How to find keyword cannibalisation

I watch for it on leej.uk because this site has both free tools and long guides, sometimes on the same subject. The split only works if each URL has a different job. The Keyword Cannibalisation Finder and a Search Console export are how I check that the split is real.

Two pages, one query, a split of attention

The simple picture is this. A person searches for a phrase you care about. Your site has two pages that both look like a reasonable answer. One week Google shows page A, the next week page B, or both appear with modest positions and neither settles. You have split the attention that could have attached to a single, clearer result.

That split can show up as:

• Two URLs on the same query in Search Console, both with impressions, perhaps alternating in average position.
• Titles and H1s that are close enough to be interchangeable, such as two pages both called some variation of “How to create a sitemap”.
• A tool landing page and a guide that repeat the same promise in the first paragraph, with no distinction about who each is for.

It is not the same as ordinary duplicates. Duplicate URLs (HTTP and HTTPS, trailing slashes, parameters) are a canonical problem, covered in the canonical tags guide. Cannibalisation is usually two deliberately published pages, each with its own content, competing for one intent.

It is also not the same as ranking for many related queries with one strong page. That is often healthy. The issue starts when two pages want the same query, in the same sense, for the same reader.

When it is a problem, and when it is not

It is a problem when:

• The pages answer the same question, at the same level of detail, for the same audience.
• Search Console shows the query hopping between URLs, or the “wrong” URL collecting the impressions.
• You keep updating both pages in the same way, because you cannot tell which one is the main result.
• Internal links and anchors treat them as interchangeable, so neither page is clearly the hub.

It is often not a problem when:

• The intents are different. “What is a sitemap” and “How to keep a sitemap accurate” can live as two pages, or as one page with a clear structure. Two almost identical how-tos cannot.
• The formats are different and the titles say so. A working tool and a guide can both mention the same topic if one is “paste your file and check it” and the other is “here is the method, including when not to use a file”.
• The pages target different countries, languages or products, and hreflang or a clear title already separates them.
• One URL is a thin leftover that already redirects, or is noindex. That is a cleanup task, not cannibalisation between two live results.

I do not treat every overlapping impression as an emergency. Search Console will show several of your URLs for a head query if the site is genuinely about that subject. Look for confusion, not for any overlap at all.

The same scoping habit I use in the iOS app MVP checklist applies: if you cannot explain, in one sentence, how page B is not page A, you probably have one page too many.

Finding it in a Search Console export

Start with evidence, not with a feeling that two titles look similar.

1. In Google Search Console, export the Performance report for the pages and queries you care about, over a period long enough to see swapping (several weeks, not two days).
2. Look at a query and list the pages that received impressions or clicks for it. Two or more of your URLs, repeatedly, is the candidate set.
3. Open each URL. Read the title, the H1, the first paragraph and the obvious intent. If they are trying to do the same job, you have a case. If one is a definition and one is a checklist, you may only need clearer titles.
4. Check the average position over time, if you have it. A clean hand-off (page A replaced by a newer page B, and A declining on purpose) is not the same as weekly swapping.
5. Cross-check indexing with the Search Console Indexing Report Analyser if a candidate URL is also excluded, duplicated or canonicalised elsewhere. A crawl issue can look like competition when one of the pages is not fully in the index.

The Keyword Cannibalisation Finder is built to highlight queries with more than one of your pages attached. The GSC Regex Builder helps you filter a messy query set, for example to group “sitemap” queries without picking up unrelated words. Neither tool can see intent; you still have to read the pages.

If you do not have Search Console access yet, you can still do a cheaper pass: list every title and H1, and look for near-duplicates. The Title and Meta Bulk Auditor is useful for that catalogue. It will not show you which query hops, but it will show you the collisions you created in the title tags.

Titles, intent and the job of each URL

Before you merge anything, write a one-line job for each page:

• Who it is for.
• What question it answers.
• What the person should be able to do after reading or using it.

Then look at the title, meta description and H1. Those should state the job, and they should not be interchangeable. Two pages that both claim to be “the complete guide to robots.txt” are asking to compete. A page titled “robots.txt vs X-Robots-Tag: which to use” and a tool titled “robots.txt Tester” are not, if the first paragraph of each agrees.

Intent is the rest of the page, not just the title. If the guide quietly contains a live checker, and the tool page quietly contains a full tutorial, you have two mixed pages. Pull the tutorial back to the guide, or the checker back to the tool, until each URL has a primary job. Writing for AI answers and search is the same discipline in the prose: answer one question clearly, and do not pretend a page is something it is not.

Compare this to a product decision such as native iOS or cross-platform. The useful output is a choice with reasons, not two articles that each argue both sides and leave the reader in the same place. Give each URL a choice to own.

Fix by clarifying, merging or redirecting

I use three fixes, in this order.

Clarify first, when both pages deserve to live. Tighten the titles, the H1s and the opening paragraphs so each query has a preferred URL. Adjust internal links so the hub page is the one you want to rank for that job, and the other page points up to it with a descriptive anchor. Add a short “if you want the working checker, use this tool” line, or the reverse. Then wait and watch the same query in Search Console. Many small-site collisions disappear once the pages stop pretending to be twins.

Merge, when the pages are the same article in two coats. Choose the stronger URL (better links, better history, or the cleaner slug). Move any unique sections across, keep the examples that are still accurate, and make one complete page. Then redirect the retired URL to the survivor. Do not leave the old page up with a canonical that points at the new one if you have no reason to keep the old URL alive; a redirect is clearer for a true merger.

Redirect without a full rewrite when one page is simply outdated, thinner, or a leftover from a rename. Confirm that the destination answers the same intent. A redirect from a how-to to an unrelated commercial page is a different mistake.

What I do not do as a first step:

• Delete a useful page because two titles felt close.
• noindex a good guide just to “focus the signals”.
• Canonicalise two different resources at each other because that looked quicker than editing. Canonicals are for the same or very similar documents, not for a tool and a tutorial, and not for two essays that should have been one.
• Stuff every overlapping keyword into a single mega-page that no longer answers any question well.

After a merge, update the internal links, as in the internal linking guide, and remove the retired URL from the sitemap.

Guides versus tools: give each a different job

This is the pattern I use on this site, and it is the one that otherwise causes the most avoidable cannibalisation.

A tool page should help someone do a job with their own input: paste a sitemap, test a header, check a contrast pair. The title can include the object (“XML Sitemap Validator”). The first paragraph should say what you put in and what you get out. A short note can point at the longer method.

A guide should teach the method: what to include, what to leave out, when the tool is the wrong instrument. The title should sound like a question or a task (“How to keep an XML sitemap accurate”). The first paragraph should answer that task. Links to the tool belong in the body, at the point you would reach for it.

When those jobs blur, both URLs start competing for “xml sitemap” or “robots.txt tester”. The fix is not to rename the tool into a vague brand word. The fix is to keep the titles honest and the content in the right place. The same idea sits behind small, focused Apple apps: one job per surface, stated plainly.

If you only have time to maintain one URL, keep the guide if the method is the asset, or the tool if the working checker is the asset, and let the other be a short page that points at it. Two half-maintained twins are how cannibalisation creeps back.

A check before you publish the next page

Add a short gate to the writing process:

1. List the primary query, and two or three close variants, for the new page.
2. Search your own site and your title catalogue for those phrases. Open anything close and read the intent.
3. If an existing page already does that job, extend it instead of publishing a sibling. If the new page has a genuinely different job, write a title that a stranger could tell apart in a results list.
4. Decide, in advance, which existing pages will link to the new one, and which query the new one will leave alone.
5. After launch, look at the query in Search Console once you have some impressions. If a parent page and the new page are swapping, clarify immediately, before both accumulate half-finished updates.

This takes minutes on a small site. It is much cheaper than merging two articles in six months’ time.

What not to do

• Do not treat every shared word as cannibalisation. Two pages that both mention “iOS” are not competing until they chase the same query with the same intent.
• Do not automatically 301 the newer page to the older one without reading them. History is useful; accuracy is more useful.
• Do not create a third “ultimate guide” as a compromise. That usually gives you three overlapping URLs.
• Do not hide one of the pages behind noindex as a temporary experiment and then forget it. Temporary tags have a way of becoming permanent leftovers.
• Do not buy into invented percentages about “how much traffic you lose to cannibalisation”. Look at the queries you actually care about, and at whether a person would be served by one clearer page.

Key takeaways

• Cannibalisation is two or more of your pages competing for the same query and the same intent, often visible as swapping URLs in Search Console. Overlap alone is not always a problem.
• Confirm it with a query-to-page export and a careful read of titles, H1s and opening paragraphs, not with a guess from similar slugs.
• Fix it by clarifying the job of each URL, or by merging and redirecting, not by deleting useful content or misusing canonicals.
• Give tools and guides different jobs, different titles and different first paragraphs, then link them to each other.
• Check for an existing page that already does the job before you publish the next one, and watch the query once the new URL has impressions.

If you want a second pair of eyes on a pair of pages that will not settle in the results, tell me about it.

Frequently asked questions

Written by Lee Johnson. Published 10 November 2026.