Fixing keyword cannibalization: assigning a practice's pages clearly to one search intent each

Let's imagine a therapy practice, anonymous and of average size. The website has grown over the years. First there was a service page, later a blog post was added, then a page for a new offering, and finally a collection of FAQs. Each addition made sense on its own. Taken together, however, five pages essentially answer the same question: what happens during the treatment, and who is it suitable for?

A typical starting point looks like this. In Google Search Console, an important query shows not one URL but several, and they take turns. Today the service page leads, next month the blog post, then the FAQ. The position fluctuates, the click-through rate stays mediocre, and nobody on the team can say which page is actually responsible.

The pattern is known as keyword cannibalization. The term is somewhat misleading, because there is no penalty mechanism. The pages simply share signals such as links, references and click behavior, and the search engine has to guess which one to display. For a practice that maintains little content anyway, that is wasted visibility.

This is particularly annoying for queries with a clear intent to act. Someone searching for an intro call who lands on a guide page that offers no way to request an appointment bounces. Someone searching for information who lands on a booking page without explanation does the same. The wrong page in the right place costs twice: once the opportunity, and once the trust of those who are just getting oriented.

With practice services, overlap is almost inevitable. Patients rarely search for the technical term. They phrase complaints, questions and reservations, and each phrasing seems to justify its own page. The practice responds with more content, and that inevitably overlaps with what already exists.

There is also a legal particularity. The Heilmittelwerbegesetz sets narrow limits on advertising for remedies and treatments. Promises of cure, guarantees of success and suggestive before-and-after depictions are taboo. In practice, this leads to cautious, often interchangeable wording, and interchangeable texts are the raw material for cannibalization. Anyone who only describes what is offered ends up writing the same thing several times.

Then there is the question of responsibility. The person who changes texts on the website is often not the same person who set the structure years ago. Without a documented assignment of page to search intent, every new post quietly creates another competing offer. The problem is rarely a single wrong decision, but the lack of a shared map.

Another reason lies in how technical language is handled. Professionals write for professionals, while searchers phrase things in everyday language. The same treatment then appears once with its technical term, once as a description of complaints, once as a question about costs. Without coordination, three pages emerge that, for the mapping, would really be a single row with three secondary questions.

Before anything is merged, a diagnosis is needed. The simplest route leads through the performance report in Google Search Console. If you filter by a query and switch to the page view, you see which URLs receive impressions for that query. If several pages appear there with notable shares, the suspicion is justified. A time series per URL additionally shows whether they alternate.

It is important not to read every multiple listing as an error. A page can be visible for a query without competing for it, for instance if it only appears at the margins. What matters is the search intent. If someone is asking informationally how a treatment works, and a booking page appears alongside, it simply does not fit the intent. That is a different problem from two pages serving the same intent.

For classification, it helps to look at the results page itself. Which page types does Google list for the query? Do guides, directories, service pages or local listings predominate? This composition reveals what kind of page the search engine expects for this intent. It determines which of the competing pages should take on the right role.

Finally, it is worth looking at the pages themselves. Does one of them carry noticeably more incoming links, has it collected references over the years, or is it simply written more thoroughly? Such differences are rarely coincidence and often give the best hint about which page is suitable as a foundation. The newer page is not automatically the better one, and the older one is not automatically worth protecting.

A principle follows from the diagnosis: every search intent gets exactly one responsible page. This is not a requirement of the search engine but a working rule that simplifies decisions. We distinguish four intents. Informational means someone wants to understand something. Transactional means someone wants to book an appointment. Navigational means someone is looking for the practice or a specific section. Local means someone is looking for a practice within reach.

For every doubly occupied intent there are three paths. Merge, when both pages have the same job and one offers the stronger foundation. Redirect, when a page has no purpose of its own after merging. Differentiate, when both pages are legitimate but should serve different intents, and only the texts have become too similar.

The alternative would be to simply keep everything and steer with canonical tags. That only works with nearly identical content and merely tells the search engine which variant is preferred. It does not solve the substantive problem that two pages want the same thing. Deleting the content everywhere would be just as wrong. Lost pages are lost entry points, and some carry links you want to keep.

An example makes the trade-off tangible. A practice has a blog post on how a treatment proceeds and a service page that repeats the same process in short form. Both want to inform. The decision is made by leaving the service page the role of overview and covering the process exclusively in the post it links to. Both pages remain, but each now has its own job.

The central tool is a mapping, at its core a table. Each row stands for a search intent, not for a page. Alongside it sit the responsible URL, the type of intent, the target audience, the next sensible step for the reader and the person responsible for editing. That sounds sober, but it is the part that keeps the structure stable in the long run.

The hub-and-spoke architecture builds on this. A hub page covers a service area comprehensively and points to in-depth pages, the spokes. The hub page serves the broad query, the spokes the specific questions, for example about procedure, duration, cost coverage or preparation. What matters is the division of labour. The hub explains and guides onward, while each spoke goes deep on exactly one topic.

For practice services, the language has to stay factual. A page may describe how a treatment works, which complaints can be a reason to seek it and what an intro call looks like. It must not promise that complaints will disappear. Recording this in the mapping from the start, for example in a review column for wording, avoids later corrections. It does not replace a legal review. That is a matter for the practice or its advisers.

Technically, the work begins with an export of all URLs including title, H1 and indexing status, for example with Screaming Frog. Data from Search Console is added. Together they form the basis for the mapping. It is important to secure the starting point before any change, so that the effect can be assessed later at all.

When merging, the content of the weaker page is carried over into the stronger one, as far as it contributes something new. The weaker URL receives a server-side 301 redirect to the target page, directly and not via intermediate stops. The sitemap and internal links are adjusted so that nothing points to the old address any more. This saves crawl effort and prevents chains.

Internal links are the steering instrument that is underestimated most often. Every page should be linked from suitable neighbouring pages, with anchor texts that describe the target. “Read more” or “click here” tells neither readers nor search engines anything. A link with the text “How the initial examination works” does. And if several links to different pages carry the same anchor text, you create exactly the ambiguity you wanted to remove.

After the restructuring comes the check. In Search Console you can verify whether the target pages are indexed and whether the old URLs redirect correctly. A fresh crawl uncovers redirect chains, forgotten internal links and pages without incoming references. This follow-up work usually takes longer than the actual restructuring, but it decides whether the new order really arrives in the index.

Not every overlap can be resolved cleanly. A common special case is regional variants. If the practice has several locations, pages with only the place name swapped are nearly identical. Merging does not help here, because the local intent is legitimate. Each location page needs its own substantial content, such as team, directions and particulars of the premises, and should only stay where it offers real added value.

A second case is pages that deliberately serve different target groups but describe the same service, for example for adults and for parents of children. These are separate intents that should differ in their language. Often it is then enough to sharpen the title, the opening and the examples. Merging would serve both groups worse.

The third case concerns pages that have links or references from outside, for example from directories. They should never be removed without a replacement. A redirect to the matching page passes on signals to a large extent, though whether completely cannot be guaranteed. Finally, for medical statements: after every merge, check that the new version is nowhere worded more sharply than the original pages. When combining pages, wording that suggests a promise creeps in easily.

A fourth case is pages that are not supposed to rank at all, such as privacy policy, legal notice or internal forms. They appear in the mapping so that it is clear they serve no intent and do not compete for enquiries. Where they do show up in results, that is more a sign of content pages that are too weak than of a problem with these pages themselves.

Do not expect a sudden improvement, but a calming. When only one page is responsible per query, fluctuating positions should become rarer, and the page that appears should fit the intent better. Whether the position rises as a result depends on competition, quality and authority and cannot be promised. This should be stated clearly, even if it sounds less compelling as a sales pitch.

Measurement happens in Search Console. For the affected queries, you compare impressions, clicks, average position and the number of URLs that appear, each before and after the restructuring. A sensible observation period covers several weeks, because Google only processes changes after crawling and evaluating them again. Seasonal fluctuations should be cushioned by comparing with the same period of the previous year.

Soft signals count as well. Fewer questions at the front desk because the page answers the most common question, or a clearer path from information to appointment request, are plausible effects that can be observed through form hits and calls. A forecast in numbers would be unserious. Only the before-and-after comparison with your own data is reliable.

Even the best reorganisation will not hold if content is added freely afterwards. That is why a simple editorial plan is part of the result. It contains no quota of posts but a rule: before every new text, check in the mapping whether the intent is already taken. If so, the existing page is extended rather than a new one created.

A second principle concerns maintenance. Every page in the mapping gets a review date, for practice pages roughly every six months. The check covers whether the statements are still correct, whether prices, times and procedures are up to date and whether the wording stays within the legal framework. This takes little time and keeps pages from becoming gradually outdated and contradictory.

A finding that can be generalised from comparable tasks: cannibalisation is rarely just an SEO problem. It is usually a responsibility problem that shows up in the search engine. Whoever documents the structure and assigns it to a person solves not only the current problem but prevents the next one. The mapping is unremarkable but works for years.

For a practice that notices that several of its own pages rank poorly, the order is decisive: first organise, then expand. Piling more content onto an unclear structure amplifies the problem. A clear mapping, by contrast, shows where there are real gaps and where there are merely duplicates.

The practical start is manageable. Export from Search Console the queries that matter for your practice and check, for each query, which pages appear. Where several show up, assign the intent and decide between merging, redirecting and differentiating. Even a handful of such decisions creates a solid basic structure.

This scenario describes an approach, not a documented project result. It shows how the task can be structured and which questions have to be answered along the way. Whether and how much your visibility changes depends on your environment and cannot be quantified in advance. The only certainty is that a clear responsibility per question makes your website easier to understand, for search engines and for people alike.

If you would rather not do this work alone, you can outsource parts of it: the diagnosis from the data, the mapping as a shared document, the technical implementation of the redirects. The content decisions and the review of medical statements should, however, stay with the practice, because that is where the expertise and the responsibility lie.