Keyword Cannibalization Checker

Start here: your export probably won’t work

This is the part most guides skip, and it will save you twenty minutes. Search Console’s export doesn’t pair queries with pages. The Performance report gives you a Queries tab and a Pages tab, and exporting produces one file of each — queries with their metrics, pages with theirs, and nothing connecting the two.

Cannibalization is a question about the connection. You cannot answer “which pages competed for this query” from a file that never recorded which page a query belonged to. So the first job is getting an export that carries both:

• Filter by page, export the queries. In the Performance report add a Page filter for one URL and export its queries. Repeat for the pages you care about and paste the files one after another — this tool reads them as a single dataset. Tedious, but it needs no tooling.
• Query the Search Console API. Requesting the query and page dimensions together returns the pair in one response.
• Use the BigQuery bulk export, which stores the pair for every row and is the right answer for a large site.

If you paste a queries-only file, the tool says so rather than showing an empty table — an empty table would read as “no cannibalization found”, which would be a lie about a question it never had the data to ask.

What the tool is actually looking for

It groups your export by query and keeps the queries where two or more of your URLs earned impressions. For each one it shows every competing URL, its average position, its impressions and its clicks, and marks the one earning the most clicks as the leader.

The number worth looking at is the impressions on everything exceptthe leader. That’s the audience that saw a page of yours other than the one you’d have chosen. It isn’t a loss estimate and the tool doesn’t pretend it is — there is no cannibalization score here, because the underlying data doesn’t support that kind of precision and inventing one would imply it does.

Why a flag isn’t a diagnosis

Two URLs on one query is normal in more cases than people expect.

• Google swaps pages over time. An export covering 90 days shows every page that ranked at any point in those 90 days. Two URLs in the list may mean Google tested one, changed its mind, and settled — which is resolution, not a problem.
• Results differ by device and country. The same query can return your product page on mobile and your guide on desktop, and both are recorded.
• Some queries genuinely deserve two pages. A broad term can legitimately surface a category page and a comparison article, and Google shows whichever fits the searcher.

The pattern actually worth acting on is a query where two URLs both earn real impressions, both sit in mediocre positions, and neither pulls away. That’s the shape of a split signal. One page at position 3 and another at 47 is not cannibalization; it’s a page that ranks and a page that doesn’t.

Remember too that these positions are averages over impressions, not ranks held — so “both around 12” may be two pages alternating between 5 and 20 rather than two pages sitting together.

Before you consolidate

The instinct is to merge the weaker page into the stronger one and redirect. Often right, occasionally expensive: the page losing this query may be earning far more across a dozen others, and folding it away costs you all of them to win one.

So check what each URL ranks for in total before you delete anything. Organic Keywords lists every keyword a page earns, which turns a one-query decision into an informed one. Afterwards, rank tracking is how you confirm the surviving page actually took the position rather than both of them dropping.

If you came here from a page-two list, the striking distance finder uses the same export and flags which of those queries have more than one of your URLs behind them — often the real reason a page won’t move. The post on which page-two wins are real makes the case that this check belongs beforeany content work, not after it hasn’t worked.