gsc-cannibalization-check

Cannibalisation check

1. Run the analysis

gsc_cannibalization with period="last_3_months". Three months beats 28 days here: cannibalisation is a structural problem and a short window hides the queries where the two pages simply alternated.

Useful arguments:

• min_pages (default 2) — raise it to 3 to see only the worst cases.
• min_share (default 0.10) — the share of a query's impressions a page must hold to count as competing. Lower it on large sites where the split is wide.
• row_limit — default 50.

Check the meta line: if truncated=true the property exceeded the 25,000 pairs the API returns in one page, so the tail was not analysed. Say so rather than implying the list is complete.

2. Read the result

Each row is one query with the competing pages, their clicks and their positions. The pattern matters more than the count:

• One page far ahead, others trailing — usually harmless. Google picked a winner. Leave it alone unless the trailing pages are near-duplicates.
• Two pages close together, both mid-page — the real problem. The signals are split and neither ranks as well as one consolidated page would.
• Different intents on the same words — not cannibalisation. A category page and a guide can legitimately both appear.

3. Recommend

For each genuine case, name the page that should win and why (more clicks, better position, better intent match), then say what to do with the others: canonical to the winner, merge the content, redirect, or differentiate the targeting.

Do not recommend deleting a page that still earns clicks without saying where that traffic will go.

4. Verify later

After the fix, gsc_inspect_url on the losing pages confirms whether Google honours the declared canonical. That check is worth scheduling two to three weeks out, not the next day.