Compare the answers before touching signals
Open both existing URLs and read the main content side by side. Compare the buyer question, product or version covered, conditions, supporting evidence and next action. Matching layouts or titles alone do not establish that two pages answer the same question.
Google’s canonicalization troubleshooting guide says “Google might choose a different canonical” because of content quality or technical signals, and asks whether its selection makes more sense for search users. That makes a different URL a reason to investigate, rather than proof that the article needs correction.
For example, two paths that deliver the same compatibility answer may reasonably share one representative page. An overview and a detailed answer for a specific configuration may have different reader purposes, even if they share an introduction. Record the substantive difference; do not try to make pages distinct by changing only their headings. A category answer gap workbook can help identify which buyer answer needs its own treatment.
Put competing targets in one comparison
Create one worksheet for the duplicate group, with a separate entry for each URL. Note your preferred URL, the reported selection and the observation date above the table. Copy exact destinations instead of marking a setting as simply present.
| Signal | Where to check | Consistent? |
|---|---|---|
| HTML canonical | The delivered HTML head; check rendered HTML if scripts can alter it | Each duplicate names the intended representative; no competing target appears |
| HTTP canonical | The response Link header, where present | Its target agrees with the intended representative and any HTML declaration |
| Redirect | The response status and every destination through to the final page | The route does not send readers toward a competing representative |
| Internal links | Navigation, article links and other links to these existing pages | Relevant links consistently use the intended canonical URL |
| Sitemap URL | The current sitemap entries for this group | Listed URLs agree with the intended representative |
In its canonical methods guide, Google calls redirects and canonical annotations “A strong signal”, and sitemap inclusion “A weak signal”. It recommends consistent internal links and warns against naming different canonicals through different methods. It also recommends choosing either HTML or HTTP canonical declarations because using both is more prone to errors; permanent redirects are appropriate when retiring a duplicate.
Work through redirects and canonical declarations first when they name different targets. Then check links and sitemap entries for remaining contradictions. An absent declaration is different from one pointing to the wrong page: label the former as absent and the latter with its actual target. Do not use a sitemap change to conceal a conflicting redirect.
Decide whether to keep or correct the selection
Keep the current choice when both URLs deliver substantially the same answer, the selected page is a suitable destination for readers, and you find no unintended conflicting target. Document why it is acceptable; a preference for a shorter path alone is a weak basis for demanding a different representative.
Open a correction task when the worksheet exposes an unintended declaration or redirect, or when the selected page fails to preserve an important difference in the answer. Describe the observed response, intended destination, affected URLs and evidence of the reader purpose. A task saying only “make Google choose this URL” does not identify a repairable defect.
Where the pages should remain distinct, explain the substantive differences that must be clear in their existing content. Where they are true duplicates, agree on one preferred representative and align the conflicting signals. The website’s technical team should implement any site changes.
Recheck the choice after the correction
Google’s URL Inspection help says to use the indexed URL Inspection result to read the “Google-selected canonical” field. Keep the user-declared value separate: it records a preference, while the selected value reports Google’s decision. The live test cannot predict which URL will be considered canonical.
After a correction, verify that the changed signal now delivers the intended target, then revisit the reported selection when Google has processed the pages again. Save the new observation date and selected URL alongside the previous result. If the choice remains different, return to the worksheet and content comparison instead of assuming that another canonical tag will settle it.
This check ends with the observed canonical choice. It does not establish inclusion in an AI answer, an indexing outcome or a ranking gain. Preparing material for getting cited by ChatGPT is a separate content and evidence task; canonical consistency cannot promise that outcome.
Want to see how AI engines describe your brand today?
Get a free growth audit