Blog · 2026-08-30

Operating on Keyword Cannibalization: 35 Terms Fighting Across 6 Pages, Fixed Without Touching the Money Page

A field report from a content site in a high-competition industry: we mapped the cannibalization from GSC query-by-page data, assigned one term to one page, used an internal-link anchor lever to fix 26 anchors by editing a single title, and never touched the money page's home turf.

Once a content site passes a hundred pages, the search report tends to develop a familiar illness: several pages rank for the same keyword, all of them stuck on page two or three, and the clicks get sliced up. We hit a textbook case on a content site in a high-competition industry, with 35 queries spread across 6 pages fighting each other, a head term with real search demand, and not one page assigned to answer it. This is the full record of the operation.

Map It First: The GSC Query-by-Page Cross Tab

You cannot catch cannibalization by intuition. The method is crude and it works: export the last four weeks of the Google Search Console (GSC) performance report along two dimensions, query and page, and flag every query that picked up impressions on two or more pages. Once this site was mapped, the illness looked like this:

• 35 queries spread across 6 pages, with 4 of those page titles all chasing the same promotional term. Which page was supposed to rank for it was something the site itself could not answer.
• The head term averaged position 23.7 and still collected 9 clicks. Demand was strong enough that people dug it out on page three, yet no page was built to catch it.
• The most ironic pairing: page A had 33 internal anchors pointing at it and got only 8 impressions, 0 clicks, position 37, because the term was nowhere in its title. Page B was the best-converting page on the site with 572 impressions and 55 clicks, but not one of its 22 inbound anchors referred to it by the head term.

With the map done, the conclusion surfaced on its own: there was plenty of content, and the division of labor was a mess. Titles, anchors and target pages all contradicted each other, so the search engine had no choice but to split authority across six pages, and all six lost.

One Term, One Page: Assign the Roles Before Editing a Word

The fixing principle is one sentence: every query gets exactly one page assigned to it, and the rest stand down. Here is the assignment table for this site:

• The head term went to the comparison page, whose title was rewritten to answer "how to choose" directly, so those 33 anchors would finally pay off.
• The money page kept the two terms it was already winning and gave up the head term. Its job is conversion, not covering everything.
• The calculator page had the duplicate terms stripped out of its keywords and now targets calculator intent only.

Assignment does not always mean stepping aside, it can also mean splitting by intent. On the same site, 7 pages were ranking for the brand term, and two feature pages between them turned 3088 impressions into 6 clicks (a CTR of 0.2 percent). The cause was titles that buried the point in a trailing suffix, so someone searching the brand could not tell what the page answered. The fix was not to delete pages, it was to front-load the brand term and give each page one intent: one answers "how to join", the other answers "what this is", so a single search takes two slots.

The Anchor Lever: Edit 1 Title, Fix 26 Anchors

Most people fix internal links by inserting them one at a time by hand. Stop and go read the template first. On this site, the "related articles" card renders the anchor text straight from the target page's title, and the new page for the head term was listed as related on 26 pages. Which means editing one title fixes 26 internal anchors at once. Measured before the edit: 0 anchors contained the head term. After: 26 out of 26.

Three supporting lessons:

• Check the template's render limit before touching internal links. This site's related list renders only the first 3 entries, but 5 articles had 4 configured. The fourth would never appear in the HTML, so configuring it did nothing.
• One page with 198 impressions at position 10.3 had exactly 1 inbound link on the entire site. After rebalancing the related lists it went from 1 to 11, and two other orphan pages went from 1 to 7 and 1 to 6.
• Always verify by counting href values in the built HTML, not by reading the config file. Config files lie, rendered output does not.

lastmod Is Only Allowed to Tell the Truth

There is one more gate after the title edits: the sitemap. This site used the publish date as the lastmod value, so pages with edited titles still carried a lastmod from four months earlier. The search engine had no reason to prioritize a recrawl, and the edits may as well not have happened.

Neither shortcut is acceptable: rewriting the publish date is lying, and stamping today's date across every lastmod on the site is worse, because the search engine learns quickly to stop trusting your sitemap. The correct fix is to split "published date" and "last modified date" into two fields, with the sitemap reading the latter and falling back to the former when it is absent. Verified after deploy: only the 3 pages that genuinely changed moved their lastmod to that day, and everything else stayed untouched.

One more detail that is easy to get wrong: internal link adjustments are navigation changes, not content changes, so do not bump the modified date along with them. The value of lastmod is that it is a trustworthy signal, and every time you water it down it is worth less.

The Money Page Red Line: Minimal Changes Plus Baseline Monitoring

The biggest risk in this whole operation was never failing to fix the cannibalization, it was breaking a page that already earns. The best-converting page on the site had 55 clicks over the last four weeks at a stable position, so the only edit we made was the part of its title that gave up the head term. Not one word of the terms it was winning was changed, and moving the promotional and calculator terms around was deliberately routed away from its home turf.

Two things happened on the day of the operation: the money page's position and click baselines went into the monitoring list so we could watch for any drop, and acceptance was measured strictly on four weeks of GSC data against the previous four weeks, never daily numbers. Daily swings scare people into bad decisions. Put plainly, the point of operating on cannibalization is to pull the sliced-up clicks back together, not to rack up a large number of visible edits. Cut wrong and damage the money page, and every impression you saved will not cover it.

Closing

The whole method condenses into one flow: map it with query by page, assign one term to one page, find the lever and edit there (one title equals 26 anchors), then let an honest lastmod invite the search engine to come and check the work, all while staying off the money page's home turf. No magic, just discipline about three things: data, assignment and verification. If your site also has a pile of terms stuck on page two or three, or you suspect pages are fighting each other but are not sure where to cut, come talk to us. We tend to draw the map first and discuss surgery second.

We Turned This Kind of Surgery Into a Service

The hard part of operating on keyword cannibalization is not finding the overlap. It is deciding which page stays, which page gets merged, how the internal links are rewired, and doing it without losing the rankings you already have.

• Content audit and consolidation: identify the pages fighting each other, decide the page to keep and the 301 redirect target
• Internal link restructuring: concentrate anchors and inbound links on the page you keep, leaving no orphan pages
• Technical fixes: align canonical tags and the sitemap so the site stops competing with itself

SEO build and monthly maintenance plans:

Specs are on iGaming SEO Site Build.