pages with different intent
One sequencing rule covers most of it: content, then internal links, then the redirect. Doing those in any other order is how a merge that should have gained you six positions costs you a quarter of recovery instead.
Give it a full crawl cycle before judging the result. The first two weeks after a consolidation look worse than the steady state, reliably enough that judging early is how people undo a merge that was working.
Finding the pairs worth merging in the first place is a job for a content audit, which is where the selection happens before any of this applies.
Running the check on your own site
Both halves of this are things I built tools for, because I got tired of doing them in spreadsheets.
Building a topical map from your keyword export is the planning-time check: it flags pairs that share an intent and most of their tokens before anything is written, including the plural case that a manual scan reliably misses.
The Internal Link Mapper handles the published-site half. Two pages competing for one query usually also compete for internal links, and the tool separates editorial links from navigation ones so you can see which page your own site is actually voting for. A page with forty inbound links, all from the footer, is not the page you are backing.
Neither replaces looking at Search Console. They narrow where you look.
Cannibalisation is a planning failure that shows up in search results months later. Diagnose it from URL instability in Search Console, not from similar-looking titles, and consider differentiating before you consider merging: consolidation is the only option you cannot undo.
The merges I regret are the ones I did quickly because the data looked obvious. Two pages at position fourteen, similar titles, merge them. Six months later the combined page sits at eleven and the second page’s long-tail traffic, which nobody had looked at, is gone.
Now I check what the weaker page actually earns before touching it. A page that ranks badly for its target and well for thirty things nobody planned is doing more work than the report suggests.
Frequently asked
How do I know if I have keyword cannibalization?
Is keyword cannibalization actually bad?
How do I fix keyword cannibalization?
Does keyword cannibalization affect rankings?
I take on a small number of engagements at a time. If you are past product-market fit and want to talk through what this would look like for your situation, the calendar is open.
Book a call