Keyword Cannibalization: When Your Own Pages Compete

You have three blog posts about the same thing. One from 2021, one from 2023, and one you published last month because you forgot the older two existed. Now Google cannot decide which one to show. It swaps them around, and none of them makes page one. That is keyword cannibalization.

What keyword cannibalization actually is

Cannibalization happens when several URLs on the same domain target the same query with the same search intent. It is not the same thing as duplicate content. The texts can be completely different — different wording, different length, different examples. The problem is intent: if three pages answer the same question, Google has to pick one.

One distinction people get wrong: two pages targeting similar words are not automatically a problem. "Women's running shoes" and "men's running shoes" are two different intents and they sit next to each other without any trouble. But "how to choose running shoes" and "what to watch out for when buying running shoes" are the same intent wearing two different coats.

Google does not officially punish cannibalization. There is no penalty with that name anywhere in the documentation. The damage happens more quietly.

How it hurts the site

Split signals. The backlinks you earned over two years get divided across three URLs. So do your internal links. Instead of one strong page you end up with three average ones.

Unstable positions. If Search Console shows several URLs rotating on one query and the average position bounces between 8 and 19, that is the classic picture. Google is testing which page performs better and is not getting a clear answer.

Lower CTR. Google often shows the URL whose title is not the best match for that query. The user skips it and clicks a competitor.

Wasted crawl budget. On an ecommerce site with thousands of URLs this stops being theoretical. The crawler works through three versions of the same thing instead of pages it has never seen.

How to find cannibalization

Search Console, Performance report

Fastest route. Open the Performance report, go to the Queries tab, click a specific query, then switch to the Pages tab. If one query returns two or three URLs with a meaningful number of impressions, you have a candidate.

What to look for specifically:

• a query where the first URL has 400 impressions and the second has 260 — that is a clear split
• a query where the URLs swap over time (compare two monthly ranges side by side)
• a query where both URLs sit between positions 11 and 25 and neither breaks higher

If one URL has 900 impressions and the other has 12, leave it alone. That is normal noise.

A site: search in Google

Type site:yourdomain.com "the phrase you care about" into Google. You will see every indexed page containing that phrase. When five come back and all of them look like they answer the same question, you know where to start.

Your own list of titles

Export the title tags and H1s of every page and sort them alphabetically. Cannibalization usually gives itself away right there — three titles that differ only in word order. This is exactly what I look for during an audit, alongside duplicate titles and thin descriptions. It is very often the same batch of pages.

Four ways to fix it

1. Merge into one page

Usually the right call for blog posts. Pick the URL with the most links and the best Search Console history. Move the best parts of the others into it: missing sections, examples, numbers. Then 301 redirect the remaining URLs to the main one.

Check before you merge: which URL has more backlinks, which is older, and which has the higher CTR in Search Console. The newer one does not automatically win.

2. Separate the intent

Sometimes you do not want to delete anything, because both pages make sense — they are just written too similarly. In that case rewrite one so it targets a different intent. Turn "how to choose a CRM" into a comparison of specific tools, and the other into an implementation guide. Change the title, the H1, the meta description and the H2 structure. Half a rewrite is not enough; Google will not read it as a new intent.

3. Deindex and remove

Pages with no traffic, no links and nothing to say can simply be deleted with a 410. Or given noindex if you still need them for visitors but not for search. This applies mostly to old tag pages, archives and ecommerce filters.

4. Canonical

A canonical tag solves cases where pages are technically near-identical — URL variants, parameters, filters. For two differently written articles on the same topic, canonical is the wrong tool; that case needs a merge and a redirect. And Google treats canonical as a hint, not a command.

Cannibalization in ecommerce

Shops have it worse than blogs, because cannibalization appears there on its own.

Category versus subcategory. "Bikes" and "Mountain bikes" fight each other when the subcategory holds most of the parent's products. The fix is in the copy and the internal links: the parent should describe the whole range, the subcategory only its slice.

Filters and parameters. ?colour=black&size=42 generates dozens of URLs with practically the same content. These belong under canonical or noindex, depending on whether anyone actually searches for that filter.

Product versus category. If a category holds a single product, both pages chase the same query. The product usually wins, and the category can be left without its own target.

Product variants. Three colours of one t-shirt as three URLs with identical copy. Either combine the variants on one page or canonical them to the main one.

How to check the fix worked

Do not guess. Write down the deployment date and compare 28 days before with 28 days after in Search Console, for that specific query. Watch three things:

1. Number of URLs on the query. After a successful fix one should dominate.
2. Average position. It should stabilise, not just jump. A two-day spike means nothing.
3. Total clicks. Clicks across all URLs before, against clicks on the merged page after. This is the only number that really matters.

Now the unpleasant part. After a 301, it takes several weeks before Google rewrites the index and moves the signals across. On smaller sites the first change usually shows up after two to four weeks; on large ones, later. Anyone telling you the result lands in three days is wrong.

Also expect a merge to reduce total impressions in the short term. Three pages ranking low produced more impressions than one page ranking higher. Look at clicks, not impressions.

How to avoid it in the first place

Keep a simple list: URL, target query, one sentence about the intent. Before writing a new article, check whether that query is already on the list. Three minutes now saves three hours of cleanup later.

Pick topics from real data, not from ideas. Queries where you already show up in positions 11 to 30 do not need a new article — they need the existing one improved. Write a new article for queries where you do not appear at all.

And finally, internal links. When a topic has one main page, link to it from the other articles with consistent anchor text. That tells Google which URL is supposed to own the query.

What I find

When I crawl a site I flag pages with near-identical titles and H1s, clusters of articles aiming at the same intent, and URLs where Search Console shows several addresses rotating on one query. For the findings that matter I write a new title and meta description ready to paste, so the pages pull apart in intent.

What I will not do automatically: I will not merge two articles for you or set up the 301. That decision belongs to a person who knows which page matters more to the business. You get the evidence; the click is yours.