Guides

How to Find Cannibalizing Pages in Search Console Without a Paid Tool

Read Time7 Min Read
Length1,613 Words
CategoryGuides
How to Find Cannibalizing Pages in Search Console Without a Paid Tool
AI Generated

Search Console has no cannibalization report. The official path is Performance, then Search results: open Queries, click the query, and open the Pages tab. Google’s own help text is direct: view the Pages tab to see which URLs were shown. That tab, not a paid dashboard, is where a real split becomes visible.

The chart can hide the problem you came to find. Performance data is aggregated by property, so if two results from the same site appear for one query, they count as a single impression in the chart total. Query-grouped totals and page-grouped totals can disagree for the same reason. Multiple URLs on one query are a candidate, not a verdict. John Mueller has said three pages in one result are not a problem just because there is more than one, and pages are not duplicates merely because they appear together.

This check stays fail-closed. Use at least the default three-month window, drop branded queries, and log a split only when impressions are material, positions compete or swap, and the intent is the same. A deep, low-impression second URL is noise. Anchors, parameters, and canonical duplicates are a different problem, and nothing in this audit is a reason to merge yet.

Summary
  • Search Console has no cannibalization report; the Pages tab shows which URLs were shown for one query.
  • Chart totals can hide a split because two results from the same site count as one impression.
  • More than one URL is a candidate, not proof of cannibalization or duplicate content.
  • Confirm only material, same-intent splits over at least the default three months, and skip branded queries.
  • The UI table stops at about 1,000 rows and one dimension, so a full query-and-page join needs the free Search Console API.

How to see which pages rank for one query

Search Console Pages tab showing which pages rank for one query

Search Console has no keyword cannibalization report, and no threshold that says merge. To see which of your pages were shown for one query, open Performance, then Search results, then Queries. Click the query and open the Pages tab. Google’s own help states it directly: view the Pages tab to see which URLs were shown.

Run that path on the exact query you might brief next, before you write. Custom regex filters, available since April 2021, can group close query variants in one view so a near-duplicate search is not missed. Stop at the Pages list. Do not treat the chart total as the list of competing URLs: query-grouped data is aggregated by property, so two URLs on one results page can count as a single impression in the chart. Only the Pages tab separates them.

Key Takeaway

The path — There is no cannibalization report. Click the query, open Pages, and read the URL list — not the chart total.

Sources

Is it bad if two pages show for the same keyword?

Keyword cannibalization check separating a real same-intent split from noise

No — not by itself. Two of your pages on the same keyword are a candidate, not a verdict. John Mueller said three pages in one result are not a problem merely because there is more than one, and Search Console only lists pages that were actually shown.

Those pages are not duplicates just because they appear together. A hand review of 80 multi-ranking keywords, drawn from about 9,700 cases, found only one that needed action.

1 of 80
Multi-ranking keywords that needed action
0.2
Flip rate above which a split looks real
90%
Click share with no flips: a double listing

Confirm a split only when it is material. A practical gate is a flip rate above 0.2 and a top click share below 70 percent. A 90 percent click share with no flips is a double listing, not cannibalization.

A deep, low-impression second URL is noise. Average position is the average of the topmost result, not a permanent rank for every URL, so a position gap alone is not proof.

Require the same intent. Pages that do different jobs stay separate. Name same text versus same intent using the existing cannibalization-versus-duplicate-content guide, and do not merge from this screen.

Key Takeaway

Not a verdict — Two shown URLs are cannibalization only when the split is material — flip rate above 0.2, top click share under 70 percent — and both pages serve the same intent.

Cannibalization cutoffs vs. real fixes

Sources

Which date range to use, how to spot a URL swap, and why branded queries don't count

Search Console date range and URL swap check for non-branded queries

Do not log a split from that Pages list until the window is long enough to show a pattern. Use at least the default three-month Performance view of clicks and impressions. A short range turns a one-off appearance into a false conflict.

To see whether Google is swapping which URL ranks, compare two periods, or open the same query filtered to each URL, and note whether the leading URL flips. Record that flip as swap evidence, not as an automatic merge.

Ignore branded queries. Homepage, About, and contact pages are supposed to show together for the brand name, so co-appearance there is not cannibalization.

The branded and non-branded filter has data from 11 March 2025, with a 16-month history from that introduction. It is unavailable on low-impression sites and on URL-prefix sub-properties. If the filter is missing, exclude brand terms by hand before you log anything.

Key Takeaway

Log only this — A three-month window, a documented leading-URL flip, and non-branded queries are what turn a Pages list into a loggable split — not a one-week blip or a brand query the homepage is meant to own.

Sources

Why chart totals and the Pages table don't match

Search Console chart impressions splitting into two Pages table rows

Once branded queries are out and the window is long enough, the chart can still look like one clean ranking. Chart data is aggregated by property, so two results from the same site on one query count as a single impression. A split can hide under that single number.

The table will not agree, and that is expected. Data grouped by Queries is aggregated by property, while the Pages tab is aggregated by page — the view that separates the URLs.

Some queries are omitted for privacy but still count in chart totals unless a query filter is applied. Grouping by page and query can also drop rows so the result can be computed with limited resources. Neither the gap nor a missing row is a hidden cannibal, or proof the site is clean.

Decide from the filtered Pages rows, not from the chart number.

Key Takeaway

Judge the Pages rows — Property-level aggregation hides a split in the chart; privacy omissions and dropped page-and-query rows explain the rest. A split is real only when the filtered Pages rows show it.

Sources

How to export past 1,000 rows — and why query groups are not the report

Exporting Search Console query and page data past 1,000 rows

The free way past that wall is not another Search Console screen. The Performance table cannot list query and page side by side and shows at most 1,000 rows, and a UI export hits the same wall. Query groups do not show competing pages either.

For a one-person site, stay in the interface on the queries you might actually brief. Use a free pull only when that cap hides the query-and-page join.

The Search Analytics API returns at most 25,000 rows per response and exposes a maximum of 50,000 rows per day per search type. BigQuery bulk export is the other free bulk door. Do not buy a cannibalization tool for this step, and do not build a pipeline just to see the join.

That list is still partial. Anonymized queries and page or query grouping drop rows. Query groups, added to Search Console Insights in October 2025, cluster similar queries. They do not show when two pages compete, so they are not a cannibalization report.

Key Takeaway

Past the cap — Stay in the UI unless the 1,000-row cap hides the join; the free API or BigQuery can extend the list, but query groups still will not show two pages competing.

Sources

Canonical problem, or two real articles?

Cannibalization log separating a canonical problem from two real articles

A partial export or a query group still will not separate a canonical duplicate from two articles. Most performance data in this report is assigned to the page's canonical URL, not to a duplicate URL, so a click on the duplicate does not create a second Pages row. If an anchor, a parameter, or a near-duplicate path still shows, fix canonicals before you call those URLs two articles.

Two real articles are two distinct URLs that each do a job. Same intent plus a material split is the only row that leaves this worksheet. Identical text is a duplicate-content problem: name the existing guide on the brief, and do not merge it from this log.

01
Log the row before you act
Record the query, both URLs, clicks, impressions, positions, click share, swap evidence, and intent.
02
Hold a blank intent
If intent is empty, the row stays. Do not merge from this screen.
03
Hand off only a confirmed split
Send same-intent material splits to the content cannibalization audit, which chooses merge, redirect, or hold.

Recheck that query before the next brief on the topic is queued, and once more after the new URL has impressions inside the three-month window. Do not recheck daily.

Key Takeaway

Hold the merge — A second URL is two articles only when both do a job, share intent, and show a material split. Anchors, parameters, and canonical duplicates get fixed first; a blank intent stays on the log.

Key Takeaways

[01]
Pages viewSearch Console has no cannibalization report; the official check is Performance, then Search results, then Queries, then the query, then Pages, and the decision belongs on those filtered rows, not the property-level chart.
[02]
Split, not a double listingTwo URLs on one query are only a candidate. Log a split when both earn meaningful impressions, the flip rate is above 0.2, top click share is below 70%, and the intent is the same; a 90% click share with no flips is a double listing.
[03]
Window and brandUse at least the default three-month range, treat a leading-URL change across periods as swap evidence rather than an automatic merge, and exclude branded queries by hand if the branded filter is missing.
[04]
Chart mismatchChart totals count the property once, while Pages are page-aggregated, so two competing URLs can look like one impression; privacy-omitted queries and grouping can also drop rows.
[05]
Free export limitsThe UI export stops at 1,000 rows and cannot pair query and page; the Search Analytics API and BigQuery are the free bulk options, and query groups cluster similar queries without showing which pages compete.
[06]
Canonical versus two articlesMost Performance data is assigned to the canonical URL, so parameters, anchors, and near-duplicates are canonical fixes. Same-intent splits stay on the log until a separate audit chooses merge, redirect, or hold.

Before you brief another page on a topic you already cover, run this Search Console check on the exact non-branded query and log a split only when the Pages rows meet the flip, click-share, and intent tests.

Frequently Asked Questions

Does Google Search Console have a keyword cannibalization report?
No. The official path is Performance, then Search results: open Queries, click the query, and open the Pages tab to see which URLs were shown.
Why can the Performance chart hide two competing URLs?
Chart data is aggregated by property. If two results from the same site appear for one query, they count as a single impression in the chart total. The Pages tab is the view that separates them.
Are multiple pages ranking for one query always cannibalization?
No. John Mueller has said three pages in one result are not a problem just because there is more than one, and pages are not duplicates merely because they appear on the same results page. Confirm a material split and the same intent before you treat it as a conflict.
How long a date range should I use?
Use at least the default three months of Search clicks and impressions. A short range can turn a one-off appearance into a false conflict. Compare periods, or compare the two page filters, to see whether the leading URL flips.
Should branded queries be included in a cannibalization check?
Skip them. Homepage, About, and contact pages are supposed to show together. The branded and non-branded filter has data from 11 March 2025 and a 16-month history from that introduction, but it is unavailable on low-impression sites and URL-prefix sub-properties.
When do I need the Search Console API instead of the UI?
The Performance table cannot list query and page side by side and shows at most 1,000 rows. A full join needs the free Search Analytics API, which returns at most 25,000 rows per response and exposes a maximum of 50,000 rows per day per search type. Grouping by page and query can still drop rows.
Sources

You Might Also Like

View All Articles