How to Fix Keyword Cannibalization on a Service Site

Most cannibalization advice is written for publishers with 400 blog posts. You are not that. You run a service business with maybe 40 pages, and the pages fighting each other are the ones that make you money: the service page, the emergency variant, the city page, and one blog post from 2024 that somehow outranks all three.

That is a different problem with a different fix. Here is how to prove it, decide what each page deserves, and stop creating the problem next quarter.

What does keyword cannibalization look like on a service-business site?

It looks like one query returning a different URL of yours every few weeks, with position bouncing between roughly 5 and 20 while nothing on the site changed. That URL rotation is the tell. A page that is genuinely losing to a competitor drifts down slowly. A cannibalized page teleports, because Google is re-picking which of your URLs to show.

The classic service-business version, using a plumber as the example:

Three of those four are legitimately different pages. The collision usually starts when the city page copies the service page's body text, or when the blog post opens with 800 words explaining what drain cleaning is — which is the service page's job. Now Google has two documents making the same argument and has to pick.

Why do your service page, city page and blog post all target the same query?

Because they were written at different times by different people with no map, and each one was independently optimized for the fattest keyword in the vertical. Nobody sat down and assigned one intent to one URL.

Three things reliably cause it:

• Template cloning. Your 12 city pages are the service page with the city name swapped. Identical H1 pattern, identical body, one word different.
• Blog posts that sell. An informational post that ends with 400 words of 'why choose us' and a quote form has quietly become a second service page.
• Agency deliverables. A previous agency was paid per page. So you got a page for 'drain cleaning', another for 'drain cleaning services', and another for 'drain cleaning company' — three URLs, one intent.

The prevention rule is one line long: one buyer intent, one URL. Not one keyword — one intent. Keyword variants belong on the same page. Ahrefs' 2017 study of 3 million search queries found the average #1 ranking page also ranked in the top 10 for nearly 1,000 other relevant keywords, with a median around 400 (Ahrefs). You do not need a URL per phrasing. One strong page harvests the variants for free.

How do you prove it in Search Console in 20 minutes?

Open Search Console, set the date range to the last 3 months, and run a page-level check on your 10 most commercial URLs — that is roughly 20 minutes on a 40-page site, no spreadsheet required. You are looking for one thing: the same query showing impressions on two different pages.

The sequence:

• Performance report, Pages tab. Click your service page URL. It filters everything to that page.
• Switch to the Queries tab. Note the top 10 queries by impressions. Screenshot it or paste it somewhere.
• Clear the page filter. Now filter by one of those queries instead, and switch to the Pages tab.
• If two or more of your URLs show impressions for that query, you have overlap. That is the raw signal.
• Sanity-check severity: overlap only matters if the impressions are meaningful on both URLs and the average position for that query is unstable or stuck outside the top 10.

Do not stop at overlap. Ahrefs makes the point bluntly in its own guide: real cannibalization needs competing pages and demonstrable harm to organic performance, not merely keyword overlap (Ahrefs). Two of your pages picking up impressions for a query while the main one sits at position 4 and books calls is not a problem to solve. Leave it alone.

Also check the Page Indexing report for 'Duplicate, Google chose different canonical than user'. That is Google telling you, out loud, that it disagrees with your page choices. If you would rather have someone else run this pass, that is exactly what our technical SEO audit covers.

Merge, differentiate, canonical, de-optimize or noindex — which does each case need?

There are five repairs, and the right one is decided by page type, not by keyword overlap percentage. Here is the decision table we work from:

Verdict: merge is the default and canonical is the most over-used. A canonical tag is a hint, not an order. Google ranks canonicalization signals in order of strength — a redirect is 'a strong signal that the target of the redirect should become canonical', rel=canonical is likewise 'a strong signal', and sitemap inclusion is only 'a weak signal' (Google Search Central). Google also says these methods stack and become more effective combined. If two pages are truly the same page, redirect one; do not leave both live behind a canonical and hope.

When you merge, use a permanent server-side redirect. Google's docs state that 301 and 308 status codes mean the page has permanently moved, and calls permanent server-side redirects the best way to make sure Search and people land on the right page (Google Search Central). Meta refresh is the fallback, not the plan.

Merging works. In Ahrefs' own write-up of consolidating two guides on the same topic, they report that after redirecting one into the other, only one page gets traffic, 'but it gets way more traffic than both pages (combined) did beforehand' (Ahrefs).

When are two pages on the same topic actually fine?

Two pages on the same topic are fine whenever they answer two different questions from two different buyers — and on a service site that is most of the time. Overlapping words are not the test. Overlapping intent is.

Fine to keep separate:

• A service page ('we do it') and a how-to post ('here is how it works') — different buyer, different stage.
• A service page and a city page — as long as the city page carries real local substance and not a find-and-replace of the service page. We cover what that looks like in location pages for service-area businesses.
• A pricing/cost post ('what does drain cleaning cost') and the service page — cost queries are research intent and convert into the service page later.
• Two genuinely different services that share a word ('drain cleaning' vs 'drain repair').

The honest test: open both pages, read the H1 and the first paragraph, and ask which one you would hand to a buyer who typed the query. If you hesitate for more than a couple of seconds, Google is hesitating too — and Google resolves that hesitation by rotating your URLs.

Why does a page drop from 3 to 19 with no update?

A clean drop from position 3 to the high teens, with no site change and no core update, usually means Google swapped which of your URLs it ranks — the query did not leave, your page did. Filter that query in Search Console and switch to the Pages tab. If a second URL of yours suddenly picked up impressions on the same date, that is a URL swap, not a ranking loss.

Two other things cause the same-looking drop, and you should rule them out before you go redirecting things:

• Intent shift in the SERP. Google re-decided what the query means — the top 10 filled with directories or map results and your page type no longer matches. Nothing to merge here; the fix is a different page format.
• Content decay. Competitors updated, you did not. Slow slide, not a teleport, and usually across many queries at once rather than one.

The distinguishing signal is shape. Cannibalization is jagged — up 4, down 14, back up. Decay is a slope. If you are staring at a graph that could be either, our post on why your service business isn't ranking walks the wider diagnostic tree.

How do you stop cannibalization before you publish the next page?

You stop it with a one-page keyword-to-URL map that every new page has to pass through — a spreadsheet with one row per URL and one column for the single intent that URL owns. Two minutes per page, and it prevents the entire problem.

The rules we apply before any page ships:

• One intent, one URL. If the intent already has an owner, the new content becomes a section on that page, not a new page.
• Site-search first. Google site:yourdomain.com plus the target keyword before writing. If a page already ranks for it, you are about to compete with yourself.
• Internal anchor discipline. Anchor text is a vote. If half your internal links say 'drain cleaning' and point at the blog post, you told Google the blog post is the drain-cleaning page. Point commercial anchors at the commercial page, always.
• City pages need local substance or they don't ship. Templated geo pages are the number-one source of self-competition on service sites. See our guide to on-page SEO for service pages for what a page has to carry.
• Quarterly re-check. Twenty minutes in Search Console, ten commercial URLs. Cheaper than a merge you have to undo.

One more thing, and it matters more than any of the above: cannibalization is a symptom of publishing without a map. Most sites that have it got it from an agency paid per page, or from a content plan that counted posts instead of intents. Fixing the pages without fixing the process just buys you 18 months.

If your rankings keep swapping between URLs and you want the overlap mapped, the primary pages picked, and the redirects written for you, that is the first thing our SEO audit does. Month-to-month, no lock-in, and you keep everything we build. Get my free audit.