Keyword cannibalization happens when two or more pages on your site target the same query closely enough that search engines can't decide which to rank. The result: split signals, lower position for both, and a confused ranking that flips between the candidates over time.
How it usually arises:
- A blog post and a product page both target "best widget" because nobody coordinated
- A glossary entry and a long-form guide both cover the same definition
- Old content was never deprecated when newer content shipped, leaving stale + fresh versions of the same topic live
The classic SEO advice — "one page per query" — exists because of this.
Why it matters for AEO
AI engines deal with cannibalization differently than search engines: instead of picking one, they sometimes cite NEITHER. Faced with two pages of similar authority on the same topic from the same domain, the LLM has no clear signal which is canonical — and may default to citing a competitor's single clean page instead.
The asymmetry: a competitor with one strong page beats your two okay pages on the same topic, even if your combined depth would beat theirs.
What b/cited does about it
- Ownership analysis flags clusters where multiple URLs on your domain rank for related queries — the cannibalization signal
- Briefs suggest consolidation or canonicalization when overlap is detected: pick one canonical URL, redirect or merge the rest, update internal links
- Audit grade counts cannibalization as a structural finding (vs a content finding)
The fix is usually painful but cheap-relative-to-the-loss: merge the strongest content into one URL, 301-redirect the others, repoint internal links. Recovery typically lands within 30 days of consolidation.