How to Find Keyword Cannibalization in Search Console

Alejandro Rioja
Alejandro Rioja
6 min read
TL;DR

Keyword cannibalization means two pages on your site compete for the same query. Search Console shows it directly: filter Performance by one query, open the Pages tab, and look for more than one URL with meaningful impressions. Only act when both pages rank within a few positions of each other and swap places. In my own check on the query 'reverse video search', the second page sat at position 53.6 with 30 impressions, so I left it alone. Fix real cases by merging, redirecting or re-targeting, and wait 28 days.

Building with AI Agents

Learn how to leverage AI agents for marketing, sales and product. Every week I share what I learned with 5,400+ readers.

Most advice on keyword cannibalization tells you to merge pages that look similar. That’s backwards. You merge pages after Search Console shows both of them fighting for the same query, and in my experience that’s rarer than the audit tools suggest.

Cannibalization means two or more pages on your site compete for the same search query, so Google keeps swapping which one it shows and neither builds a stable position. The check takes about twenty minutes and needs nothing but Search Console. Here is the routine, with a live example from my own site.

Step 1: start from a query, not from a page

Cannibalization is a property of a query. Pick the queries you care about first: your top five by clicks, plus any query where you sit between positions 5 and 20.

  1. Open Performance → Search results and set the date range to the last 28 days. Use 90 days if the site is small and queries are thin.
  2. Click + New → Query, choose Exactly matching the query, and type the query.
  3. Switch to the Pages tab under the chart.

If the tab lists one URL, you’re done with that query. There’s no cannibalization. If it lists two or more, go to step 2.

Step 2: read the rows for three things

For each URL ranking on that query, write down clicks, impressions and average position. Then ask:

  1. Do both pages have meaningful impressions? A page with 12 impressions in a month isn’t competing with anything. It’s noise. I treat anything under about 5% of the leading page’s impressions as noise.
  2. Are the positions close? Real cannibalization shows two pages within a few positions of each other, usually both on page one or page two. A gap of ten positions or more means Google has already picked a winner.
  3. Do they trade places? Add the Date dimension (or compare two date ranges) and filter to one page at a time. If Page A holds position 6 for two weeks, then Page B takes over at position 8 while A drops off, that’s the swap. A stable winner with a distant runner-up is healthy.

All three must be true before you do anything.

A live example from my own site

I ran this on my reverse video search query for 20 October to 18 November 2026. Search Console returned four URLs for “reverse video search”:

URLClicksImpressionsAvg. position
/tools/reverse-video-search/ (the free tool)2382,91610.5
/de/reverse-video-search-explained/11510.4
/reverse-video-search-explained/ (English guide)03053.6
/it/reverse-video-search-explained/019.0

On paper that looks like trouble: a tool page and a guide about the same subject. Run the three tests and it isn’t. The tool holds 98% of the impressions, the English guide sits at position 53.6 with 30 impressions, and the German and Italian guides barely register. The tool wins, the guide supports it by linking to it, and the guide targets a different job, working out where a clip came from. I changed nothing.

That’s the usual result. Pages that overlap in topic often serve different intent, and Google is good at telling them apart. A tool that lets you do the thing and a guide that explains the thing can both be right.

I’d also check the other side. The 90-second version of this is to pull your top 20 queries and open the Pages tab for each. Most will show one URL.

Step 3: confirm it’s two pages and not one page in two languages

On a multilingual site, don’t mistake a translation for a competitor. My German guide above shows up on an English query at position 10.4, which is a different-language page showing on an English query, not a conflict. Check that the competing URLs aren’t the same post under different language folders. If they share a translation_key or the same slug, they’re translations, and the fix is hreflang, not merging. I cover how to read that split in how to read Search Console by language.

Step 4: pick the fix that matches the cause

When all three tests say a real conflict exists, choose one of four moves:

  1. Merge and redirect. The two pages say the same thing for the same reader. Move the better parts of the weaker page into the stronger one, 301 the weaker URL to the stronger, and update every internal link to point at the survivor. This is the right call about half the time.
  2. Re-target one page. They serve different intent but share a keyword. Change the weaker page’s title, H1 and opening paragraph so it chases a different query, such as a “how to” angle instead of a “best” angle. Check which one fits using the clustering approach in what is keyword clustering.
  3. Point internal links at the winner. If one page is clearly the one you want ranking, find the pages that link to the other with the query’s wording as anchor text and repoint them. It’s slow and dull, and it works more often than people expect.
  4. Add a canonical. Only for near-duplicates that must both stay live, such as a printable version. Don’t use it as a shortcut for two pages with different content.

Change one page, note the date, and wait 28 days before judging it. If the site is mid-rollout of a Google core update, wait until it finishes, because rankings move on their own while it runs. The November 2026 core update began on 10 November and was still rolling out when I pulled these numbers, so I’d have waited on any fix I’d made today. My routine for that is in how to read Search Console during a core update.

When a paid tool is worth it

The manual check covers your top queries. If you manage a client site with thousands of queries, going one query at a time stops being practical. Semrush’s Position Tracking and Organic Research reports can flag queries where more than one of your URLs ranks, which gives you a shortlist to verify in Search Console. I use Semrush for that when I need the whole site at once (affiliate link, and I’d say the same without it). Treat its list as candidates, not diagnoses. It can’t tell you whether the pages trade places, and Search Console can.

The prompt I paste into Claude

Export the Search Console Performance report with both the Query and Page dimensions (the API export in Looker Studio or the Google Sheets connector gives you query and page together), then paste or attach it:

I’m attaching a Search Console export with one row per query and page pair for the last 90 days. Find every query where two or more of my URLs have at least 50 impressions each and their average positions are within five places of each other. For each query, list the URLs, impressions, clicks and positions, and sort by total impressions. Output a table only. Do not recommend merging anything and do not guess at intent.

Before you trust the answer, open Search Console for the top two queries on the list and confirm the numbers. Models misread columns in exported sheets more often than you’d expect, and one bad row sends you to merge the wrong page.

If you’d rather skip the export, SiteWide MCP is the connector I use to pull Search Console into Claude directly, and the query-then-pages check above is one question to it.

Advertisement
Keep reading

Related posts

Keep reading

Get the AI playbook in your inbox

Learn how to leverage AI agents for marketing, sales and product. Every week I share what I learned with 5,400+ readers.

↵ ↵ to see all results esc esc to close