Keyword Cannibalization: Find and Fix Overlapping B2B Pages

Keyword cannibalization is what happens when two or more pages on the same site try to answer the same search for the same reader, so they compete with each other for one place. "Cannibalization" is borrowed from retail, where a new product takes sales from the company's own existing product instead of from a competitor. The fix is to decide which page owns the task, then merge, differentiate or leave alone according to evidence. Repeated keywords alone do not establish a problem.

That last point matters for manufacturers. A guide to choosing a material and a product page selling that material share most of their vocabulary and serve different readers. They can and should coexist, and the blog versus landing page comparison explains how to keep their jobs separate. Two near-identical buying guides are the real candidates.

Five kinds of overlap on a manufacturer's site

The label covers several situations with different remedies. Identify which one you have before changing anything.

Kind	What it looks like	Usual remedy
Two posts, one question	"How to choose a stainless steel grade" and "Best stainless steel for fasteners"	Merge into the stronger post and redirect the other
Post versus product page	A blog post appears for a product query the product page should answer	Differentiate. Narrow the post to the learning task and link it to the product page
Archive versus article	A tag or category archive appears for a query instead of the guide it lists	Give the guide contextual links and review whether the archive needs to be indexed
Same page, several addresses	Tracking parameters, filter URLs, print versions, http and https copies	Technical duplication. Handle with canonicals or redirects. No rewriting needed
Market or language versions	A UK page and a US page with nearly the same English text, or distributor pages per country	Usually legitimate. Mark the versions so the right one is shown in each market

The fourth row is not an editorial problem at all, and treating it as one wastes time. Google's guidance on duplicate URLs lists redirects and rel="canonical" as strong signals for the preferred address and sitemap inclusion as a weak one.

The fifth row is the one exporters meet most. Google's documentation on localized versions recommends hreflang annotations for similar content with small regional variations in one language, as well as for full translations. Regional pages should differ where the market differs: units, standards, voltage, stocked sizes, the local contact. Pages that vary only by a city or country name and funnel visitors to one destination fall under what Google's spam policies call doorway abuse.

Find the overlap in Search Console

Suspicion is cheap. A single search where two of your pages appear proves little. Collect evidence across several queries and a few months.

1. Open the Performance report for search results and set the date range to the last three months or longer.
2. Add a query filter for the phrase you suspect, using the "contains" option so that close variants are included.
3. Switch to the Pages tab. If one URL appears, there is no overlap for this phrase. If two or more appear with meaningful impressions, write them down.
4. Remove the query filter and add a page filter for the first URL. Switch to the Queries tab and export the list. Repeat for the second URL.
5. Compare the two lists. Count the queries that appear in both, and read them. Shared queries about the same task are the signal. Shared vocabulary across different tasks is normal.
6. Use the date comparison to check whether the URL shown for those shared queries changed between periods. A pair that keeps swapping is a stronger candidate than a pair where one page is consistently preferred.
7. Open both pages and read them side by side. This step settles more cases than the data does.

One caution about the numbers. Google's Performance report documentation explains that data grouped by query is aggregated by property, while data grouped by page is aggregated by page. Totals therefore differ between the two tabs, and impressions for a query cannot simply be added up across pages. Compare patterns and read the lists. Do not build the decision on small differences in totals.

These patterns identify a candidate. They do not prove that the overlap caused a drop in visits; seasonality, a competitor's new page or a site change can produce the same picture. Keeping observations apart from recommendations, as in the Search Console content plan, makes the later decision easier to defend.

Decide: merge, differentiate, canonical or leave

Evidence	Decision
Both pages answer the same question for the same reader, and one is clearly better	Merge into the better page, redirect the other
Both answer the same question, and each has parts the other lacks	Merge the content into one page at the URL with more links and history, redirect the other
The pages share queries, and you can state a different reader or task for each in one sentence	Differentiate: rewrite titles and openings, and link each to the other
The pages are copies at different addresses	Canonical or redirect. No editorial work
The pages share words but the queries are about different tasks	Leave both alone
One page has no impressions, no links and no distinct purpose	Remove it, or redirect it if a relevant destination exists

The threshold for "shared queries" is yours to set. As a working default of ours, not a Google rule, we look harder at a pair when more than about a third of each page's queries also appear on the other. Use it to choose which pairs to read first.

How to merge

Pick the surviving URL. Move across every useful paragraph, table and image from the other page, then edit the result into one coherent article, because two posts pasted together read like two posts. Set up a permanent redirect from the retired URL to the survivor. Google's redirect documentation says a permanent redirect is used as a signal that the target should be canonical, while a temporary redirect is not. Then search your own site for links to the retired URL and point them at the survivor directly. The update or write new guide covers how to combine the content without losing what worked.

How to differentiate

Write one sentence per page naming its reader and task. Rewrite each title and opening paragraph to match. Remove from each page the sections that belong on the other and replace them with a link. If you cannot write two different sentences, the pages should be merged.

When a canonical is the right tool

Use it for genuine duplicates that must both stay accessible, such as a printable version or a filtered product listing. It is a poor patch for two articles with different text, since the canonical is a signal and Google may select another URL. The same guidance advises against using robots.txt for canonicalization, and against using noindex to steer canonical selection within a site because it removes the page from Search entirely.

Example: four pages about one material

Example. Brightline Fasteners is a fictional supplier, invented for this article. Its marketing lead suspects overlap around "stainless steel fasteners" and finds four URLs appearing for related queries.

URL	Reader and task	Finding	Decision
Product category: stainless steel fasteners	A buyer evaluating the range and requesting a quote	Appears for supplier and product queries	Keep. This page owns the commercial task
Post: "How to choose a stainless steel grade"	An engineer selecting between grades for an environment	Appears for comparison and "which grade" queries	Keep as the surviving guide
Post: "Best stainless steel for fasteners"	The same engineer, the same choice	Most of its queries also appear on the grade guide. Contains a useful table on galling	Merge into the grade guide, keep the galling table, redirect
Post: "Stainless fasteners at our 2024 trade stand"	Nobody searching	A handful of impressions on the brand name	Leave. It competes with nothing

After the merge, the grade guide links to the category page where it names each grade, and the category page links back to the guide beside its grade filter. One URL was retired. Two of the four pages needed no change, which is a common result of an audit and a useful one.

Check the result and keep it from returning

Save a baseline before changing a URL: its queries, its inbound internal links and a copy of its content. After the change, confirm that the retired URL redirects, that the survivor's canonical points to itself, and that no internal link still targets the old address. The URL Inspection tool in Search Console shows the canonical you declared next to the one Google selected. Compare later performance with the baseline over a similar period, and attribute changes cautiously.

Prevention is cheaper than cleanup.

• Keep a topic register. One row per published page: URL, reader, task, main phrase. Before commissioning a post, search the register and the site. Overlap is nearly always created by publishing without looking.
• Search before briefing. A site search for the planned title's main phrase takes a minute.
• Repeat the Search Console check twice a year for your ten or twenty most valuable phrases.
• Record decisions. A note saying why two pages were merged stops the next writer recreating the retired guide under a new title.

Blogo's what to write next recommendations are built from the site's existing pages and its Search Console data, which helps with the look-before-you-write step. It does not merge or refresh old posts for you, so the decisions in this article stay with the site owner.

Frequently asked questions

What is keyword cannibalization in one sentence?

Two or more of your own pages competing to answer the same search for the same reader.

How do you fix it?

Identify the kind of overlap, confirm it with query and page data, then merge duplicates of the same task, differentiate pages with separate tasks, and use canonicals or redirects for technical duplicates. Leave pages alone when the evidence shows no conflict.

Can two pages from one site appear for the same keyword?

Yes, and it is often fine. A product page and a guide can both appear for a broad phrase because searchers using it want different things.

Should we always remove the lower-traffic page?

No. It may serve a specific application or buyer that the broader page does not cover. Compare purposes before comparing traffic.

Will a canonical tag fix it?

Only for true duplicates. Two different articles with unclear purposes remain unclear with a canonical added.

What is the 80/20 rule in SEO, and does it apply here?

It is a rule of thumb borrowed from the Pareto principle: a small share of pages tends to account for a large share of results. It is not a Google rule and the ratio is not a measurement. The practical use in an overlap audit is to start with the pages and phrases that matter most to the business and stop when the remaining pairs are harmless.