How to Fix Keyword Cannibalization: Find the Splits in GSC and Pick a Winner

There is a special kind of ranking problem where your competitor is you. Two of your own pages take turns ranking for the same query, neither climbs, and every week the Performance graph draws a staircase where a trend line should be.

That is keyword cannibalization, and the fix is rarely writing more. It is a decision: which page owns the query, and what happens to the other one. This guide is the workflow I use to find the splits in Search Console, confirm they are real, and consolidate them without losing what the losing page earned.

What cannibalization is, and what it is not

Keyword cannibalization is two or more of your pages competing for the same query with the same intent. The damage is signal splitting:

• Internal links split. Half your anchors point at each page, so neither accumulates the authority one page would have.
• Clicks and engagement split. Each page earns half a history, and Google reads two lukewarm candidates instead of one proven answer.
• Relevance blurs. Neither page is the obvious best match, so Google keeps re-testing which to show, which is the swapping you see in the position graph.

Just as important is what does not count. Two pages on one query is normal, sometimes ideal, when the intents differ: a comparison post and a product page can share a query happily. The test is never "same keyword". It is "same job".

Step 1: Find the splits in GSC

Search Console will not name cannibalization for you, but it records every symptom. For a query you already suspect:

1. Open the Performance report, date range set to the last 3 months.
2. Add a query filter for the exact query.
3. Switch to the Pages tab. Every URL listed collected impressions for that query.
4. One URL means you are fine. Two or more with real impressions means you have a candidate.

For the site-wide sweep, export queries with their pages to a sheet and flag every query where two or more URLs show impressions in the same window. Group suspect queries by topic first with a regex filter; our free GSC regex generator builds the pattern, and the brand-exclusion version keeps navigational queries out of the sweep.

The position wobble, your fastest tell

The signature I check first: open the suspect query's position graph. A healthy ranking drifts. A cannibalized one moves in steps, because the number is an average and the URL behind it keeps changing. Pair the graph with the Pages tab and you can usually watch the two URLs trade places window by window.

Step 2: Confirm the split is real

Every fix in this guide costs a URL something, so do not skip the confirmation. Three checks, in order:

• Search the query yourself. In an incognito window, look at what Google actually shows. If it lists both your pages at once as a double listing, you are being rewarded, not punished. Leave it alone.
• Read both pages as the searcher. Would the person typing this query be equally served by either page? Equally served means same intent, and same intent means real cannibalization. A guide and a pricing page sharing a query are doing different jobs.
• Ask what you intended. If one page was never meant to rank for this query (an old changelog entry, a tag page, a near-forgotten 2023 post), the fix is retargeting or removal, and you already know the winner.

Step 3: Pick the winner with data

The winner is the page you consolidate into, so pick it with evidence, not attachment. Score both pages on four things:

• Clicks and impressions for the query, from the same GSC Pages tab you used in step 1. The page Google already prefers starts ahead.
• Links pointing at each page. Run both URLs through our free internal link analyzer to see which one your own site already endorses, and check external links in whatever backlink data you have.
• Engagement and conversions, from your analytics. A page that ranks slightly worse but converts twice as well is usually the better keeper.
• Content quality and freshness. Which page would you honestly send someone? That page, updated, is almost always cheaper than dragging the weaker one up to standard.

This weighing is the step CrawlRaven does for you: its topic clusters flag the collision, and the join puts each page's GSC clicks, GA4 engagement and crawl health side by side, so the winner is an evidence-backed pick rather than a coin flip.

Step 4: Choose the right consolidation

One split, one fix. Match it to the situation rather than reaching for the same hammer every time:

• Merge and redirect when the pages are near-duplicates and one is clearly stronger. Fold the loser's best sections into the winner, then 301 redirect the loser. This is the default fix, and the most permanent.
• Canonical when both URLs must stay live. A canonical tag points signals at the primary while keeping the secondary reachable. Know its limit: it is a hint, and Google ignores hints that contradict visibly different content.
• Retarget the loser when the two pages have different jobs but drifted onto one query. Rewrite its title, H1 and overlapping sections toward the query it should own, and cut the sections that trespass on the winner's.
• Repoint internal links when the right pages exist but the wrong one ranks. Anchors that name the query should point at the intended winner, everywhere.

The mechanics of a clean merge

Merging is the fix people rush, and rushed merges leak the value the merge exists to keep. In order:

1. Move the loser's unique value first. Sections, data, examples and FAQ answers the winner lacks get folded in as real content, not appended as an afterthought.
2. 301 the loser to the winner, to the page itself, never to the homepage or a category hub. A redirect to a vaguer page reads as a soft 404.
3. Update every internal link that pointed at the loser so it points at the winner directly. Chains through the redirect leak crawl time and clarity.
4. Leave the redirect in place permanently. Old links and bookmarks keep resolving, and removing it later reopens the 404.

Step 5: Aftercare, then a 28-day window

Consolidation is a change Google has to notice, digest and re-rank, so the last step is patience with instrumentation:

1. Record the date of the fix against the query, in a sheet or as an SEO annotation kept against your traffic timeline, as CrawlRaven's annotations do.
2. Request indexing for the changed URLs in GSC's URL inspection, then let it be.
3. Compare 28 days against the previous 28, filtered to the query. What you want first is not a better position but a calmer graph: one URL, one line, the wobble gone.
4. Then push the survivor. A freshly consolidated page is very often sitting just off page 1 with everything it needs to climb. That handoff is exactly the loop in our striking distance keywords guide.

Tips from splits I have untangled

• Fix the biggest split first, not the easiest. Rank candidates by the query's total impressions across both URLs. One consolidated money query beats ten tidied long-tails.
• Check what else the loser ranks for before merging. Its Pages-tab twin, the Queries view filtered to the losing URL, tells you what the 301 will hand over. A loser with its own healthy queries should be retargeted, not merged away.
• Publishing calendars cause most of it. Yearly "best X" posts, near-identical service pages per city, and a blog that revisits topics without checking the archive are the usual suspects. The prevention is a keyword map: one query, one page, checked before anything new is briefed. The cure for the accumulated backlog is a proper content pruning pass.
• Titles do more than you expect. Two pages whose titles both lead with the same phrase invite the swap. Often sharpening the loser's title alone settles which page Google shows.
• Do not fix a split during a core update. Rankings are already in motion; you will not be able to attribute anything. Annotate, wait for the dust, then consolidate.

Why the pages are still trading places

Consolidated a month ago and the wobble persists? Work through these four, in order:

• Is the canonical being ignored? Check the secondary URL in GSC's URL inspection: it reports which canonical Google selected. When Google picks itself over your tag, the pages are too different for a hint, and the honest fix is a merge or a retarget.
• Do internal links still endorse the loser? Rerun the internal link check. Templates, footers and old related-post modules keep resurrecting links you thought you moved.
• Did the retargeted page actually let go? If its old sections still answer the winner's query, Google still sees two candidates. Retargeting means removing the overlap, not just adding a new focus on top of it.
• Is a third page joining in? Recheck the Pages tab. Sites that cannibalize once usually have a template or a calendar habit doing it elsewhere, and sometimes the fix promoted a new contender into the swap.

Tools that make this easier

Everything above runs on Search Console and a spreadsheet. The cost is the sweep: rechecking every multi-URL query, every quarter, by hand. Where tooling picks up the load:

• Free, on this site: the GSC regex generator for topic-group and brand filters, the internal link analyzer for the winner decision and the aftercare check, and the SERP preview when a retargeted title needs to fit before it ships.
• CrawlRaven, where the sweep is standing: the keyword map holds one page per target query, topic clusters flag collisions as they appear in your Search Console data, and the join attaches the winner-picking evidence: clicks, GA4 engagement and crawl findings for both contenders in one view.
• Honest alternatives: SEO Gets has a dedicated keyword cannibalization tool over GSC data, and any rank tracker that records the ranking URL will show you the swap. Both surface the split; the diagnosis and the winner call stay manual.

However you run it, make the sweep recurring. Cannibalization is a byproduct of publishing, so every quarter of new content plants the next split. Catch it while it is one query, not a pattern.