Documentation
Keyword Cannibalization: find URLs competing for the same search
See the searches where two or more of your URLs collect meaningful impressions, which page leads each one, and how much traffic the others draw away.
Keyword Cannibalization (the report is titled URL Cannibalization) finds searches where Google shows more than one of your URLs often enough that they compete with each other. When that happens, rankings tend to alternate between the pages, neither builds up the signals one page would, and clicks are split. The report tells you which searches are affected, which URL is strongest for each, and how much traffic lands on the others.
Set up a run
- Analysis Period: the last 7, 14 or 31 days (14 by default), ending on the most recent day Search Console has data for.
- Folder Filter (optional): limit the run to one section, such as
/blog/. After the first run, the folders of your site appear as chips with their page counts. - Brand Keywords (optional): queries containing these terms are left out, because several pages ranking for your own name is normal. Paste a list, or press Use property brand terms to reuse the terms saved in Property Setup.
- Min Impression Threshold: 5%, 10% (default), 15%, 20% or 25%. A second URL only counts as competing when it earns at least this share of the top URL's impressions for the query.
Results are kept for 24 hours under Previous runs, so you can reopen a recent run without fetching Search Console again.
How a query is flagged
AutoPSEO fetches query-by-page rows for the period and, for each query:
- Sorts the URLs that appeared by impressions. The first is the top URL.
- Keeps the URLs whose impressions reach the threshold share of the top URL's.
- Flags the query when at least two URLs remain.
Impressions Lost is the sum of impressions that landed on the competing URLs rather than the top one; clicks lost are worked out the same way. Up to 2,000 flagged queries are listed, busiest first.
Read the results
The cards show Cannibalized Keywords, Affected Pages, Total Impressions Lost and Avg Pages Per Keyword, and Biggest conflicts charts the 15 queries with the most impressions lost.
Two views hold the detail:
- Cannibalized keywords: one row per query with the number of competing URLs, clicks, impressions, average CTR and position, impressions lost, and the URLs themselves with the top URL starred.
- Affected URLs: one row per page with how many flagged queries it competes for (Keyword Count), for how many of them it is the strongest page (Top for), and its traffic.
A URL with a high Keyword Count and a low Top for is usually the weaker duplicate.
Decide what to do with each conflict
Not every overlap is a problem. Google says some duplicate content on a site is normal and not a violation of its spam policies (canonicalization), so this is about focus, not a penalty. A category page and a product page can both rank for a broad search and serve different intents. Act when the pages answer the same question:
- Merge the weaker page into the stronger one and redirect it permanently, so links and signals consolidate.
- Differentiate the pages so each targets its own intent, and change titles and headings to match.
- Point internal links for the search at the page you want to rank, using descriptive anchor text.
- Canonicalise genuine duplicates (filters, parameters, print versions). Google describes the options in consolidating duplicate URLs.
The playbook Fix keyword cannibalization walks through the decision in more detail. For a statistical view of how spread out each query is, see Cannibalization Entropy.
How AutoPSEO prevents and fixes it
The content engine is built not to create this problem. Every proposed topic passes a coverage check against your existing pages (titles, headings, slugs and the searches each page already wins), and a reviewing model sees the five closest pages before a topic is queued; a topic your site already covers is skipped or turned into a new section of the existing article (how topics are chosen). When two managed articles do answer the same searches, they can be merged: the weaker article is taken down and your site is told to answer its old address with a 301 to the one that stays (Keeping articles fresh). Your AI assistant can also query conflicts through the cannibalization tool of the MCP server.