Keyword cannibalization: how to find and fix it

What keyword cannibalization is, with an example, is in the glossary; this guide is about proving that two of your pages compete, and fixing it.

Not every overlap is a problem. Google can show more than one page of a site for a search: its site diversity system generally shows no more than two listings from the same site in the top results, and may show more where it judges that especially relevant^[1]. Two of your pages on the first page for one search, each serving a different reader, is a strong position, not a fault. Where pages are very similar, Google’s deduplication systems show only the most relevant results^[1], so the copy may simply not appear, and the one that does is not always the one you would choose.

Search Console’s Performance report holds everything needed to find competing pages. The work is to read it by query and by page together.

1. 1Start from the searches that matter: the ones that bring buyers or enquiries, and the ones your important pages were written for.
2. 2Filter the report by one query, then open the Pages tab to see every URL that earned impressions for it. To group the variants of a search, use a custom regular expression filter on the query, which matches several queries at once and is not case-sensitive by default^[2].
3. 3Set a long date range and compare periods. One week of two URLs trading places is noise; three months of it is a pattern.
4. 4For each URL, note impressions, clicks, average position and the dates it was shown. Grouped by page, each URL in a result is counted separately, even when several point to the same site^[3], which is what lets you see the split.
5. 5Repeat the other way round: filter by one page and open the Queries tab, to see which searches it is chasing that another page was written for.

Two limits matter. Most performance data is assigned to a page’s canonical URL, the version Google chose to show, not to a duplicate, and a click on a duplicate counts toward the canonical^[4]: a copy that Google already treats as a duplicate will not show up as a competitor here. And filtering by query or URL can change the report’s totals, because of data truncation and the omission of anonymized queries^[2], so judge the split by its shape over time rather than by one exact count.

For a site with many pages, export the query and page data and build one table: one row per query and URL, with impressions, clicks and position. Sort it by query, and every search with two or more URLs earning a meaningful share of its impressions is a candidate.

The numbers say that two pages compete; only the pages say whether they should. Open both, side by side, and read them as the searcher would.

• Do they answer the same question for the same reader? If one is for someone learning and the other for someone ready to buy, they serve a different search intent, and the fix is to make each clearly about its own.
• Which one is the better answer today, on its own merits, not on its history?
• Which one has the stronger record: links from other sites, internal links, conversions, and the longer history of ranking?
• Which address would you rather show in results: the shorter, clearer one, the one in the main navigation, the one that sells?
• Is either page worth keeping for a different search it already earns? A page that loses one query may still own three others.

Write the decision down before acting: which page owns the search, what happens to the other, and which figures you will judge the result by. The figures to judge it by are the two pages’ combined impressions, clicks and position for the search before the change, which become the baseline.

Each competing pair gets one of five decisions. The choice follows from the reading of the pages, not from which fix is quickest.

The sections below take the first four in turn.

Merging is the strongest fix, and the right one when two pages answer the same need and one would serve the reader better than both.

1. 1Choose the page to keep by its record: links, conversions, history and the better address.
2. 2Move every section of the other page that readers came for into the page you keep, rewritten to fit. A redirect moves the address, not the content.
3. 3Redirect the old URL to the page you keep with a permanent server-side redirect. Google treats 301 and 308 as permanent, uses the redirect as a signal that the target should be canonical, and shows the target in results^[5].
4. 4Update the internal links that pointed at the old URL so they point at the kept page directly, and remove the old URL from the sitemap.
5. 5Record the date of the merge and the pair’s combined figures before it, and read the kept page against them over the following weeks.

Google’s guidance on duplicate URLs lists the methods in order of how strongly they influence the choice of canonical: redirects, then rel="canonical" annotations, then inclusion in a sitemap, and notes that the methods stack, so that using two or more raises the chance of your preferred URL appearing in results^[6]. A merge that redirects, links and lists the same URL sends one clear signal.

Often neither page should go: each is a good page aimed at the wrong search. A blog post that explains a treatment and a service page that sells it can both rank for the treatment’s name, and both do worse for it. Re-targeting gives each its own job.

• Decide which page answers which need: the guide informs, the service or product page is where the buyer acts.
• Rewrite the title and main heading of each around its own need, and remove the sections that belong to the other.
• Give the informing page the questions around the search, and a clear link to the page that sells.
• Give the selling page what a buyer needs: prices, availability, proof, the next step.
• Where a page ranks for searches it was never written for, write the page those searches deserve, and link it from the one that ranked.

A canonical tells search engines which of several duplicate or very similar pages you want shown^[6]. In our view it is not a way to choose between two different pages that happen to share a topic. Use it where both versions must stay live, such as a product reachable at two addresses or a printable copy of a page, and point the copy at the original. Google advises against naming different canonicals for the same page through different methods, and against using noindex to choose a canonical within a site, because noindex blocks the page from Search entirely^[6].

Internal links are the quiet fix, and they matter in every other fix too. Google recommends linking to the canonical URL rather than a duplicate, because consistent links help it understand your preference^[6]. It asks for link text that is descriptive, reasonably concise and relevant to the page it points to, and says that every page you care about should have a link from at least one other page on your site^[7].

1. 1List the pages that link to either of the competing pages, with the text of each link.
2. 2Where a link’s text names the search, point it at the page that owns the search.
3. 3Add links from related pages to the owner, with text that says what the owner is about.
4. 4Link the other page to the owner once, in context, so that a reader who lands on the wrong page finds the right one.

Cannibalization returns whenever a site publishes without checking what it already has. Three habits keep it away.

• Keep a map of searches to pages: every important search, the one page that owns it, and its intent. Check the map before a new page is briefed, and update it when one is published.
• Prefer improving the page that owns a search to writing a new one beside it. A new page needs a need of its own.
• Check the query and page table each quarter, alongside the review for pages that are slowly losing ground.

Never answer one search with a set of near-identical pages. Google’s spam policies name doorway abuse, pages created to rank for specific, similar search queries, and give as an example substantially similar pages that are closer to search results than a clearly defined, browseable hierarchy^[8]. What starts as cannibalization can become a policy problem at scale.

A page that loses ground while no other page of yours gains on its searches has a different problem, set out in Content decay: how to find and reverse it. Where both kinds of fix belong in the month’s plan is in Monthly SEO report: what to include.