Yesterday, a report for a query showed an article; today, it shows a service page. This change alone does not prove that the pages are competing with each other. The conditions under which the results were checked may have changed, the search engine may have selected a different user intent, or the data may relate to different variations of the query.
In practical analysis, cannibalization means unwanted overlap: several pages target the same user need, and the site does not provide a clear preferred answer. An investigation should identify the specific overlap and its consequences. Automatically merging every URL that appears for the same word is risky: some of those pages serve different needs.
Gather the query and page history
Choose one exact query. Record the search engine, region, device, period, and data source. Separately save variations containing words such as “price,” “buy,” and “how to choose,” as well as the service name without any qualifiers. They may be lexically similar but express different needs.
Create a table by date with the query, URL found, position when available, impressions, clicks, and measurement conditions. If you use both manual search results and a search console, do not combine them into a single numerical series. A search results snapshot captures observations under specific conditions; a console provides aggregated history for its own set of impressions.
In Search Console, most performance data is associated with the canonical URL. So before analyzing two rows, check whether they represent variations of the same document. The grouping details are described in Google’s help on report dimensions.
Save the original export. A working table with notes will help you make a decision, but without the original, you cannot reconstruct the filters or check whether some observations were lost.
Describe the user need for each URL
Open the pages and fill out a short card for each: who visits, what they want to know, what answer they get, and what next action is available. Do not focus on the title alone. Two different headings may conceal the same text and the same inquiry form.
A hypothetical article, “How to choose ventilation maintenance,” helps readers compare options. A page titled “Ventilation maintenance” describes a specific offer and its price. Their appearance for overlapping queries is acceptable if their content and links support these roles.
Another scenario: two service pages describe the same maintenance service in one city, contain the same form, and differ by just one word. In that case, a potential merge calls for a more detailed review. Check whether either page has a distinct offer, audience, or limitation that needs to be preserved.
For each pair, define the expected query split. The article should address selection and preparation; the service page should address placing an order. This is an editorial hypothesis to test against data, not a promise to the search engine.
Rule out technical duplication
First compare status codes, final URLs, canonical tags, main content, and internal links. Two URLs may turn out to be one page with a parameter, an old address left over after a redesign, or a version with a trailing slash. In these cases, the task is to align the URLs, not rewrite two texts.
Google describes canonical as a signal indicating the preferred document among identical or very similar pages. For distinct, standalone content, it is not a general instruction to “show this page for the query.” Using canonical requires a substantive reason. Google’s documentation on canonical URLs.
Check the preference signals: which page is included in the sitemap, where the navigation leads, which pages articles link to, and whether any redirect still points back to the old URL. If the site supports different URLs at the same time, resolve this conflict before evaluating the results of any edits.
Scenarios involving true duplicates are covered separately in the guide to duplicate content. Use it as a technical reference while keeping the current analysis focused on the specific pair of documents.
Assess the impact of a page change
A URL change matters if visitors get an unsuitable answer or the target page loses significant traffic. The number of URLs involved does not show the impact. Sometimes search results switch to a more suitable page without any noticeable change in the overall result.
Compare four observations: whether the URL found matches the query, whether total impressions for the group change, whether clicks are lost, and whether the available target action performs worse. Check the last item using working analytics or manually; explicitly note when no measurement is available.
A hypothetical query, “maintenance cost,” may lead to an article with no prices or link to place an order. This is a reproducible mismatch with the user’s need. But it does not prove that the article is technically suppressing the service page. In your log, distinguish the fact “the result was an unsuitable answer” from the hypothesis “the pages compete because they have the same signals.”
When data volumes are low, avoid making decisions based on a single day. Choose a control period and use a separate status, “not enough observations.” This is more useful than an arbitrary redirect of a page that users need.
Choose an action based on the type of overlap
Use a matrix tied to the actual pair:
| What you found | Action | What to check afterward |
|---|---|---|
| Two URLs for the same document | Align the primary URL and technical signals | Status codes, canonical, links |
| Two pages with the same user need and no differences | Consider merging while preserving useful content | Completeness of the answer and transition from the old URL |
| Different user needs, but nearly identical text | Differentiate the content and next actions | Clarity of each page’s role |
| Different user needs and useful answers | Keep the pages and clarify navigation | Whether links match expectations |
| A change was observed once | Continue monitoring | Whether it recurs under the same conditions |
When merging, list what must not be lost: instructions, specifications, support answers, images, and links. Prepare the target page first, then configure the redirect. Do not remove the second page before checking the resulting answer.
When separating page roles, rewrite the main sections to address the user’s need, not just the title. Two pages with different headings but the same content remain difficult to explain to both visitors and editors.
Give the editor and developer a change map
For each page, record its target user need, primary query group, content to preserve, sections to change, and the user’s next step. Add the old and final URLs, the expected HTTP status, and a list of internal links to update.
Review the internal linking strategy: a link with wording about placing an order should lead to a page where the offer is available, while a link about choosing should lead to a page with comparisons. This is not a universal ranking formula; it is a way to make the site architecture consistent.
A reChecker site audit can check technical findings across the pages collected. An editor familiar with the offer should confirm the content overlap and whether the merged answer is equivalent. A tool does not replace the decision about the purpose of each document.
Evaluate the changes separately from their search impact
Technical acceptance is immediate: the target page is accessible, the old URL is handled as planned, links are updated, canonical is aligned, and useful materials are preserved. Editorial acceptance checks whether the content makes the distinction between the pages that remain clear.
Continue monitoring search results for the same query and under the same conditions. Record the release date, URL history, and data limitations. If metrics change, check for other events during the period instead of attributing the entire effect to one merge.
The investigation should conclude with a decision about the specific pair, supported by reasons and an acceptance test. The decision may be to merge the pages, separate their roles, or keep them as they are. None of these options guarantees that the search engine will select the desired URL or improve its position.