Keyword Cannibalization Tool

Each row must connect a query to its associated page, with clicks, impressions and average position for the same reporting period. Daily mode also needs a date on every row. Keep those relationships intact when preparing your file.

The separate Queries and Pages tabs do not provide a reliable way to reconstruct query-page pairs. A query total does not tell you how its impressions were divided among pages, and a page total does not identify the queries behind it. Joining those separate totals would introduce relationships the export never observed.

Record the source reporting period and search type. Preserve country and device dimensions when available. Search Console uses three-letter country codes and desktop, mobile or tablet device labels. When a dimension is absent from the export, leave it blank throughout rather than adding a guessed value.

Use the SEO tools hub to choose another workflow when your evidence answers a different question. Content similarity, keyword planning and technical indexability checks complement performance analysis; each needs its own inputs and interpretation.

Enter the website and choose domain scope, including subdomains, or exact hostname scope. The distinction matters when a blog, store or regional host should be reviewed separately. Check that the imported page addresses belong to the intended scope.

Select the same start date, end date and search type as the export. The form provides Web, Image, Video and News choices. Keep unlike search types in separate reviews so their metrics retain a clear context.

Choose date-range totals for records already aggregated across the reporting window. Choose daily records when every row has its observation date. A date-range total cannot reveal when a page led or how often the leading page changed.

Identify export coverage as accurately as possible: unknown, all returned rows, a filtered export or a sample. “All returned rows” describes what the source returned, not a complete census of every query or page. Privacy omissions and source limits can still leave gaps.

Do not mix all-device totals with their mobile breakdown, or all-country totals with country-specific rows. Those are overlapping observations. Aggregate omitted dimensions upstream where necessary so the same evidence is not counted twice.

1. Prepare the source. Confirm that query, page, clicks, impressions and average position appear together on each row.
2. Set the context. Enter the website, URL scope, reporting dates, record type, search type and coverage.
3. Add the records. Upload CSV or TSV, or paste the paired records into the input area.
4. Read the records. Review the parsed count and the column assignments before requesting the overlap review.
5. Map the dimensions. Assign the date column for daily records and country/device columns when the export includes them.
6. Complete verification. Complete the configured human verification. Built-in codes expire after ten minutes and work once; request a replacement when required.
7. Review the evidence. Inspect the relevant query segment and its page metrics before selecting work.
8. Keep the context. Download the evidence you need and retain the original export with your project notes.

Check dates and mappings when a file behaves unexpectedly. A row-count message is useful import feedback, but it does not confirm that a completed overlap report is correct. The public Example is illustrative evidence for learning the workflow, not measured results from a campaign.

A multi-page segment contains the same normalized query with positive impressions for more than one URL within the same country and device selection. That is a signal to inspect the pages, not a declaration that they damage one another.

A single observed page means the supplied records show one page with the relevant evidence. It does not establish that no other URL has appeared for the query. A filtered file, privacy omission or source limit may hide another page.

Open the underlying pages and ask whether they serve the same need. A product page and a practical guide may both help someone using a broad query. Two interchangeable guides may need a different review. Their presence in the same segment does not settle that editorial question.

Use the Search Intent Checker when a supplied search snapshot can help clarify the expected page format. Compare the actual results and page purposes before deciding that one URL should replace another.

The first page in a segment has the most imported impressions. That does not automatically make it your preferred destination, highest-converting page or most useful answer. Treat its ordering as a description of the supplied evidence.

Average position is weighted by each row’s impressions. It remains an average across observations, not a single live rank. When combining daily records, a day with many impressions contributes more than a day with very few.

The impression-share column divides a page’s imported impressions by the sum of imported page-level impressions in that query segment. It is not a share of all searches, unique people or the entire market. It should not be compared directly with a property-level Search Console total, which uses a different aggregation basis.

Keep small samples in perspective. A handful of impressions may justify reading two pages, but it is a weak basis for a major content change. There is no universal impression percentage that proves harmful cannibalization.

Daily mode identifies the page with the most impressions on each observed day. It counts a change only between consecutive days with unique leaders. A tie or missing day interrupts that comparison.

For example, if two pages tie on Tuesday, a different leader on Wednesday does not create a measured Monday-to-Wednesday switch across the tie. If Thursday is missing, do not connect Wednesday and Friday as though they were adjacent observed dates.

A changing impression leader is not necessarily a ranking-position change. One page can attract more impressions because demand or audience mix changed. A lower average position can also coexist with more impressions. Keep the leader definition visible when discussing daily changes with colleagues.

Date-range totals leave this daily history unmeasured. They can still show multiple pages receiving impressions, but they cannot tell you when the lead changed. Do not label an unmeasured history as zero changes or stable ownership.

Read both pages and compare their purposes, broader query coverage and business value. Determine whether the overlap reflects two useful answers, unclear positioning or genuinely interchangeable content.

Where purposes differ, make them easy to understand. Review titles, introductions, navigation and internal links so visitors can choose the right resource. Removing a shared phrase from every secondary page is not a substitute for a clear information structure.

Record your preferred page and the reasons for that choice in the Keyword Mapping Tool. The mapping records a future plan; the performance export records past observations. Keep the decision separate from the page that happened to receive the most impressions.

If consolidation seems appropriate, examine useful sections, conversions, backlinks and other query traffic before planning it. A URL that contributes little to one query may be valuable elsewhere. This report alone does not justify deleting, redirecting or noindexing it.

The published URL rules keep schemes, paths, query strings and trailing slashes distinct. Fragments are removed and reported. Different addresses may therefore remain separate evidence even if you suspect they lead to similar content.

Use the Canonical Tag Checker to inspect canonical declarations when duplicate URL handling is part of the question. A declared preference is different from Google’s selected canonical; the overlap report does not establish either one by itself.

For suspected crawl or indexing restrictions, use the Indexability Checker. A missing performance row cannot diagnose a robots rule, indexing decision or HTTP problem.

When your plan calls for clearer site navigation, the Internal Link Opportunity Finder can support a separate review of linking opportunities. Check page relevance and the visitor’s next step before adding a link. The overlap data does not determine an ideal link count.

Duplicate query-page records for the same date and segment are rejected to prevent counting an export twice. If additional dimensions were omitted during preparation, aggregate them correctly upstream instead of deleting one of several legitimate contributions.

Zero-impression rows remain visible in the report and full input exports but do not create overlap. A second URL with zero impressions is different from a second URL with positive evidence. An absent row is also different from a measured zero.

Search Console does not expose every query. Privacy protections, filters and export limits affect the observed set. Keep the coverage declaration and reporting context with your conclusions so another reviewer can assess what the evidence actually supports.

Queries are normalized to Unicode NFC, lowercase and consistent spacing. The report exposes that key and up to 20 original variants per segment; the full input evidence retains each original query. It does not merge semantic synonyms or infer search intent.

Original URL text is retained alongside the normalized address. Fragments are removed and counted. Schemes, paths, query strings and trailing slashes remain distinct. The server validates site scope; it does not crawl addresses or resolve redirects.

Filtered pages CSV includes all matching query segments, across result pagination. Full input CSV preserves every accepted record, including zero-impression rows. Full report JSON contains the complete report and timelines. TXT creates a concise full-report review brief. Admin controls can disable exports or the TXT brief.

CSV neutralizes spreadsheet formula-like text. Use JSON when you need the original text without CSV safety prefixes. Save your report before reset or leaving this page; the tool has no saved project history.

Downloads retain reporting context for a later review. Keep that context with the original source file, and record any filters or preparation applied before import. Download what you need before clearing evidence.

The plugin does not save imports or reports as a report history. Records are sent to this website for analysis; processing is not browser-only. Shared abuse protection records a salted IP hash, tool identifier and usage date. Hosting and security-provider policies apply separately. This does not establish that all processing happens in your browser or that the surrounding website records no information.

The tool does not fetch your pages, infer search intent or change content, canonicals, redirects or indexing settings. A completed evidence review remains a planning step; any website changes require a separate, deliberate decision.

No. Multiple useful results can coexist. Compare page intent, wider query coverage, conversions and backlinks before deciding that overlap is harmful. This report does not prove lost traffic.

No. Each row needs the query and its associated page. Separate aggregate totals cannot reconstruct those pairs. Prepare a query-page export through the Search Console API or a compatible connector.

No. It summarizes observations from the reporting period. The tool weights supplied row averages by impressions. Neither that calculation nor the impression leader is a live ranking lookup.

Not necessarily. Check the eligible consecutive-day pairs. Missing days, ties and zero-only days break comparisons. Date-range totals do not measure daily changes at all.

No. It produces evidence for review and does not fetch or modify your website. Content edits, redirects, canonicals and indexing changes require a separate decision.

Filtered pages CSV contains every segment matching the display filters, across pagination. Full input CSV, full report JSON and the TXT review brief use the whole report. Downloads depend on the administrator settings. CSV neutralizes formula-like text; JSON preserves original text.