How to Find and Fix Keyword Cannibalization

Keyword cannibalization is when two or more pages on your site answer the same search intent so closely that Google keeps switching between them, or settles on the weaker one. This guide shows you how to tell real cannibalization from harmless overlap, how to find it with Search Console query-and-page data, and how to choose the right fix for each case: merge, differentiate, canonicalize or leave alone. Once you have decided to merge, our consolidation guide covers the execution in detail.

What keyword cannibalization is, and what it is not

The textbook definition is “two URLs on the same domain ranking for the same keyword.” That definition flags far too much. The useful one is narrower: your pages are cannibalizing each other when one searcher would be equally well served by either page, and the competition between them is costing you clicks.

That makes it an intent problem that shows up as keyword overlap. A product page for “CRM software” and an article called “What is a CRM?” can share dozens of queries without competing, because they answer different questions. Two blog posts with different target keywords can still collide if both answer “how do I reduce churn?” for the same reader.

Two facts from Google keep the problem in proportion:

• It is not a penalty. Google’s SEO Starter Guide says duplicate content is not a violation of its spam policies and will not cause a manual action. The cost is inefficiency: split signals and the wrong page in front of searchers.
• Google already limits same-site results. Its ranking systems guide describes a site diversity system that generally shows no more than two listings from one site in the top results, and deduplication systems that show only the most relevant of very similar pages. Two of your pages ranking together is not, by itself, a problem.

Real cannibalization or harmless overlap?

Use this table before you touch anything. Most “cannibalization” reports from rank trackers land in the right-hand column.

What it actually costs

The wrong page ranks

This is the most expensive version and the easiest to miss. A blog post ranks for a commercial query your product page should own, so the visitors who were ready to act land on an article. Traffic looks fine; conversions do not.

Signals split across URLs

Links, internal links and engagement that could point at one page are spread across several. Google lists “to consolidate signals (such as links) for similar or duplicate pages” as one of the reasons to choose a canonical URL. Google also states that PageRank is still part of its core ranking systems, so where your links point matters.

Rankings that flicker

When Google rotates two URLs for one query, the title and snippet searchers see keep changing, and neither page builds a stable position. You will see it as two URLs taking turns in Search Console, which the worked example below shows.

Reporting that hides the problem

When a topic’s traffic is spread over four URLs, the section total looks healthy while no single page is strong enough for the head term. Every one of those pages also needs its own updates, accuracy checks and link maintenance.

Crawl budget, mostly only for big sites

Google’s crawl budget guide is written for sites with roughly a million or more pages, or 10,000 or more pages that change daily. If your site is smaller, duplicate URLs are rarely a crawling problem. If it is that large, Google’s advice is to eliminate duplicate content so crawling focuses on unique pages.

Where cannibalization comes from

• Publishing without a map.Five years of blogging produces “How to improve page speed” (2021), “Core Web Vitals guide” (2023) and a case study covering the same ground. Nobody checked what already existed.
• Redesigns and migrations. A new hub page is built from old articles, and the old articles are never retired or redirected.
• Templates and parameters. Faceted navigation, sort orders, tag archives and tracking parameters create many URLs for one piece of content. A Site architecture review shows where your URL patterns create these duplicates.
• Production at volume.Writing twenty articles a day, with or without AI, multiplies overlap unless each brief starts from what already exists. At the extreme it stops being a quality problem and becomes a policy one: Google’s spam policies name doorway abuse, including “substantially similar pages that are closer to search results than a clearly defined, browseable hierarchy,” and scaled content abuse.
• Teams publishing in parallel.Marketing, product docs and the blog each write “how to set up single sign-on” for good internal reasons, and no one owns the shared keyword map.

How to find it

Search Console: queries with more than one URL

Search Console is the best free source, because it is Google’s own record of which of your URLs it showed for each query. One correction to advice you will see elsewhere: exporting the Performance report from the interface gives you separate Queries and Pages tables, not query-and-page pairs, and the export is capped at 1,000 rows. You need the pairs.

1. For a few queries you already suspect: open Performance, add a filter for the exact query, then open the Pages tab. Every URL listed earned impressions for that query, with its own clicks and average position.
2. For the whole site: pull the query and page dimensions together through the Search Analytics API, the Looker Studio connector or the bulk data export to BigQuery. Use the last three months, which is long enough to catch rotation and short enough to reflect the current site.
3. Group by query and count URLs. Keep queries where two or more URLs earned impressions. Remove branded queries; they almost always show several of your pages, legitimately.
4. Filter out the noise.A rule of thumb that works as a starting point: keep a query only if it has at least 100 impressions in the period and the second URL has at least 20% of that query’s impressions. Below that, the second URL is an occasional appearance, not a competitor. Adjust both numbers to your site’s size.
5. Look at the weekly pattern. For each remaining query, break impressions down by week and URL. URLs that take turns are the strongest evidence. Two URLs that both appear every week at stable positions are usually Google showing two useful pages.

Three caveats from Google’s own documentation. Performance data is assigned to the canonical URL Google selects, so duplicates it has already folded together appear as one row. Queries issued by too few people are anonymized and left out of query-level tables, so long-tail overlap is under-reported. And this method only finds overlap Google has already acted on; two new pages on the same topic will not show up until they compete.

Worked example: two URLs taking turns

An illustration with made-up numbers, not client data. A software company has a template page, /templates/expense-report, and a blog post, /blog/how-to-write-an-expense-report, that also offers the template as a download. For the query “expense report template” the weekly Search Console rows look like this:

Google picks one URL each week. In the weeks it picks the blog post, clicks roughly halve, because a searcher looking for a template does not want an article about writing reports. The two pages do serve different intents, so merging them is the wrong fix. The right one is to differentiate: take the download off the blog post, keep the post for “how to write an expense report,” and link from the post to the template page with the anchor “expense report template.” Success looks like the template page appearing every week and the blog post dropping out of that query.

Read the results page

Before deciding that two pages compete, search the query yourself. If Google shows a mix of guides and product pages, it sees mixed intent and two of your pages with different jobs can both earn a place. If every result is the same kind of page, only one of yours should be competing for it.

The site: operator, as a spot check only

Searching site:yourdomain.com "topic" is a fast way to find pages you forgot existed. It is not an audit: Google says the site: results are not exhaustive, and recommends the URL Inspection tool to check whether a specific URL is indexed.

Content similarity, before Google finds it

Search data only shows collisions that already happened. To catch them earlier, compare what pages are about. For a handful of pages, read them side by side and ask: would the same searcher be satisfied by either? For thousands, you need a short, consistent description of each page’s primary intent that you can sort and compare. The cannibalization audit use case shows how that works in Morrison.

dash.morrison.app/lumenpay.com/chat

Illustration

lumenpay.com/Shared queries check

What do you want to know about lumenpay.com?

Ask about your website content...

Search ConsoleSerpApi

Claude Sonnet 5.5

Internal links that disagree

If your articles link to three different URLs with the anchor “content audit,” your own site is telling Google you have three answers. An internal link audit across your content surfaces these quickly.

Classify the overlap: the type decides the fix

Two notes on the last rows. A search intent alignment review is the quickest way to spot mixed-intent pages. For regional versions in the same language, Google’s multi-regional guidance is to pick a preferred version and use rel="canonical" together with hreflang so searchers get the right regional URL.

Choose the fix

Merge and redirect

When two pages answer the same question, combine the best of both on the stronger URL and permanently redirect the other. Google treats 301 and 308 redirects as permanent, and a strong signal that the target should become canonical. 302, 303 and 307 redirects are temporary and a weak signal, so use them only if you might undo the change. Redirect straight to the final URL: Googlebot follows up to 10 hops, but Google still recommends avoiding chains. Choosing the winner, merging the content and the full redirect checklist are covered in the consolidation guide.

Differentiate

When both pages have a reason to exist, change them until they no longer answer the same question. Pick one split and commit to it:

• Intent: one page teaches (“how to choose a CRM”), the other sells (“CRM for agencies”).
• Audience: beginner versus practitioner, with different depth and examples.
• Format: a full guide versus a checklist or template.
• Scope: the broad topic versus one subtopic that deserves its own page.

Then make every signal agree: title, H1, introduction, and the anchor text of internal links pointing at each page. Changing the title while the body still answers the same question rarely works.

Canonical tags

Use rel="canonical" when duplicate URLs must stay reachable: parameter variants, print versions, syndicated copies. Google describes it as a strong signal, not a command; it groups duplicates and may still pick a different canonical if other signals disagree. Google also asks you not to point the same page at different canonicals with different methods, and not to use robots.txt or the URL removal tool for canonicalization. Canonicals are for true duplicates; for two editorially different pages with overlapping intent, merge or differentiate instead.

Parameters and faceted navigation

Older guides tell you to configure parameter handling in Search Console. That tool no longer exists: Google retired the URL Parameters tool in 2022, saying its crawlers now work out which parameters matter on their own, and pointing to robots.txt for site owners who want more control. For faceted navigation, Google’s current guidance is:

• If filtered URLs do not need to appear in search, disallow them in robots.txt, or put filters in URL fragments, which Google generally ignores.
• Canonical tags and nofollow on filter links are, in Google’s words, generally less effective in the long term.
• If a filter combination should rank, such as /black-shoes, give it a clean URL, keep filter order consistent, and return a 404 when a combination has no results.

noindex

noindexremoves a page from Google’s results while keeping it live for visitors: thank-you pages, internal search results, short-lived campaign pages. It does not pass the page’s value to another URL, and Google does not recommend it for canonicalization within a site, because it blocks the page from Search entirely. Two mechanics to get right: the page must not be blocked in robots.txt, or Google never sees the noindex, and Google may take a while to recrawl a page and drop it. If a page you want gone has backlinks, redirect it to the surviving page instead.

Realign internal links

Every fix ends here. Once a topic has one primary URL, point your internal links at it, directly rather than through a redirect, and use anchors that describe that page. After differentiating, make sure each page’s anchors match its new job. The internal linking guide covers how to do this across a large site.

The audit, start to finish

A complete content inventory makes step 4 much faster, because you can see every page on a topic, not only the ones Google has already tested.

What to fix first

A rule of thumb for ordering the queue: fix first where a commercial query is involved and the evidence is clear; leave until last, or leave alone, low-volume informational overlap.

After the fix

Give Google time. For site moves, Google says it can take a few weeks or more for medium-sized sites to start showing new URLs, and longer for large ones; a handful of merges usually settles faster, but plan on four to eight weeks before judging. Check three things: the query’s Pages tab in Search Console shows one URL; retired URLs return the redirect you set (URL Inspection shows how Google sees them); and combined clicks for the query cluster hold or grow compared with the same weeks before the change.

Where it usually comes from, by site type

For publishers with years of archive, a content pruning analysis is often the fastest way to find the pieces to merge or retire.

Stop it coming back

Keep a keyword-to-page map

A keyword-to-page map assigns every topic or query cluster to exactly one primary URL. A spreadsheet is enough. The columns that matter are below; the full method, with a test for grouping queries and status codes for every row, is in keyword-to-page mapping.

The map only works if it is checked before every new brief and updated after every merge or launch.

Put the check in the brief

Every brief should name the primary intent, the query cluster, the existing pages on the topic and whether this piece is new or an update. The decision to write a new page or improve an existing one belongs before drafting, not after publication. Our guide to writing content briefs shows where this fits.

Give every content type a place

Define your content types (hub, supporting article, glossary entry, product page, comparison) and how they relate. Topical map and cluster planning makes it explicit: one hub per topic and a set of supporting pages, each with its own subtopic. If a cluster already has a page of that type for that intent, a new idea becomes an update.

Recheck on a cadence

Rerun the Search Console check monthly or quarterly, and always after a big publishing push, a migration or a team change. Those are the moments governance slips.

Where Morrison helps

The judgment calls in this guide stay with you. Morrison takes on the reading and data-joining that make them slow:

• Before you brief. Ask Morrison in chat which pages already cover a topic. It has read your whole site, so it finds pages by what they are about, not only by keyword.
• Checking a suspected pair or small cluster.In the same chat, Morrison pulls each URL’s Search Console queries, reads the pages and tells you whether they serve the same intent.
• Across a whole section.Every crawled page is classified by page type and search intent, so you can filter, for example, for informational pages sitting in a commercial section. To go further, a workflow runs the same steps on up to 500 pages in a section, such as a one-line statement of each page’s primary intent, and writes the results into your URL grid, ready to export and compare.
• On a schedule. An automation can run the Search Console check for you, weekly or monthly (up to daily on Pro), and email the queries where more than one of your URLs earned impressions.
• After the fix. Each recrawl records every URL’s HTTP status and redirect target, so you can confirm retired URLs redirect where you intended. Recrawls run monthly on Starter and weekly, every two weeks or monthly on Pro and Business.

The cannibalization audit, keyword-to-page mapping and content consolidation use cases show each of these in practice.

The short version

• Cannibalization is an intent problem, not a keyword one, and not a penalty.
• Confirm it with query-and-page data from Search Console: rotation or the wrong page ranking.
• Classify before fixing: technical duplicate, same intent, mixed intent, cluster without a hub, or cross-property.
• Merge same-intent pages with 301s, differentiate pages with different jobs, canonicalize true duplicates, and leave stable pairs alone.
• Handle parameters with robots.txt and clean URLs; the Search Console parameter tool is gone.
• Prevent it with a keyword-to-page map checked at the brief stage.