Keyword cannibalization
Keyword cannibalization is what happens when two or more pages on the same site target the same search term or the same question, so they compete against each other instead of against outside competitors. Instead of one page building up authority, several split it. Search engines can't settle on which one to rank, and an AI system looking for a single citable source finds several half-answers instead of one whole one.
Why it matters
When several pages chase the same term, they divide the clicks, links, and trust signals that one consolidated page could have earned outright. In classic search, that shows up as ranking instability: page A shows up this week, page B next week, and neither climbs far because neither one accumulates the full signal. You end up competing against yourself for a spot you could otherwise own outright. Fixing this isn't just tidying up — it's usually the fastest way to improve a term you already half-rank for, because you stop splitting the one thing that would have pushed a page over the line.
How it arises
Almost nobody sets out to cannibalize their own rankings. It builds up over time: a guide gets written, then a near-duplicate gets written a year later by someone who forgot the first one existed. A product page and a blog post end up covering the same buying question. Old campaign landing pages stay live and overlap with newer ones. Category pages and the detail pages underneath them start competing for the same phrase. You can spot it by searching your own domain plus the keyword and seeing how many of your own pages show up, or by watching whether one query in your search console keeps flipping between two different URLs — that flip is the tell.
How you solve it
Start with an inventory: list every page that targets the same term or answers the same question. Pick one as the page of record. Fold the others into it — move their useful content over, then 301-redirect the old URLs so the authority they built transfers rather than evaporates. Where two pages genuinely need to stay separate, retarget one of them to a distinct, related keyword so they stop overlapping. A canonical tag can help when near-duplicate content has to exist for other reasons, but it's a patch, not a fix — internal linking is what actually does the work, because consistently linking to the one page you want to win tells both search engines and AI crawlers which address is the authoritative one.
Relation to AI recommendations
Cannibalization is more costly for AI visibility than for classic search, because language models don't return ten blue links — most cite one, maybe a handful, of sources per answer. If three of your own pages half-cover the same question, none of them reads as the definitive one, and the model may cite a competitor's single, unambiguous page instead. Note that AI citation and Google ranking aren't the same contest: research from Ahrefs found only 6-8% of URLs ChatGPT cites overlap with Google's top 10 for the same query, and roughly 80% of ChatGPT-cited URLs don't rank in Google's top 100 at all — so a page that resolves your internal cannibalization can gain citation ground even before it moves in classic search. What consistently correlates with getting cited is being the clear, mentioned source on a topic: Ahrefs' analysis of roughly 75,000 brands found web-mention frequency correlates with AI citation rate at about 0.664, roughly three times the strength of backlinks. A single unambiguous page is easier to mention, link to, and cite than three competing ones.
Example
An online bike retailer publishes three articles: "Buying an e-bike," "Finding the right e-bike," and "E-bike buying advice." All three answer the same underlying question. Instead of one strong guide, three mediocre ones split the traffic and take rankings away from each other — in Google the top spot keeps alternating between them, and none reaches page one. Ask ChatGPT which e-bike buying guide to trust and it has no single clear source to point to, so it names a competitor instead. The fix: merge the three articles into one comprehensive guide, redirect the old URLs to it, and link to it consistently from elsewhere on the site.
Common questions
How do I quickly recognize keyword cannibalization?
Search "site:yourdomain.com keyword" and see how many of your own pages show up for the same term, or check your search console data for a query where the ranking URL keeps switching between two or more of your addresses. That switching is the clearest sign — a stable single ranking page usually means no cannibalization.
Is duplicate content the same as cannibalization?
No. Duplicate content means near-identical text living at multiple addresses. Cannibalization means multiple pages compete for the same search intent even when the wording is completely different. The two problems often show up together, but fixing one doesn't automatically fix the other.