Cannibalization (similarity 0.9–0.99)

Embeddings
CannibalizationKeyword Cannibalization
Cannibalization (similarity 0.9–0.99) is when pages compete for the same queries, detected when semantic analysis shows 90-99% content similarity.

Cannibalization occurs when two or more pages on a site compete for the same Google queries, weakening each other. Embeddings allow automatic detection: pages with cosine similarity 0.9–0.99 are cannibalization candidates — for example, 'What is SEO' and 'SEO Basics' may target the same query.

The solution is content consolidation (merging into one stronger article) or differentiating intent (rewriting one article to target a more specific angle). Unlike duplicates (similarity 1.0), cannibalization requires search intent analysis, not simple removal. Without embeddings, cannibalization is difficult to detect on large sites (500+ pages), but with them — it's a matter of a few minutes in Google Colab.

In practice, after detecting cannibalization, check Google Search Console to see which cannibalizing pages generate more clicks — that should become the target page.

Source: AI Semantic SEO Expert, Robert Niechciał (sensai.io)