The reliable way to find keyword cannibalisation is to look at Google's own data, not to eyeball your content. Search Console shows you where two of your pages compete for one query. A quick site search and a check of your titles and H1s confirm it. Ignore keyword density, it is not the signal.
Where do you start?
Cannibalisation is an intent problem, so you are hunting for queries where two of your own pages compete for the same searchers. If you want the background first, see what keyword cannibalisation is and how to fix it, or the quick definition in the glossary. To find it, start with the one source that cannot lie to you: Google Search Console.
How do you find it in Search Console?
The Performance report is where cannibalisation shows up in black and white:
- Open Search Console, then Performance, then Search results.
- Set a meaningful date range, three months is a good default.
- In the Queries table, click a query you care about.
- With that query still filtered, open the Pages tab. It now shows only the pages that rank for that query.
- If two or more of your pages earn real impressions or clicks for the same query, that is cannibalisation.
For a whole-site sweep, export the query and page data and group by query, then count the distinct pages per query. Anything with two or more pages on a term you care about goes on the list.
What is the flip-flop test?
There is a second tell in the same report. For a suspected query, compare two date ranges, or watch it over a few weeks. If the URL that ranks for the query keeps changing, Google cannot settle on a winner between your pages. That instability is classic cannibalisation, and it quietly caps how well you can rank.
What does a site: search show?
For a fast sanity check, search site:yourdomain.com "your keyword" in Google. If several of your own pages surface for the exact phrase, note them. This is not proof on its own, queries and intent still matter, but it is a quick way to see which pages are in the running before you confirm with Search Console.
How do you check on-page prominence?
Now look at the suspects themselves. Pull the titles, H1s and opening lines of the pages Search Console flagged, by hand on a small site, or with a crawl on a large one. The pages that compete are the ones carrying the same target term in those prominent spots. Prominence in the title, H1 and intro is what makes pages compete, not how many times a word appears in the body.
Why confirm before you act?
A shared keyword is not automatically a problem. Before you merge or rewrite anything, confirm the two pages actually serve the same intent. If they answer genuinely different questions, leave them alone. Once you have a confirmed cluster, move on to how to fix it. Limecube's Site Audit does this whole sweep for you: it combines Search Console overlap with on-page prominence and clusters the results, so you skip the manual export. The 7-day trial covers it.