Updated
Keyword cannibalization is when two or more pages on your site compete for the same search, so Google splits impressions between them or switches back and forth, and often neither ranks as well as one strong page would. The free way to check for it is your Search Console data: export queries and pages together, and find searches where two of your URLs each get a real share of the impressions. Then fix only the conflicts that hurt, by merging, re-targeting or linking.
What a keyword cannibalization checker should look at
Many keyword cannibalization tools compare the keywords you “target” or crawl your titles for duplicates. That finds pages that might compete. Search Console shows pages that do compete: which of your URLs Google actually showed for each search, how often, and at what position. That is the data a checker should use, and you can read it yourself.
How to check for keyword cannibalization in Search Console
For one search
- Open Performance > Search results and pick the last 3 months.
- In the Queries tab, click the search you care about. This adds a query filter.
- Switch to the Pages tab. Every URL listed was shown for that search. Turn on impressions and position.
- If two or more URLs each have a good share of the impressions, look closer. One URL with 95% and another with 5% is not a conflict.
For the whole site
- Export query and page together through the Search Analytics API, the Looker Studio connector or the bulk export to BigQuery. The Performance report’s export stops at 1,000 rows (Google’s note on limits).
- Clean the URLs. Strip
#fragmentsand tracking parameters (utm_,gclid,fbclid) so one page isn’t counted as two. - For each query, add up impressions per URL and each URL’s share of the query’s total.
- Keep queries where at least two URLs have 10% or more of the impressions, and the query has at least 20 impressions in total. These are the thresholds DirectoryLens uses; they filter out one-off appearances.
- Drop brand searches and
site:queries. Several of your pages showing for your own name is normal. - Rank by impressions and evenness. An even 50/50 split on a big search matters more than 85/15 on a small one. DirectoryLens scores the split with Shannon entropy: 1 bit is an even split between two pages.
Then check one more thing per conflict: which page Google preferred each day. If the top URL flips every few days, Google can’t decide, which is the clearest sign the pages compete. If one page has led for months and the other shows up occasionally, it is a smaller issue.
When it is not a problem
- Both pages rank well. Two of your pages at 2 and 4 take two spots on the page. Leave them.
- Different intents. A guide on “how to choose a plumber in Austin” and the
/plumbers/austin-txcategory page can both show for broad searches; one serves research, one serves hiring. Make sure each is clearly about its own job. - Brand searches. Your home page, about page and contact page all showing for your name is expected.
- A listing and the page that lists it. On a directory, a search for “joe’s plumbing austin” showing both
/listing/joes-plumbingand/plumbers/austin-txis normal and usually fine. - Tiny numbers. A search with 25 impressions split between two pages isn’t worth a redirect.
How to fix keyword cannibalization
| Situation | Fix | Why |
|---|---|---|
| Two pages say the same thing | Merge into the stronger one and 301 redirect the other | One page keeps the content, links and history. See Google on redirects. |
| Both pages should exist but overlap | Re-target: change one page’s title, H1 and content to its own search | Each page then has a clear job Google can match. |
| Variants of one page (parameters, sort orders, print versions) | rel=canonical to the main URL | Tells Google which version to show; see consolidating duplicate URLs. |
| The wrong page wins | Point internal links and anchor text at the page you want | Internal links are your strongest signal of which page matters for a topic. |
| Page 2 of a list competes with page 1 | Keep pagination crawlable, make page 1 the strongest, don’t canonicalize page 2 to page 1 | Page 2 has different items; see pagination SEO. |
Avoid noindexing one of the pages as a quick fix if it has links or traffic: you lose what it had without passing it on. And don’t delete a page without a redirect.
After the fix, check again in a few weeks: the query should show one URL with most of the impressions, and its position should settle. Track the clicks of the surviving page, not each URL on its own.
On a directory: the usual conflicts
| Conflict | Example | Usual fix |
|---|---|---|
| City hub vs category-in-city | /austin-tx and /plumbers/austin-tx for “plumbers austin” | Link the hub to the category page with “Plumbers in Austin” anchor text; keep the hub about the city as a whole. |
| Two URLs for one place and category | /plumbers/austin-tx and /austin/plumbers | 301 one to the other and fix the links that create the second. |
| Category page vs category-in-city | /plumbers and /plumbers/austin-tx for “plumbers austin” | Usually resolves itself as the city page gains links; link the national category page to its top cities. |
| Filter or sort URL vs clean URL | /plumbers/austin-tx?sort=rating and /plumbers/austin-tx | Canonical to the clean URL, or stop generating crawlable sort links. |
| Page 2 vs page 1 | /plumbers/austin-tx/page/2 and /plumbers/austin-tx | Make page 1 the stronger page; see pagination. |
| Blog post vs directory page | “Best plumbers in Austin” post and /plumbers/austin-tx | Re-target the post to a research question, or merge it into the category page. |
| Overlapping categories | /plumbers/austin-tx and /plumbing-contractors/austin-tx | Merge the categories, or make one clearly about a different service. |
Directories create these conflicts by design: every category, place and filter makes a URL. That is also why doorway pages and index bloat come up on the same sites.
Keyword cannibalization vs duplicate content
Duplicate content is two URLs with the same or nearly the same content, such as a page reachable with and without a parameter. Google usually picks one as canonical and drops the other from results, and Search Console reports it in the Page indexing report. Keyword cannibalization is wider: the pages can be quite different, both indexed, and still compete for one search. Duplicates are fixed with canonicals and redirects; cannibalization often needs a decision about which page owns which search. See duplicate without user-selected canonical for the duplicate case.
How to prevent it
- One URL per place and category. Decide the pattern (
/plumbers/austin-txor/austin-tx/plumbers) and redirect the other everywhere it can appear. - Give each page type its own job. City hubs cover the city across categories; category-in-city pages own “[service] [city]” searches; guides answer questions. Write that down so new templates follow it.
- Check before publishing. Before adding a page, search your own Search Console data for its main search. If an existing page already ranks for it, improve that page instead.
- Keep filters and sort orders out of the index unless a filter has its own real demand.
- Merge categories that mean the same thing (plumbers and plumbing contractors) before they each grow a city page in every city.
- Recheck monthly. New conflicts appear as pages are added and rankings move.
Cannibalization and the other quick wins
Competing pages often show up as striking distance keywords: two URLs at 9 and 14 for one search, neither climbing. They can also look like content decay, when a new page takes impressions from an old one. Check for a second URL before you rewrite a page that seems stuck or slipping. For the full order of checks, see SEO quick wins.
How DirectoryLens checks it
Competing pages (under Advanced) runs the method above on your full Search Console data for the dates you pick: every query where two or more of your URLs each get at least 10% of at least 20 impressions, after removing fragments and tracking parameters, with site: queries left out. Each conflict shows its share and position per page, how evenly it is split, and a likely reason: a parameter URL against a clean one, page 2 against page 1, two pages for the same place and category, a city hub against a category-in-city page, or a guide against a directory page.
Brand searches, and business-name searches showing a listing plus a page that lists it, are marked as expected and hidden unless you ask for them. Click a conflict to see which page Google preferred each day over the last 90 days.