Keyword cannibalization in WordPress: how to tell a real problem from a harmless one.

Most articles about keyword cannibalization start from the assumption that any two pages ranking for the same search are a problem, and end by telling you to merge them. That advice fixes some sites and quietly damages others. A better starting point is to ask what the two pages are each for, and what the numbers say is actually happening.

This guide is about telling those cases apart on a WordPress site: the habits that create competing pages, a step-by-step way to find them in Google Search Console, and a decision table for which ones to leave alone. The detailed fixes have their own guide on merging, redirecting and re-targeting competing posts, which this one refers to as “the fixes guide” from here on.

What keyword cannibalization is

Keyword cannibalization (also spelled “cannibalisation”) is when two or more pages on your site go after the same search, with the same purpose, so they compete with each other instead of one of them being the clear answer. The word is borrowed from retail, where a new product “cannibalises” sales of an old one. Here, what gets split is clicks, links and Google’s confidence about which of your pages to show.

Two things have to be true for it to be cannibalization rather than coverage:

• The same query. Both pages get impressions for the same search, in the same period, not just for vaguely related ones.
• The same intent. A reader typing that search would be equally happy with either page, because they answer the same need. This is the part most audits skip. Search intent is what the person wants, not the words they used.

A recipe blog with “Easy banana bread” and “Banana bread with oil instead of butter” has two pages that will both show for banana bread recipe. A reader who wants the classic version does not want the oil version, and the reverse. That is two answers to two needs that happen to share words. A travel site with “Lisbon in three days” and “Three-day Lisbon itinerary (updated)” has two answers to one need. Only the second is a problem.

Nobody is penalised for this; Google has never described a penalty for it, and its own staff talk about it as a question of focus, not of rules (more on that below). The cost is indirect: links and attention are split, the URL Google shows can keep changing, and neither page settles as high as one clear page might. That last part is the common industry view rather than something Google has documented, so treat it as a hypothesis to test against your own numbers.

What it isn’t: when two results are fine

Google’s own people have been openly sceptical of the term. In September 2025 John Mueller replied on Bluesky to a question about tracking cannibalization in Search Console. He pointed out that Search Console only reports pages that were actually shown, said “it’s not really ‘cannibalization’ if it’s theoretical”, and added that three of your pages in the same results is not a problem just because it is more than one. His advice was to cut unnecessary duplication and put the energy into one excellent page, while remembering that pages are not duplicates simply because they appear for the same search.

Parts of the SEO industry say the same. Ahrefs’ own guide to keyword cannibalization argues it is a problem only when overlapping pages demonstrably cost traffic, and quotes Patrick Stox dismissing the idea that several pages simply “confuse” Google. One reason to be careful: pages rank for far more than one keyword. In Ahrefs’ 2017 study of three million queries, the average number-one page also ranked in the top ten for nearly 1,000 other relevant keywords (the median was around 400). Merge two pages to win one query and you can lose some of the hundreds of others the weaker page was quietly earning.

That gives you three situations that look like cannibalization and usually are not:

• Two of your pages on page one, together. Google can show more than one result from a site. In 2019 it announced a site diversity change that generally limits a site to two listings in the top results, which tells you two is a normal, allowed outcome. If a hobby site on fountain pens has its buying guide at position 3 and its nib-size explainer at position 6, it has two shots at the click.
• Different intents behind one phrase. A WooCommerce outdoor-gear shop’s category page for trail running shoes (wants to browse and buy) and its blog post “How to choose trail running shoes” (wants to learn first) serve different moments. If the results for that search mix shops and buying guides, there is room for both.
• A page that ranks for thousands of queries. Any decent post will pick up impressions for long-tail searches that a neighbouring post owns. A handful of impressions at position 60 for someone else’s keyword is noise.

Symptoms worth taking seriously

These are the patterns that are worth your time, roughly from most to least telling.

1. The ranking URL flips. For one query, Google shows post A for a week or two, then post B, then A again. In the illustrative chart below, the query’s total impressions hardly change; what changes is which page gets them.
2. Impressions are split, and both pages sit lower than one used to. Say a personal-finance blog’s guide ranked around position 6 for sinking funds before a second post went up, and now the two share the impressions at positions 14 and 18. The before-and-after is what makes the case; a split on its own does not.
3. The wrong page ranks. A local plumber wants the “Boiler repair in Leeds” service page to show for boiler repair leeds, but Google keeps choosing a 2021 blog post about a boiler repair job, which has no booking form.
4. A low click-through rate on a decent position. If the URL Google picks has a title that does not match the search, people skip it. This is often the wrong-page symptom seen from another angle.

How WordPress creates competing pages

WordPress makes it easy to publish the same thing twice without meaning to. These are the usual sources, and most sites have at least three of them.

Where the switches are in Yoast and Rank Math

Most of the archive problems above are one setting each. The menu paths below are from each plugin’s current documentation; labels move between versions, so if yours differ, look for the same words nearby.

Sources for those paths: Yoast’s help pages for its Advanced settings and its Categories & tags settings, and Rank Math’s knowledge-base articles on Titles & Meta, noindexing URLs and its General Settings. Noindexing a tag archive is the right tool here rather than a redirect: the archive still exists for readers who click a tag, it just stops competing in search.

Three of the causes deserve a closer look.

Attachment pages

Older WordPress sites give every uploaded image its own page. In version 6.4, core disabled attachment pages for new installations, but sites that existed before the upgrade keep them switched on. If your site is older than late 2023, check: open any image in the media library and click through to its attachment URL. If it shows a page, turn them off with your SEO plugin’s setting or the wp_attachment_pages_enabled option.

Pagination

It is tempting to point the canonical of /page/2/ at page one. Google specifically says not to: its pagination guidance says don’t use the first page of a paginated sequence as the canonical, and to give each page its own. If archive pages are outranking your posts, the archive is not the problem; the post is not clearly the best answer yet, and probably needs more internal links from related posts.

The yearly update post

This is a very common cause on review and how-to sites. Each January somebody publishes “Best X for 2025” at a new URL instead of updating last year’s. After three years you have three posts on one search, the oldest with the most links and the newest with the freshest content. Keep one URL with no year in the slug, update it each year, and put the year in the title only.

Finding it in Google Search Console

Search Console is the only place that tells you which of your URLs Google actually showed for a query. Start here, with one query you care about.

1. Open Performance → Search results and set the date range to the last three months. Shorter ranges are too noisy for this.
2. Filter to one query. Click + Add filter → Query, then choose “Exact query” to see one search, or “Queries containing” to catch variations like sinking funds list and what are sinking funds.
3. Switch to the Pages tab. It now lists every one of your URLs that appeared for that query, with clicks, impressions and average position for each. Google notes in its Performance report documentation that tables grouped by page are aggregated by page, so each URL gets its own count even when two appeared in the same search.
4. Look for a second URL with a real share. One page with 95% of the impressions and another with 5% is normal. Two pages with 60% and 40% is worth a closer look.
5. Check for flipping. Click one URL in the table to filter to it, switch the chart to show impressions, and look at the line. Then do the same for the other. If one rises as the other falls, Google is taking turns.
6. Compare pages directly. Remove the page filter, add a Page filter with Compare, and put the two URLs side by side, still filtered to the query. The table then gives you a difference column for each metric. Google’s help page on advanced filtering and comparison covers the options, including regex query filters.

Here is what a real-looking result for the personal-finance example looks like:

Two posts with a 56/44 split of impressions, both outside the top ten, for the same informational search: that is the pattern worth fixing. The category page at position 41 is noise.

Checking a whole site with a spreadsheet

Checking one query at a time is fine for your top twenty searches. For a whole site you need every query and page pair in one table, and the export from the Performance report will not give you that: its Queries and Pages tabs are separate lists. There are two ways to get the pairs.

• Looker Studio. Add the Search Console connector and choose the URL Impression table (the Site Impression table counts by property, so it cannot tell your pages apart). Build a table with Query and Landing Page as dimensions and Impressions, Clicks and Average Position as metrics, then export it to Google Sheets.
• The Search Console API. Its searchanalytics.query method takes query and page as dimensions and returns up to 25,000 rows per request, paged with startRow. The script below exports a quarter. The service account’s email address has to be added as a user on the property first.

# pip install google-api-python-client google-auth
# Read-only. Exports every query + page pair for one quarter to query-page.csv.
from google.oauth2 import service_account
from googleapiclient.discovery import build
import csv

SITE = "sc-domain:example.com"   # or "https://www.example.com/"
creds = service_account.Credentials.from_service_account_file(
    "key.json", scopes=["https://www.googleapis.com/auth/webmasters.readonly"])
gsc = build("searchconsole", "v1", credentials=creds)

rows, start = [], 0
while True:
    resp = gsc.searchanalytics().query(siteUrl=SITE, body={
        "startDate": "2026-07-01", "endDate": "2026-09-30",
        "dimensions": ["query", "page"],
        "rowLimit": 25000, "startRow": start,
    }).execute()
    batch = resp.get("rows", [])
    rows += batch
    if len(batch) < 25000:
        break
    start += 25000

with open("query-page.csv", "w", newline="") as f:
    out = csv.writer(f)
    out.writerow(["query", "page", "clicks", "impressions", "position"])
    for r in rows:
        out.writerow([*r["keys"], r["clicks"], r["impressions"], round(r["position"], 1)])

Open the file in Google Sheets or Excel with the columns in this order: query (A), page (B), clicks (C), impressions (D), position (E). Then add two helper columns.

F2  pages with 10+ impressions for this query:
    =COUNTIFS(A:A, A2, D:D, ">=10")

G2  this page's share of the query's impressions:
    =D2 / SUMIF(A:A, A2, D:D)

Then filter: F >= 2, G between 0.15 and 0.85, D >= 100.
Sort by query, then by impressions (largest first).

What you are left with is every query where two or more of your pages each take a real share of the impressions. The thresholds (10 impressions, a 15–85% share, 100 impressions per query) are working rules, not anything Google publishes; loosen or tighten them until the list is short enough to read. Then group the rows by the pair of URLs involved, because the same two posts tend to compete on a cluster of related searches, and you fix the pair once.

For example, a SaaS help centre might find 31 queries with a split that come down to only nine pairs of articles, most of them an old article and the rewrite that replaced it, both still published. Nine decisions is an afternoon’s work; 31 looks like a project.

Quick checks: site: searches and free tools

Search Console only knows about searches where you already get impressions. Two quicker checks catch the rest, including competing pages that are too new or too weak to show up yet.

• A site: search. Type site:yourdomain.com sinking funds into Google. Google’s documentation for the site: operator says it does not necessarily return every indexed URL, so it is not a complete list or a ranking report. It is still a fast way to see which of your pages match the phrase. If five posts come back with near-identical titles, you have a cluster to look at.
• Your own titles. In WordPress, go to Posts → All Posts, click the Title column to sort alphabetically, and scan for repeats; use the search box for “best”, “guide” and each recent year. Ten minutes of reading catches most yearly-update duplicates.
• A free checker. Our keyword cannibalization checker gives you a first shortlist of pages that may be competing, so you know which pairs to open in Search Console first.

Deciding whether it’s a real problem

Now the important part. For each pair you found, decide whether there is anything to fix. Intent is the deciding question; the numbers tell you how urgent it is.

If you cannot tell whether the intent is the same, search the phrase yourself and look at what Google shows. If the top ten are all one kind of page (all recipes, all product listings, all step-by-step guides), Google has decided the intent, and only one of your pages fits it. If the results are mixed, there is room for both of yours.

The fixes, in brief

Once you know a pair is a real problem, there are six things you can do with it. Here they are in brief; each one has a full walkthrough, with WordPress steps and server examples, in the fixes guide linked at the top of this page.

The internal-link side of the problem

Internal links are where many competing-post problems are made worse, and the cheapest place to fix them. Google says in its link best practices that it uses links as a signal when judging the relevance of pages, and that good anchor text describes the page it points to. Your own links are the part of that signal you fully control. If half your posts link to one sinking-funds post and half to the other, you are describing both as the answer.

Three things to check for every pair you decide to fix:

• Which page do your links point at? Count the inbound internal links to each URL. A crawler will give you exact numbers; Search Console’s Links report (Links → Internal links → Top linked pages) gives a rough count, though Google describes it as a sample. The winner should have clearly more.
• Which words do they use? If links with the anchor text “sinking funds” go to both posts, pick one destination for that phrase. The guide to anchor text for internal links explains why one phrase should mean one page, and the one on exact-match links covers how much repetition is too much.
• Do any links go through a redirect? After a merge, every link to the old URL still works but takes an extra hop. Update them to point straight at the winner.

For the wrong-page problem, repointing links is the cheapest thing to try first, before anything you cannot easily undo. Picture a travel site whose tag archive for Kyoto keeps outranking its Kyoto itinerary. A crawl shows only two posts link to the itinerary in their text, while every Kyoto post links to the tag through the theme. Adding contextual links to the itinerary from six related posts costs half an hour and changes nothing else, so if the itinerary takes over in the following weeks, you know why. There is no published study that says how often this alone is enough; it is a low-risk first step, not a guaranteed fix.

Competing pages also tend to turn up during a broader audit; the internal link audit checklist includes them as one step, and the internal linking strategy pillar explains, more broadly, how to decide which pages your links should favour.

What to do on Monday

1. Pick your ten most valuable queries

   The ones that bring traffic or sign-ups. Run the six-step Search Console check on each.

2. Look for flipping and real splits

   Two URLs with at least 15% of the impressions each, or a ranking URL that changes from week to week.

3. Decide by intent

   Use the table above. Write down the verdict for each pair, including “leave it”.

4. Repoint internal links first

   For every pair you are fixing, point links and the matching anchor text at the winner.

5. Then merge, redirect or re-target

   Following the step-by-step guide, one pair at a time, with the date noted so you can measure it.

Where Silotree fits

Silotree detects competing posts from your Search Console queries, suggests which one should be the winner, and repoints your internal links to it in one click. Links that go through redirects after a merge can be fixed in one click too, on every plan. The merging and redirecting itself stays your decision, made in WordPress. If you want to see the rest of your link problems first, the free internal link checker needs no account. If you already use Rank Math, the comparison of Silotree and Rank Math covers what each one does with internal links, and the pricing page lists what each plan includes.

Questions people ask