Type "best running shoes for flat feet" into Google and watch which page ranks. Now open Search Console and look at the queries driving traffic to that URL. If the phrase that earns the click is nothing like the phrase you optimized the page for, something is off. Not broken, exactly. Just messy. That's usually the first moment people realize two pages on their own site have been quietly fighting over the same search term for months.
Keyword cannibalization doesn't announce itself. There's no alert, no red flag in your dashboard. What you get instead is a slow leak: rankings that wobble between positions you can't explain, a blog post that outperforms the product page it was supposed to support, and a support article that keeps stealing clicks from your main landing page. I've watched this happen on sites with fewer than 40 pages and on sites pushing 5,000. The pattern is the same.
Key takeaways
- Keyword cannibalization means two or more URLs on your site compete for the same query or the same search intent.
- The clearest sign isn't a ranking drop, it's an unexplained rotation: two URLs swapping position on the same query week after week.
- Roughly 10% of queries on a typical mid-sized site show overlapping URLs, but only a small fraction—often under 1%—actually justify a fix.
- Merging and redirecting solves intent overlap; a canonical tag or de-optimization solves duplication without merit.
- Doing nothing is a legitimate answer. Not every overlap costs you traffic.
What keyword cannibalization actually means (and what it doesn't)
Two pages. One query. Both trying to rank. That's the short version. The longer version matters more, because a shallow definition leads people to "fix" problems that aren't problems.
When Google can't decide which of your URLs best answers a query, it rotates them. One week the blog post sits at position 4. The next week the category page takes that slot. From your end, it looks like instability. From Google's end, it looks like your site hasn't made up its mind.
Cannibalization vs cannibalisation
Same word, two spellings. If you're searching for "keyword cannibalisation," you're reading British English results; "keyword cannibalization" is the American version. The concept is identical. No separate strategy required. I mention it because people sometimes land here confused about whether the two terms describe different phenomena. They don't.
Not the same thing as duplicate content
Duplicate content is when two URLs serve near-identical text. Cannibalization is when two URLs target the same intent, even if every sentence is different. A 2,000-word guide and a 400-word FAQ page can cannibalize each other perfectly well without sharing a single paragraph. That distinction trips people up constantly, and it's the reason fixing "duplicate content" often does nothing for an actual cannibalization problem.
Here's a scene from a client project two years ago: their support team published a troubleshooting article, and their marketing team published a "how to fix X" blog post three months later. Different writers, different tone, zero overlap in wording. Both ranked for the same query. Google alternated between them for eleven weeks. Neither reached the position either would have held alone.
Signs of keyword cannibalization you can actually measure
The diagnostic signals are concrete. You just have to know where to look, and most of the standard advice skips that part entirely.
1. URL rotation on the same query
Open Search Console, go to Performance, filter by a specific query, then click the Pages tab. If two URLs appear for one query with meaningful impressions on both, you have overlap. If they alternate which one gets the impressions month over month, you have active cannibalization. This is the single most reliable test, and it takes about ninety seconds per query.
2. Ranking oscillation you can't attribute to anything
A page that flips between position 3 and position 7 with no algorithm update, no competitor change, and no on-page edit is telling you something. Something is competing with it. Often that something is yours.
3. Low CTR for a high position
If a URL sits at position 2 with a 3% click-through rate, that's abnormal. Sitelinks can absorb some clicks, but a CTR that flat at a top-three position usually means a second URL is appearing nearby and splitting attention. Check the SERP manually. Search the query yourself and see what else from your domain shows up.
4. A blog post outranking the page it's meant to support
This is the most expensive version of cannibalization in SEO. Your informational article pulls the traffic, but the product or service page—the one with the conversion path—never gets seen. In e-commerce, it's the category page losing to a thin "buyer's guide" post. I've seen a single blog post absorb 60% of the organic clicks intended for a product page that had ten times the commercial intent behind it.
5. Pages that get impressions but never get clicks
A URL with 4,000 impressions and twelve clicks over three months is a page that ranks for something, but loses every click to a sibling. It's not dead. It's just being outcompeted by its own neighbor.
What are the types of cannibalization?
There are essentially two: intent-based and duplication-based. Intent-based is when two pages serve the same search goal with different content—a guide and a listicle, for example. Duplication-based is when two pages are functionally equivalent: an old product page and its updated replacement, a category and a tag archive, a paginated series and its canonical index. The first type needs a strategic decision about which page deserves the query. The second type is usually a technical cleanup with a canonical tag or a redirect.
There is also a third situation worth naming: intra-SERP overlap, where your homepage and a deep page both target the brand's primary term. That's rarely a bug. It's often useful, because both URLs can appear in the same result set. Don't "fix" it without checking whether it's actually earning you an extra listing.
What is the 80/20 rule in SEO—and why it matters here
The 80/20 rule says most of your results come from a minority of your effort. In SEO, the practical version looks like this: a small share of your pages will drive the vast majority of your organic traffic, and a similarly small share of your queries will be worth optimizing aggressively.
Applied to cannibalization: on a site with a few hundred URLs, you'll typically find somewhere around 10% of queries showing overlapping URLs. Of those, the share that actually justifies intervention is small—often under 1% of your total query count. Pull up a mid-sized site and run the check yourself. You'll see hundreds of overlaps, and you'll only need to touch a handful.
This matters because the alternative—fixing every overlap—is a trap. You'll spend weeks merging pages that were never hurting each other, and you'll lose the traffic you were trying to protect. The rule is a filter, not a schedule.
How to fix keyword cannibalization without breaking what works
Pick the fix based on which kind of overlap you've got. There's a clear decision tree, and it's not the one most guides describe.
| Situation | Best fix | When to avoid it |
|---|---|---|
| Two pages, same intent, both have backlinks | Merge into one, 301 the weaker URL | If the weaker page has unique traffic sources you can't replicate |
| Old page and new page, near-identical content | Canonical tag pointing to the survivor | If you need both URLs indexed for internal linking reasons |
| Two pages, related but distinct intent | De-optimize one: strip the target keyword from title and H1 | If traffic to the de-optimized page is growing on its own |
| Homepage and inner page on the brand term | Usually leave it. Both may rank. | Only if the inner page is outranking the homepage on non-brand terms |
| Paginated series and its canonical | Rel=canonical on paginated URLs | Almost never—this one is nearly always correct to fix |
Merge and redirect: the heavy option
When both pages deserve to exist on their own merits but target the same query, merge the content into one URL. Preserve the stronger backlink profile. Keep the better-performing URL as the survivor. Redirect the other with a 301. Expect a temporary dip—I typically see 2 to 4 weeks of soft rankings after a merge before the consolidated page stabilizes, often higher than either original.
Canonical tag: the light option
If both URLs need to exist for structural reasons, point a rel=canonical at the one you want indexed. Google respects canonicals most of the time, but not always—so test it. If the wrong URL still shows up in Search Console two weeks later, escalate to a redirect.
De-optimization: the counterintuitive option
Sometimes the right fix is to stop trying. Remove the target keyword from one page's title and H1, shift its focus slightly, and let the other page own the query. This works well when the two pages serve meaningfully different audiences. It fails when they don't—you'll just create confusion.
When not to fix it at all
If both pages rank in the top ten for a query and the combined clicks are stable or growing, leave it alone. You're not losing traffic. You're occupying more of the SERP. A client of mine kept two pages ranking for the same informational query for over a year—positions 4 and 8—and the combined click volume was 40% higher than either page achieved alone. Merging them would have been a mistake.
The instinct to fix everything is the real problem. Run the diagnostic. Sort by whether the overlap costs you traffic. Fix the ones that do. Ignore the rest.
The short checklist I actually use
- Filter Search Console by query, then look at the Pages tab.
- Flag any query with two or more URLs holding meaningful impressions.
- Confirm which URL has the better backlink profile and conversion path.
- Choose your fix from the table above.
- Document the change and check back in 30 days.
- Leave overlaps that aren't costing you anything.
Keyword cannibalization gets treated like a disease, but it's closer to a symptom. It's your site telling you that two pages are doing the same job, and the search engine is doing the deciding for you. Whether that's a problem depends entirely on whether the outcome is worse than having one page do that job alone. Most of the time, it isn't. The rest of the time, you know within ten minutes of checking which URL rotation is real and which is noise.
