Keyword cannibalisation: finding it and fixing it

Cannibalisation is when two or more of your own pages compete for the same query, and neither does as well as one page would have.

It is a real problem, and it is also over-diagnosed to the point of being unhelpful. Plenty of sites have pages that *mention* the same phrase, which is not cannibalisation. Plenty have pages that legitimately serve different intents around a shared topic, which is not cannibalisation either. Ripping those apart makes things worse.

So the first job is telling the real thing from the false alarm.

What it actually is

Cannibalisation requires two conditions together:

1. Two pages target the same search intent — not the same words, the same *question from the same person*.
2. Google is visibly uncertain which to show, or is showing the weaker one.

Both. Two pages containing the phrase "content strategy" is not cannibalisation if one is a guide for beginners and the other is a case study — those are different questions.

How to confirm it

Three checks, in order. If the first two do not show anything, you probably do not have the problem.

Check 1: does one query flip between URLs?

The clearest evidence. In Search Console, look at a query and its landing pages over a few months. If the URL Google shows for it changes back and forth, Google is genuinely uncertain — that is the signature.

A query that consistently maps to one page is fine, even if other pages mention the term.

Check 2: does the site: search show the wrong page winning?

Search site:yourdomain.com <the query> and see which of your pages Google puts first. If it is not the page you intended to rank, that is a mismatch worth understanding. It may be cannibalisation; it may simply be that the other page is genuinely better at answering it, which is useful information either way.

Check 3: read the two pages and state their questions

Write down, in one sentence each, the question each page answers.

If you cannot make the sentences meaningfully different, you have one page written twice. If you can — genuinely, without straining — you probably do not have a problem, and the real issue is something else.

That last case is common. A page underperforming for reasons of intent or depth gets blamed on cannibalisation because it is a more comfortable diagnosis. The diagnostic checklist covers the alternatives.

The four fixes

Ordered by how often they are the right answer.

1. Merge

The best outcome when two pages genuinely serve one intent. Take the stronger URL — usually the one with more links or history, not necessarily the better-written one — move the best material from the weaker into it, and redirect the weaker to it permanently.

You end with one page that is better than either, holding the combined history. This is nearly always preferable to deleting one and hoping.

2. Differentiate

When the pages *should* be different but have drifted together. Rewrite each so it clearly owns its own question, and make the titles and headings reflect that difference plainly.

This is the right fix when the underlying topics are genuinely distinct and the execution blurred them.

3. Consolidate the internal links

Sometimes both pages are fine and your own linking is the cause: half your internal links point at one, half at the other, so you are telling Google both are the main page.

Pick the one that should win. Point the internal links at it. This alone resolves a surprising share of cases and costs an hour.

4. Canonical, occasionally

Where two pages must both exist for real reasons — a printable version, near-identical variants — a canonical tag tells Google which is the primary. Google's documentation on consolidating duplicate URLs covers the mechanics.

Do not use a canonical to paper over two pages that should have been merged. It is for genuine duplicates, not for editorial indecision.

Doing a merge properly

The fix people get wrong in execution rather than in diagnosis, so the mechanics are worth spelling out.

1. Decide the surviving URL first, using the criteria in the next section. Everything else follows from it.

2. Move the content across before redirecting anything. Take what the weaker page did better — a clearer explanation, an example, a section the survivor lacks — and merge it in properly rather than appending it. The result should read as one page, not two stapled together. If the merged page is now disorganised, restructure it; a badly merged page can rank worse than either original.

3. Redirect the retired URL to the survivor, permanently. A permanent redirect is what consolidates the signals. A temporary one tells Google the old URL is coming back.

4. Redirect to the specific page, not the homepage. Redirecting a retired article to the homepage is treated as a soft 404 and throws away everything the old page had.

5. Update your internal links. Every link on your site that pointed at the retired URL should now point at the survivor directly. The redirect handles the outside world; you can edit your own pages, so edit them.

6. Leave it alone for a month. Then check whether the query now maps consistently to the survivor.

The step most often skipped is five, and the consequence is a site quietly relying on redirects for links it could simply have corrected.

Which page should win?

When merging, the instinct is to keep the better-written page. Usually wrong.

Keep the URL with:

• more inbound links from other sites — the hardest thing to replace
• more history — an older URL with accumulated signals
• the better-fitting URL slug, if the others are equal

Then move the better *content* onto it. Content is portable; links and history are not.

Preventing it

Cannibalisation is a structural problem and it is far cheaper to prevent than to unpick.

One page, one question, written down. Before publishing, state the question the page owns and check that no existing page owns it. This is the single most effective habit.

Keep a list of what you have published and what each piece answers. Most sites over fifty pages have no such list, which is exactly why the problem creeps in.

Structure by subject rather than by date. Clusters make overlaps visible as you build, because you can see the whole territory. A dated blog hides them — see topic clusters and pillar pages.

Be suspicious of the second article on a topic. If you are writing something that feels adjacent to an earlier piece, check whether you are actually about to write it again with different words.

When two similar pages are correct

Worth stating, because over-correction is a real cost.

These are legitimately separate:

| Page A | Page B | Why both | |---|---|---| | What a thing is | How to do the thing | Different reader stages | | Guide for beginners | Reference for practitioners | Different audiences | | The general case | A specific sector's case | Different situations | | An argument about a topic | A definition of a term in it | Different intents entirely |

If your two pages fit one of those patterns and the query data shows no flipping, leave them alone.

The ecommerce version

Cannibalisation looks different on a shop, and it is more often structural than editorial.

Category versus article. A "running shoes" category page and a "best running shoes" buying guide both target the same buyer. Usually the category page should win the commercial query and the guide should target the comparison one — and the guide should link to the category prominently, which is what most sites forget.

Category versus subcategory. "Running shoes" and "men's running shoes" are genuinely different queries. "Running shoes" and "running footwear" are one query with two URLs, and one of them should not exist.

Filtered URLs. Faceted navigation can generate hundreds of near-identical pages — colour, size, price band — each competing weakly with the category they came from. This is the largest source of self-competition on most shops, and it is fixed by deciding which facets deserve an indexable page at all rather than by editing content.

Product variants. The same product at several URLs, one per size or colour. Canonical tags to a single product page are the standard answer, and most platforms do this by default if you let them.

The diagnosis is the same as everywhere else: does one query flip between URLs, and can you state each page's question distinctly? The fixes just happen in the platform settings rather than in the text.

What to expect after fixing it

Merging usually helps within weeks, faster than most SEO changes, because both pages are already indexed and understood.

Expect a brief dip as the redirect is processed and the signals consolidate.

Do not merge in bulk. Fix two or three, watch what happens, then continue. If you merge fifteen pages at once and something goes wrong, you will not know which one caused it.

The honest limit

Fixing cannibalisation makes a site tidier and lets one page take the position two were splitting. It does not make an unwinnable query winnable. If the pages above you are strong and specific and yours are not, consolidation moves you from twelfth and fourteenth to eleventh — which is real, and small.

Whether the query is winnable at all is a separate question, covered in judging whether you can realistically win a keyword. If it is not, the routes are a narrower query or paying for placement outright, and the current prices for that are on the rankings board.

Common questions

Is having a topic mentioned on many pages a problem? No. Mentioning is not targeting.

Should I noindex the weaker page? Rarely. Merging and redirecting keeps its value; noindexing discards it.

Can category and article pages cannibalise each other? Yes, and it is common on ecommerce sites. The category page usually deserves to win commercial queries.

How do I find it without Search Console? With difficulty. A site: search reveals the worst cases, but query-level flipping is only really visible in Search Console.