Keyword cannibalization happens when two or more of your pages compete for the same query and Google cannot decide which one to rank — so it alternates between them, or ranks the weaker one, or ranks neither well. It is real, it is common on sites over a few hundred pages, and it is also the most over-diagnosed problem in SEO.
The clarifying test, up front: cannibalization is only a problem if it is costing you. Two pages ranking for the same term is not automatically bad. Two pages swapping positions week to week, both stuck outside the top ten, with impressions split between them — that is bad. Learn to tell those apart before you start merging pages.
What Cannibalization Actually Looks Like
The symptoms are specific. Watch for these:
- URL flapping. Your ranking URL for a query changes between checks — page A on Monday, page B on Thursday, page A again next week. This is the clearest signal.
- Position ceiling. Both pages hover between 8 and 20 and neither breaks through, despite decent links and content.
- Impression split. In Search Console, a single query shows meaningful impressions distributed across two or three URLs.
- The wrong page ranks. Google surfaces a blog post for a commercial query when you have a purpose-built product page.
- A drop after publishing. You publish a new article and an existing page’s position for a term degrades within a few weeks.
Note what is not on that list: two pages both appearing on page one for the same query. That is double occupancy, and it is a win — you own more of the SERP. Do not merge those.
Detecting It in Google Search Console
Search Console is the only free source of truth here, because it reports impressions and clicks per query per page for your actual site. Here is the process.
- Open Performance → Search results and set the date range to the last three months. Comparison periods hide the pattern.
- Click the Queries tab and pick a query you care about.
- With that query filter still applied, switch to the Pages tab.
- Enable the Impressions, Clicks, and Average position columns.
- Read the result. One page with 95% of impressions means no cannibalization. Two or three pages each holding 20–50% means you have a real split.
To find cases you did not already suspect, work in the other direction. Export the last 3–6 months of query and page data (the Search Analytics API or a Looker Studio connector gives you far more rows than the 1,000-row UI export), then group by query and count distinct URLs with more than 10 impressions. Sort by total impressions descending. The top of that list is your prioritized fix queue.
One more check worth thirty seconds: the site search operator. Run site:yourdomain.com "your query phrase" and see how many of your own pages Google considers relevant. If it returns eight near-identical posts, you have a content strategy problem, not just a cannibalization incident.
When Two Pages Targeting Similar Terms Is Fine
This is the part most guides skip, and it causes real damage — teams merge pages that were working and lose traffic from both.
Two pages can coexist safely when they serve genuinely different intents. “Best CRM software” (commercial investigation) and “what is a CRM” (informational) share vocabulary and share nothing else. Google treats them as separate queries with separate SERPs, and having a page for each is correct.
They also coexist fine when they sit at different levels of a hierarchy — a category page for “running shoes” and a subcategory for “trail running shoes” — or when they target different locations, different audiences (“for enterprise” versus “for freelancers”), or different formats where the SERP itself is mixed.
The practical decision rule: compare the SERPs. Search both queries in an incognito window and count overlapping results in the top ten. If six or more URLs overlap, Google considers them the same query and one page should serve both. If three or fewer overlap, they are distinct and separate pages are appropriate. Four to five is a judgment call — lean toward one strong page unless you have a clear reason not to.
The Fix: Consolidate, Differentiate, or Redirect
Once you have confirmed a genuine split, there are exactly three options. Walk them in order.
1. Consolidate — when both pages are thin and cover the same ground
Pick the winner: the URL with more inbound links, better existing position history, and a cleaner URL structure — usually in that order. Merge the unique, valuable sections of the loser into the winner so the combined page is genuinely more complete, not just longer. Then 301 redirect the loser to the winner and update every internal link that pointed at the old URL. Do not leave the old page live with a canonical tag if you have removed its content; redirect it properly.
Expect the consolidated page to wobble for two to four weeks while Google reprocesses the redirect, then settle above where either page sat individually. If it does not improve within eight weeks, the problem was never cannibalization.
2. Differentiate — when both pages have real, distinct value
Rewrite each page around a sharper, separate intent. Change the H1 and title to make the distinction explicit. Rework the opening paragraph so the first thirty words tell both Google and a reader exactly which question this page answers. Then fix the internal linking, which is usually the actual cause: if your anchor text says “keyword research” pointing at both pages, Google gets the same mixed signal you sent it. Standardize anchors so each page owns its term, and link between the two pages so their relationship is explicit.
3. Redirect or remove — when one page has no reason to exist
Some pages are simply redundant: an old post covering a topic you have since covered properly, a duplicate created by a migration, a tag archive competing with a real category page. Redirect it to the closest relevant page. If nothing relevant exists and the page has no links and no traffic, a 410 is cleaner than keeping dead weight in the index.
Fixing the Root Cause
Cannibalization is almost always a symptom of publishing without a keyword map. If nobody records which URL owns which primary term, writers eventually cover the same ground twice — especially across a team, or across years.
Build the map. One row per URL, one primary keyword per URL, plus the secondary terms it may also serve. Before commissioning anything new, check the map. If the primary term is taken, the correct action is updating the existing page, not writing a new one. This single habit prevents more cannibalization than any audit ever fixes.
A saved keyword pool per project, feeding both the writer and the tracker, enforces this structurally — it is how SEO Rocket links research to content so the same term does not get assigned twice, and any tool or spreadsheet that keeps the mapping visible at commissioning time will do the same job.
Verifying the Fix Actually Worked
Give it four weeks minimum before judging, and measure with the same Search Console view you used to diagnose. You are looking for three things: impressions consolidating onto one URL, average position for the query improving on the surviving page, and total clicks for that query rising above the previous combined total.
Two honest caveats. Daily position movement of two or three places is normal noise, so read a trend line, not a single reading — a consolidated page bouncing between 6 and 9 is holding steady, not failing. And rank tracking that records the ranking URL, not just the position, is what makes ongoing detection cheap: URL flapping shows up in that column weeks before it shows up in your traffic.
Finally, resist the urge to run a site-wide cannibalization purge. Fix the five queries with the most impressions at stake, verify the outcome, then decide whether the next five are worth it. Most sites find the tail is not worth touching — and every unnecessary merge is a redirect, a lost URL, and a small risk taken for nothing.
