Keyword cannibalization: the audit that only pays off sometimes
A study that undercuts the standard advice
The standard cannibalization advice treats any overlap as a fault to fix: find every keyword with more than one ranking page, pick a winner, redirect or merge the rest. Andrew Witts of Studio 36 Digital tested that assumption directly, publishing the results on 14 January 2026 and updating them on 13 June 2026. He pulled the top 25 traffic-driving keywords for 100 high-authority sites across 10 industries through the Ahrefs API, then counted how many URLs from the same domain ranked in the top 100 on Google UK for each keyword.
The successful sites in the sample were not running one page per keyword. The average was 4.7 competing URLs per top keyword, and 68% of sites showed what Witts classifies as significant cannibalization, five or more URLs ranking for the same term. Only 12% of sites kept to one or two URLs per keyword, the pattern most cannibalization guides describe as correct. One site had 59 separate URLs ranking for a single keyword. If cannibalization by itself reliably cost rankings, the highest-authority sites in the study would not be the ones doing it the most.
Where the overlap actually costs you
The study's more useful finding sits in the breakdown by authority tier rather than the headline average. Domains rated DR 75 and above showed no consistent relationship between URL count and ranking position. Domains under DR 50 did: sites with three or more pages competing for one keyword consistently settled at positions 8 to 20, well outside the results most searches actually click. Industry shaped the raw averages too. News and media sites averaged 8.2 competing URLs per keyword and marketing and SEO tool sites 7.9, both categories where publishing volume and topic overlap are structural. Government and reference sites averaged 1.3, where one authoritative page per topic is closer to the norm.
What the study does not contain is a before-and-after number. It documents how many competing URLs exist and where they rank, not what happened to traffic or position after a site consolidated or redirected the losing pages. That gap matters for anyone about to spend a week merging content on the strength of this one dataset alone, and it is the reason the check below asks you to look at your own authority tier and your own rankings before assuming the fix applies to you.
What Google's own team actually says
In response to a question posted on Bluesky about Search Console showing multiple URLs from one site ranking for a query, Google's John Mueller pushed back on treating that as evidence of a problem, a response reported by Roger Montti in Search Engine Journal on 22 September 2025. Mueller said Search Console "shows data for when pages were actually shown, it's not a theoretical measurement," and that "if you have 3 different pages appearing in the same search result, that doesn't seem problematic to me just because it's 'more than 1.'" He pointed to queries like cheese as an example where several legitimate, non-competing pages, a shop, a recipe, a knife guide, can reasonably appear together. His actual recommendation was narrower than "audit everything": "reduce unnecessary duplication and spend your energy on a fantastic page."
Read together, the Studio 36 data and Mueller's comment point the same direction from two different angles. The data shows multiple rankings are the default state for successful sites, not an anomaly. Mueller's comment says the presence of multiple pages is not itself the signal to act on. What both leave out is any claim that cannibalization never matters, only that counting URLs per keyword is the wrong first filter.
The audit worth running
Skip the keyword-by-keyword URL count as a first step. It flags the same pattern on a DR 90 site that is working exactly as intended and a DR 30 site that is actually losing rankings to itself, and the study above says those two cases behave differently. Run the narrower version instead.
- Pull your own domain rating first. If you are at DR 75 or above, the study found no consistent relationship between competing URLs and rank at all. Multiple pages ranking for a related term is not automatically worth your time.
- Under DR 50, check position, not just count. Find keywords where three or more of your own pages rank in the top 100. If those pages are clustered at positions 8 to 20, that matches the pattern the study found actually costing rankings.
- Separate real overlap from adjacent intent. Two pages both trying to rank for the identical phrase with the identical intent is overlap. A product page and a buying-guide page that both mention the same term, Mueller's cheese example, usually is not. Read each page before merging it.
- Consolidate the weakest, not the newest. Where real overlap exists, keep the page with the strongest existing position and internal links, then redirect and fold the content from the others in rather than deleting it outright, which matches the safe-migration approach our website migration guide covers in more depth.
This is the same lens our on-page SEO guide uses for title tags Google rewrites anyway: the leverage is not in a blanket rule, it is in the specific pages where the documented pattern actually applies to you. It is also the same instinct behind not publishing on autopilot, which our guide to how much content you should actually publish covers from the volume side, since a content calendar with no plan for overlap is exactly how a site ends up with three thin pages competing for one term in the first place.
What we could not verify
The Studio 36 Digital study measures which sites have competing URLs and where those URLs rank. It does not include a controlled test of what happens after a site consolidates or redirects the losing pages, so the positions-8-to-20 correlation on lower-authority sites is evidence of a pattern, not proof that fixing it recovers rankings by a specific amount. The study also covers Google UK results only, drawn from Ahrefs' index rather than Google's own data, and Ahrefs' coverage and ranking estimates do not always match Google's live results exactly. Mueller's comment was made informally in response to one Bluesky question rather than in Google's published documentation, and it addresses whether multiple rankings are inherently problematic, not whether consolidating them ever helps. Neither source settles the question for every site; both narrow it considerably from "any overlap is a problem."
Sources
The cannibalization prevalence, authority-tier and industry figures come from Andrew Witts, Studio 36 Digital, The 2026 Keyword Cannibalisation Study, published 14 January 2026, updated 13 June 2026. John Mueller's comments are from his response to a question posted on Bluesky, reported by Roger Montti, Google Answers SEO Question About Keyword Cannibalization, Search Engine Journal, 22 September 2025.
Common questions
Does keyword cannibalization actually hurt rankings?
Sometimes, and mostly on weaker sites. A January 2026 study of 100 high-authority sites found the average successful site already ranks 4.7 different URLs for each top keyword, with no consistent cost on domains rated DR 75 or higher. On domains under DR 50, three or more competing pages correlated with those pages settling at positions 8 to 20.
Does Google penalize sites for having multiple pages target the same keyword?
Google's John Mueller has said no, not inherently. Responding to a question about Search Console showing several URLs from one site for the same query, he said that alone "doesn't seem problematic," and recommended reducing unnecessary duplication rather than treating overlap itself as a fault.
How do I know if cannibalization is actually a problem on my own site?
Check your domain rating and ranking position together. The correlation with lost rankings only showed up on sites under DR 50 with three or more pages stuck at positions 8 to 20. A strong page at position 3 with a second page ranking 40 on the same site is not the same problem.
The row-by-row fix, not a blanket rule
On-Page Optimization audits up to 10 pages, flags real keyword overlap against your own authority and rankings, and rewrites titles, headings, internal links and schema page by page, with the before state captured so the lift is provable.