Free Tool

Keyword cannibalization checker

Paste or drop Search Console data with one row per query and page. See the queries where two or more of your pages get impressions or clicks, with each page's share. The file stays in your browser.

Your Search Console data

Stays in your browser. Your file is read and checked on this device. This page sends no part of it to any server.

You can also drop a CSV, TSV or JSON file on the box.

Needs one row per query and page, with impressions. Reads up to ⁦200,000⁩ rows and ⁦20 MB⁩. Rows past the limit are skipped and the result says so.

Point links at the page you want

Found two pages on one query? Check which internal links point at each one, and with what anchor text.

Open the internal link checker

Check what the searcher wants

Two pages can rank for one query because they answer different needs. Sort the queries by intent first.

Open the keyword intent classifier

Get query and page pairs

The checker needs one row per query and page. Here is what Google says about each way to get that.

The Performance report export does not have pairs

In the Performance report, the table is grouped by one dimension at a time: queries, pages, countries and so on. The export holds the chart and that table, so a query row has no page and a page row has no query. The export also stops at 1,000 rows.

Google says:

Google says:
"The dimension defines how the data in the table is grouped." Google
Google says:
"Both chart and table data are exported." Google
Google says:
"The maximum you can export through the Search Console user interface is 1,000 rows of data." Google

Search Console API

Call searchanalytics.query with the dimensions query and page. Google's reference page has a Try it now panel that runs the call in your browser after you sign in. Paste the JSON response here as it is. For more than 25,000 rows, page with startRow and join the rows into one CSV.

POST https://www.googleapis.com/webmasters/v3/sites/sc-domain%3Aexample.com/searchAnalytics/query
{
  "startDate": "2026-08-01",
  "endDate": "2026-10-31",
  "dimensions": ["query", "page"],
  "type": "web",
  "rowLimit": 25000,
  "startRow": 0
}

Google says:

Google says:
"The grouping dimension values are combined to create a unique key for each result row." Google
Google says:
"The default returned by the API will be 1,000 rows, but you can use rowLimit to increase it to 25,000 and startRow to pull the rows 25,001 to 50,000 using pagination." Google
Google says:
"When you group by page and/or query, our system may drop some data in order to be able to calculate results in a reasonable time using a reasonable amount of computing resources." Google
Google says:
"Data is typically available after 2-3 days" Google

Data Studio, formerly Looker Studio

Add a Search Console data source and pick the URL Impression table. Build a table chart with the query and landing page dimensions plus impressions, clicks and average position. Export the chart as CSV. Some queries come as one blank row, which the checker skips.

Google says:

Google says:
"Looker Studio is now called Data Studio." Google
Google says:
"In the Tables panel, choose either Site Impression or URL Impression." Google
Google says:
"In this case, Data Studio may aggregate these queries into a single row with a blank value." Google

BigQuery bulk data export

If your property has the bulk data export turned on, run this query on the searchdata_url_impression table and download the result as CSV. It sums the daily rows and turns sum_position into an average position the way Google describes.

SELECT
  query,
  url AS page,
  SUM(impressions) AS impressions,
  SUM(clicks) AS clicks,
  SUM(sum_position) / SUM(impressions) + 1 AS position
FROM `your_project.searchconsole.searchdata_url_impression`
WHERE data_date >= DATE_SUB(CURRENT_DATE(), INTERVAL 3 MONTH)
  AND search_type = 'web'
  AND NOT is_anonymized_query
GROUP BY query, url
HAVING SUM(impressions) > 0

Google says:

Google says:
"searchdata_url_impression: Contains performance data for your property aggregated by URL." Google
Google says:
"To calculate average position (which is 1-based), calculate SUM(sum_position)/SUM(impressions) + 1." Google
Google says:
"Using the bulk data export feature, you'll see all the performance data available to Search Console for your property, with the exception of anonymized queries." Google
1

Get query and page pairs

Pull them from the Search Console API, a Data Studio URL Impression table or BigQuery. The steps are below.

2

Paste or drop the file

CSV, TSV or the API JSON. The checker finds the separator and guesses the columns from the header row.

3

Set the thresholds

Minimum share per page, minimum impressions per query, brand words to leave out. The list updates as you type.

4

Open each query

See every page with its impressions, share, clicks, CTR and position. Download the list as CSV.

When two of your pages show for the same query, impressions and clicks are split between them. Often that is fine. Sometimes one page would do better alone. This list shows you where to look.

Two pages on one query is not always a problem

Google does not use the word cannibalization in the pages cited below. The checker shows data, not a verdict. Before you merge or redirect anything, look at these:

  • Brand and home page queries. Your home page and a deep page often both show for your brand name. That is normal.
  • Different intent. A guide and a product page can both fit a query when some searchers want to learn and others want to buy.
  • The trend. Compare two date ranges. A problem shows when the pages swap places and neither one holds a steady position.
  • Which page you want to rank. If it is the one with fewer impressions, point internal links and anchor text at it.
  • Real overlap. If two pages answer the same question for the same reader, consider one stronger page and a redirect from the other.

How the checker reads your file

  1. The separator is found from the header row: comma, tab or semicolon. Quoted fields can hold commas, quotes and line breaks.
  2. Columns are matched by header name, in English and in the French, German and Spanish names Google's help uses. You can change every match.
  3. Numbers can use 1,234 or 1 234 for thousands and 0,4 for decimals. A value of 1,234 with three digits after the comma is read as one thousand two hundred thirty four.
  4. Queries are trimmed, lowercased and joined on single spaces. The spelling with the most impressions is shown.
  5. Rows with the same query and page are added up, for example one row per day, country or device. Position is averaged by impressions.
  6. A page's share is its impressions divided by the impressions of all pages listed for that query. Google counts page data per page, so one search that shows two of your pages counts once for each.
  7. A query is listed when it has at least the minimum impressions and either 2 or more pages each reach the minimum share, or 2 or more pages got clicks.
  8. Rows with an empty query are skipped. Some exports put anonymized queries in one blank row.
  9. Up to ⁦200,000⁩ rows and ⁦20 MB⁩ are read. Nothing is uploaded.

Google says:

Google says:
"Data on the chart is aggregated by property. For example, if two results from the same site appear for one query, they count as a single impression in the chart total." Google

Sources

All pages read on ⁦2026-11-12⁩. The quotes on this page are copied from them.

Worked example on sample data

The sample file is made up: a fictional coffee shop on example.com, 27 rows. It is not real Search Console data. With the default settings and brewlab as the brand word, the checker lists ⁦7⁩ queries. These are the top three by impressions:

  1. ⁦espresso grinder⁩: ⁦3⁩ pages, ⁦3,930⁩ impressions
  2. ⁦how to descale an espresso machine⁩: ⁦3⁩ pages, ⁦3,150⁩ impressions
  3. ⁦best espresso grinder⁩: ⁦2⁩ pages, ⁦1,725⁩ impressions

The blog post holds 54.5% of the espresso grinder impressions and the product page 37.7%. Whether that is a problem depends on intent: a buyer may want the product page, a researcher the guide.

Download the sample CSV

Frequently Asked Questions

What is keyword cannibalization?

It is an SEO term for two or more pages on one site showing for the same query, so impressions and clicks are split. Google does not define the term in its Search Console help. This checker shows where it happens in your data and leaves the call to you.

Can I use the export from the Performance report?

No. That export groups each table by one dimension, so you get queries and pages in separate tables with no link between them. It also stops at 1,000 rows. Use the Search Console API with the query and page dimensions, a Data Studio URL Impression table or the BigQuery bulk data export.

Is my Search Console data uploaded?

No. The file is read with your browser's file API and checked on your device. This page has no server step for it, and the tool area is excluded from analytics capture.

Why do the shares not match the totals in Search Console?

Google counts page data per page. If one search shows two of your pages, each page gets an impression. The query total in the chart counts that search once. Google also leaves out anonymized queries and can drop rows when you ask for query and page together.

What should I do with a query that shows two pages?

Look first. Check the intent, the trend over two date ranges and which page you want to rank. Often it is fine. If the pages really answer the same question, point internal links at the stronger one or merge them and redirect.

Want us to fix this for you?

We read your Search Console data, decide which page should own each query and fix the internal links, content and redirects.