Structural Cannibalisation.
The silent ranking killer.

What it actually is

Cannibalisation is not duplicate content. It is duplicate intent. Two pages can be 0% identical in their wording and still cannibalise — if they both answer the same buying question, they compete for the same ranking slot. Google will not rank both. It picks one and demotes the other.

The signature is unmistakable once you know what to look for:

• Two pages on the same site rank in positions 7 and 12 for the same query — instead of one ranking at 3
• Rankings oscillate between two pages week-to-week as Google flip-flops on which to favour
• Both pages have decent on-page SEO, but neither ranks where it “should”
• A previously top-ranking page drops after a new page on the same topic is published

Why standard SEO audits miss it

Every audit tool checks each page in isolation. Title tag, meta description, internal links, page speed. Cannibalisation is a relational problem — it lives between two pages, not inside either one. The audit tool sees both pages individually and both look fine.

That is why most sites have cannibalisation issues and have no idea they exist. The audit said the site was healthy. The rankings say otherwise.

The four ways IA resolves it

For every cannibalisation pair IA detects, the diagnosis includes a resolution plan from one of four canonical strategies:

1. Merge — consolidate the two pages into one stronger page, 301 the loser to the winner. Used when both pages are weak but their combined topical depth would justify a single strong page.
2. Differentiate — rewrite one page to target a clearly different intent. Used when both pages have unique value but accidentally overlap.
3. Redirect — 301 the weaker page to the stronger and recover the lost authority. Used when one page is clearly dominant and the other has no independent reason to exist.
4. Cluster & canonicalise — demote one page to cluster status, link it upward to the other as canonical. Used when both pages should exist but one is the pillar and the other is supporting.

How IA finds them

The IA engine reads every page on your site, builds an intent fingerprint for each one, and compares them pairwise. Pairs above a Jaccard similarity threshold or with overlapping ownership keys are flagged with a severity rating — critical, high, medium, low — and a recommended resolution. The five governance layers verify each resolution will not create a new cannibalisation downstream before the plan ships.

See the diagnosis pipeline →