NovCannibalisation is about overlapping jobs, not repeated words
Two pages can compete even with different titles when they promise searchers essentially the same answer.
This article reflects the named expert’s practical perspective. See NovAsia’s editorial policy for how material is prepared and reviewed.
Cannibalisation is often diagnosed with a vocabulary test: find pages that mention the same keyword and assume they are competing. That catches some obvious duplicates, but it also produces a lot of false alarms.
Pages can share terminology because they belong to the same subject while doing completely different work. The market-data and economy-and-finance sections inevitably use some of the same language, yet a reader looking for observed market indicators is not asking exactly the same question as someone looking for economic context.
The more useful unit of analysis is the page job. Is this URL an overview, a comparison, a process guide, a data page or a project record? What does the reader expect to know or decide when the page is finished? Two pages become genuinely difficult to distinguish when those answers are nearly identical, even if the titles have been made linguistically unique.
A typical problem looks like this: an older guide explains how foreign buyers purchase property in a country, then a newer article is created around a slightly different keyword but ends up covering the same ownership rules, documents, payment steps and cautions. Both pieces may be well written. The issue is that the site has created two candidates for one answer.
Query data can help reveal that overlap. If the same group of searches repeatedly surfaces different URLs over time, it is worth inspecting the content and internal linking. But alternating URLs are not proof on their own. Search results vary by wording, device, location and many other factors, so the diagnostic has to come back to the actual function of the pages.
Internal links are often where the ambiguity becomes visible. If half the site points to one guide as the main answer and the other half points to a second guide using similar anchor language, the architecture is undecided. A rewrite that adds a few different examples will not resolve that conflict.
Once the roles are clear, the solution can vary. One page may become the broad definitive guide while the other is narrowed to a specific scenario. Two weak pages may be consolidated into one stronger resource. An obsolete page may be retired and redirected once its useful content has been absorbed elsewhere. The technical action should follow the editorial decision, not replace it.
Canonicalisation is not a substitute for this thinking. A canonical hint can help search engines understand preferred duplicates in appropriate situations, but leaving two near-identical user-facing answers in place still leaves readers with a messy architecture. If both pages deserve to exist, they need distinct jobs. If they do not, the content model should acknowledge that.
Cannibalisation is therefore less about a forbidden amount of keyword overlap and more about ownership. A healthy site can say, with reasonable clarity, which URL owns the broad answer and which URLs handle the narrower branches. When that ownership is explicit, shared vocabulary stops looking like a problem and duplicated intent becomes much easier to spot.