Keyword cannibalisation is when several pages on one site compete for the same query. The usual result is that none ranks as well as a single consolidated page would, and rankings fluctuate as the search engine alternates between them.
It is also over-diagnosed. Two pages ranking for the same term is not automatically a problem.
What it is and is not
It is a problem when two pages target the same intent, the search engine visibly swaps between them in results over time, both rank mediocrely, and clicks are split between them.
It is not a problem when one page is informational and another commercial for related terms, when a page incidentally ranks for a term it does not target, or when you legitimately occupy two positions for one query — which is a good outcome.
The deciding question is intent. “Project management software” and “what is project management” share words and serve entirely different needs. Two pages about choosing project management software do not.
Diagnosing it properly
Search console performance data answers this directly.
- Export queries and pages for the last six months.
- Group by query and count distinct URLs receiving impressions for each.
- Flag queries where two or more URLs have meaningful impressions.
- For each flagged query, plot position by URL over time. Genuine cannibalisation shows as alternation — one URL ranks, then the other, repeatedly.
Then confirm by reading both pages. If a reader searching that query would be well served by either, they are competing. If each serves a different need, they are not.
A quicker informal check is a site-restricted search for the term, which shows which of your pages the engine considers relevant. Useful as a first pass, not as evidence.
Fix 1: Merge
The right answer when two pages genuinely serve the same intent and neither is comprehensive.
Choose the stronger page as the survivor — more backlinks, better position history, better URL. Fold the unique value from the other into it, so the merged page is better than either was. Then redirect the removed URL to the survivor with a 301, and update every internal link to point at the survivor directly rather than through the redirect.
Merging is the most effective fix and the one teams avoid, because deleting content feels like loss. Three mediocre pages consolidated into one strong page almost always outperforms the three.
Fix 2: Differentiate
The right answer when both pages should exist but have drifted into the same territory.
Give each a distinct angle, retarget the titles and headings, and adjust the content so the difference is obvious within the first paragraph. Then make the internal linking reflect the distinction: link to each from contexts that match its angle.
This works well for a commercial page and an informational page that have blurred, which is the most common version of this situation.
Fix 3: Canonicalise
The right answer when pages must exist separately for user or technical reasons but should be treated as one for search — printer-friendly versions, near-identical location or variant pages, parameterised URLs.
Set a canonical tag on the secondary pages pointing at the primary. They remain accessible to users; ranking signals consolidate on the primary.
Note that canonical tags are a hint rather than a directive. If the pages differ substantially, the engine may ignore it, which usually means the real answer was merge or differentiate.
Fix 4: Remove or noindex
The right answer when a page has no value to anyone — an old post superseded entirely, a thin tag archive, a duplicate created by a migration.
Noindex if it still serves a purpose for users. Remove and redirect if it does not. Removing without a redirect is acceptable only for pages with no links and no traffic.
Where cannibalisation comes from
Four causes, all preventable.
No content inventory. Writers produce a piece that already exists because nobody checked. A simple spreadsheet of published pieces with their target queries prevents most of it.
Keyword-first planning. Treating each keyword variation as a separate page produces near-duplicates by design. Plan around questions and intent instead.
Site structure that generates pages. Tag archives, author archives, pagination and faceted navigation can create large numbers of near-identical URLs. This is a technical configuration problem — see the technical SEO audit checklist.
Migrations. Old and new URLs both live, or a replatform creating parallel structures.
Preventing it
Maintain a content map listing every page, its target query and its intent, and check it before commissioning anything. Group content into deliberate topic clusters with defined boundaries. And run the search console diagnosis quarterly rather than waiting for a ranking problem.
The underlying discipline is deciding, before writing, which page owns which question. Cannibalisation is almost always a planning failure that shows up as a ranking symptom.
What to expect after a fix
Merges show results fastest, often within four to eight weeks, and the merged page commonly ranks better than either original did. Differentiation takes longer because the engine has to re-evaluate both pages. Canonicalisation can take several weeks to be respected, and sometimes is not.
Record the before state — positions, impressions, clicks for the affected queries — so you can tell. A fix applied during a seasonal dip will look like a failure without a baseline.
The merge, step by step
Merging is the fix most likely to be done badly, and the damage is avoidable.
- Record the baseline. Positions, impressions, clicks and backlinks for every page involved.
- Pick the survivor on backlinks first, then ranking history.
- Write the merged page properly. Not a concatenation — a single coherent piece that covers everything the originals did, better. A merged page that reads like two articles stapled together performs like one bad article.
- Publish the merged page at the survivor’s URL. Changing the URL as well adds an unnecessary variable.
- 301 the removed URLs to the survivor, one to one.
- Update internal links to point at the survivor directly. Leaving them pointing at redirects wastes a little value and makes future audits noisier.
- Resubmit the sitemap without the removed URLs.
- Watch for eight weeks against the baseline.
Do not delete the old content from your records. If the merge underperforms, you want the ability to reconstruct what was lost.
Cannibalisation in paid and organic together
A related question comes up often: should you bid on terms you already rank first for organically?
This is not cannibalisation in the technical sense — paid and organic do not compete for the same slot — but the budget question is real. The evidence is mixed and situational: occupying both positions usually increases total clicks, and the incremental clicks are expensive when the organic result already captures most of them.
The way to settle it is a holdout test rather than an argument. Pause paid on those terms in one region or for a period, and measure total clicks and conversions rather than paid clicks alone. Teams that run this test usually find the answer differs by query type, with branded terms behaving very differently from generic ones.
Frequently asked questions
Is ranking twice for one query bad?
No, that is a good outcome. The problem is alternation and split signals, not the presence of two results.
Will merging lose the traffic from the removed page?
A properly redirected page passes most of its value. Expect a short dip while the change is processed, then recovery and usually improvement.
Can internal linking cause cannibalisation?
It can worsen it. Linking to two competing pages with the same anchor text reinforces the confusion. Consistent anchors pointing at one owner page help resolve it.
How do I know which page to keep?
Backlinks first, then historical ranking, then URL quality, then content depth. The page with external links is usually the survivor even if the other is better written, because links are harder to replace than prose.
