What happens when two of our pages target the same question?
Search systems pick one, it may not be the one you want, and the choice can rotate — which usually shows up as unstable rankings rather than as an obvious duplicate problem. Fix it by deciding which page owns the query, differentiating the other so it answers a genuinely different question, and pointing internal links with matching anchor text at the page you chose. Consolidate only when neither page is distinct.
How to spot it
In Search Console, filter to a single query and look at the pages tab. If two or more of your URLs appear for that query, and especially if the top-ranking URL for it changes across weeks, that is cannibalization. The position line for the query looks noisy rather than trending.
It rarely announces itself as a duplicate content problem. It presents as a page that keeps almost ranking and never settles, which teams usually misdiagnose as needing more content — often by writing a third page on the same question.
Why answer hubs are especially prone to it
Questions asked in different words are frequently the same question. "How long does a water heater last" and "when should I replace my water heater" feel distinct in a planning spreadsheet and are close to identical in search intent.
As a hub grows past a hundred pages, nobody remembers everything already published, and the queue starts generating near-duplicates from the same underlying demand. This is a structural problem, not a discipline problem, and it needs a structural fix. Any answer hub past a few hundred records needs a question registry the way a codebase needs a symbol table.
Differentiate, consolidate, or canonicalize
- Differentiate when both pages have real distinct substance that got blurred. Sharpen each to answer its own question and change the titles to match. Preferred outcome when it is available.
- Consolidate when one page is clearly weaker and adds nothing. Merge its unique material into the strong page and redirect. Keep the URL with the stronger history.
- Canonicalize when both pages must exist for user reasons — a service page and an answer covering overlapping ground — but only one should compete. Use the canonical to point at the intended winner and keep internal anchor text consistent with that choice.
Prevention is a registry, not a rule
Keep a single list of every question the hub has answered, with its slug and its primary query. Before anything enters the queue, check it against that list — including semantic near-matches, not just exact string matches.
This is trivial to automate and nearly impossible to sustain by hand past a few hundred pages. It is one of the highest-value pieces of machinery in a content program and it is almost always the last thing anyone builds, usually after discovering forty overlapping pages during an audit. Building it into the publishing pipeline at the start costs a fraction of the cleanup.
Topics: cannibalization · duplicate questions · consolidation · internal links
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.