Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

SEO › On-Page SEO

Finding and Fixing Keyword Cannibalisation

What keyword cannibalisation is, how to detect competing pages in Search Console, and when to consolidate, differentiate or remove each one.
Guideintermediate8 min readUpdated

What the term means

"Keyword cannibalisation" is practitioner vocabulary; Google has no feature by that name and documents no penalty. What it describes is real and follows from how ranking works: for a query, the engine evaluates candidate pages and generally shows one per site. If a site offers three candidates with similar relevance, the choice is unstable, and the signals that make a page strong — internal links, external links, engagement, freshness — are divided among them. The symptoms are rankings that fluctuate between URLs, a position that is worse than the site's authority should earn, and a click-through rate (CTR) that lags because the listing's title and snippet keep changing.

Two things that are not cannibalisation: pages that share keywords but serve different intents (a category page and a buying guide; a tool and the article that explains the metric), and exact duplicates, which are a canonicalisation problem with a technical fix. The search intent guide explains why format and intent, not words, decide whether two pages compete.

How it happens

  • Growth without a map. Articles written over years by different people on overlapping topics, each reasonable alone.
  • Content types colliding. A blog post, a service page and a FAQ entry all answering the same question.
  • Variants treated as topics. Pages for "cheap", "best", "affordable" and "top" versions of one query that the results page treats as one intent.
  • Location and product multiplication. Near-identical pages per city or per colour variant with no distinct content.
  • Refreshes published as new pages. The updated guide and the old one both live.
  • Category and tag pages on content platforms generating thin listing pages for every term.

Generated archives are the one cause that a template rule can settle at once, and the index rules for tags, archives and utility pages decide which become hubs, which are merged and which stay out of the index.

Detecting it

Query-level data is the evidence; everything else is suspicion.

  1. Search Console (GSC): the performance report filtered to a query lists every page that received impressions for it. Several pages with meaningful impressions for one query, or one query whose top page changes between periods, is the signature. Export page-and-query data for the site's important queries and group by query.
  2. Rank tracking over time. A tracked term whose ranking URL alternates is cannibalisation in motion.
  3. Site crawl. Pages with the same or near-identical titles, H1s or target phrases are candidates; confirm with query data, because similar titles do not always mean the same intent.
  4. A site search in the engine (site:example.com query) shows which pages the engine considers relevant and in what order; it is a quick check, not a diagnosis.
  5. Internal anchors. The same anchor text pointing at different pages is both a cause and a clue (the internal linking guide covers anchors).

Rank the findings by the value of the query: a swap between two pages for a head term matters; two old posts sharing a long-tail query may not be worth the work.

When a single page holds the query steadily but it is not the page the map assigns, the pages are not competing at all, and the page on Google ranking the wrong page for a keyword traces that pattern to internal anchors, titles, canonicals and where links from other sites land.

Deciding what to do

For each competing group, the decision is one of three, made by asking whether the pages have distinct reasons to exist.

SituationDecisionWhat it involves
The pages say the same thing for the same intentConsolidateMerge the best content into one URL; redirect the others to it; update internal links and the sitemap
The pages could serve different intents or audiences but currently blur togetherDifferentiateRewrite each around its own intent (format, angle, level, audience); retitle; link them to each other with anchors that state the difference; update internal links so each query's anchors point to its owner
One page is thin, outdated or exists only for a variantRemoveRedirect to the page that should own the query if there is one; otherwise return 410 and remove links
Both pages legitimately rank together (brand, navigational)LeaveMonitor; make sure the preferred page is the stronger one through links

Consolidation is the most common right answer and the most resisted, because each page has an owner, a history or a date. The test is the reader: would one page serve them better than a choice between two? If yes, consolidate.

Executing a consolidation

  1. Choose the surviving URL: usually the one with the most links and the best current ranking, unless its URL is poor and a migration is already planned.
  2. Merge content: the strongest sections from each, restructured as one page around the intent, not a concatenation. The headings guide covers the outline.
  3. Redirect the retired URLs to the survivor with permanent redirects (Google documents redirects as the strongest consolidation signal). Where a retired page must remain reachable for some reason, a canonical tag (Canonical Tag) pointing to the survivor is the weaker alternative.
  4. Update every internal link that pointed to the retired pages, including navigation, related blocks and contextual links, so nothing relies on the redirect.
  5. Update the XML sitemap and any hub pages.
  6. Record the change, with the date and the queries involved, so the effect can be read in Search Console over the following weeks.

Differentiation follows the same discipline with different content: each page gets its own intent reading, its own title, and internal anchors that reflect the split.

Illustrative example — not a real result

Three pages, one query

A software company's Search Console shows the query for its core feature's name splitting impressions across a feature page, a blog post explaining the feature, and a help-centre article — with the ranking URL changing month to month and none above the middle of the page. The decision: the feature page owns the commercial intent ("what is it, why choose it") and keeps the query; the help article is differentiated to the how-to intent ("set it up"), retitled accordingly and linked from the feature page; the blog post, which repeated the feature page, is consolidated into it and redirected. Internal anchors for the feature name now point to the feature page only. Within a few recrawls the query has one stable URL at a better position, and the help article begins ranking for the set-up queries it was previously losing to the blog post.

Preventing it

Cannibalisation is prevented at the planning stage, by the keyword mapping tutorial: every query cluster is assigned to exactly one URL, and every proposed new page is checked against the map before it is written. Three habits keep the map honest:

  • Refreshes update the existing page, under the existing URL, rather than creating a new one.
  • Variants are reviewed against the results page: if the engine shows the same results for two phrasings, they are one cluster with one page.
  • New content types are mapped too. A help centre, a blog and a product catalogue are one site to the engine; their queries go on one map. The keyword map template records each cluster's single owner URL, its current state and the consolidate, split or create decision in one row.

This site's content pipeline enforces unique titles and descriptions across every page and warns when two pages in a cluster share an intent and similar titles, which is an automated first line of defence.

Cannibalisation in stores and multi-location sites

Two site types produce it structurally. Stores generate it through filters and variants: a filtered category view and a subcategory page for the same refinement, or colour variants each with a page when the query is the product. The faceted navigation guide decides which views are pages and canonicalises the rest, and the category and product guide settles variants. Multi-location businesses generate it through location pages that differ only by town name, which the engine treats as one page's worth of content with a dozen URLs; the fix is genuinely local content per page, or fewer pages, as the local section describes. In both cases the detection method is the same, and the fix is usually a template rule rather than a page-by-page decision.

Measuring the outcome

After a fix, the signals to watch in Search Console are: one stable URL per query, position improving toward what comparable pages on the site earn, click-through recovering as the listing stops changing, and total clicks for the query group rising even where the retired pages' clicks disappear. The SEO KPIs guide covers reading these over time; the recrawl lag means weeks, not days.

Overlapping pages often surface in Search Console as excluded URLs; the page on crawled pages Google declined to index shows how to tell a near-copy from a thin page.

Common mistakes

  • Treating shared words as cannibalisation. Different intents are allowed to share vocabulary.
  • Deleting instead of redirecting. The retired page's links and history are thrown away.
  • Concatenating pages. A merged page that is two pages stapled together serves neither intent.
  • Leaving internal links on the redirect. Signals flow through a detour indefinitely.
  • Fixing one group and not the cause. Without a map, the next year recreates the problem.
  • Keeping both pages because of ownership. The reader does not care who wrote which.

The technical SEO audit tutorial pairs with this guide as the two reviews to run on any existing site; the SEO audit expertise page shows the author's audit approach.

Frequently asked questions

Is cannibalisation a Google penalty?

No. Google does not document any penalty for having several pages on a topic. The term describes a self-inflicted dilution: when several pages are plausible answers to a query, the engine picks one per search and may switch between them, and signals such as links and engagement are split across them. The effect is weaker rankings, not a sanction.

Can two pages rank for the same query on purpose?

Occasionally both appear, which can be a good outcome for brand or navigational queries. For most informational and commercial queries the engine shows one page per site, so two candidates means one is wasted. The question is whether both pages have a distinct reason to exist.

Is it cannibalisation if two pages share a keyword but not an intent?

No. A product page and a how-to guide can share words without competing, because the results page wants different formats for the two queries. Cannibalisation is about the same intent, not the same words.

Terms used on this page

Tools to try

Frameworks

Where this page is used

Part of these learning paths

Sources

  1. Consolidate duplicate URLs — Google Search Central

Related