Method
We fetched seven public URLs once from the Hermes VPS and parsed the server-rendered HTML. The collector recorded every rel=canonical value without executing client-side code. This checks the signal a crawler receives in the first response, not a browser extension or an SEO tool's interpretation.
We compared the declared canonical with the final fetched URL. A match is treated as self-referencing. A mismatch is recorded as an implementation defect, not proof that Google selected a different canonical.
Observed evidence
https://sitesignalhub.com/Self-referencing and expected.https://sitesignalhub.com/Incorrectly pointed to the homepage.https://sitesignalhub.com/Distinct article identity was not declared.https://sitesignalhub.com/articles/software-renewal-reviewThis template already emitted its own URL.Findings
- Six sampled non-home routes inherited the root layout's homepage canonical.
- Titles and headings were unique, so the canonical conflict came from metadata inheritance rather than duplicate titles.
- The automated article template behaved differently from the fixed guide template, making the defect inconsistent across content types.
Decision and resulting changes
Remove the site-wide canonical default and require every indexable route to declare its own canonical. Legacy articles under revision should declare their URL but remain noindex until their evidence is upgraded.
- Removed the homepage canonical from global metadata.
- Added route-specific canonicals to indexable pages.
- Added canonical assertions to rendered HTML tests.
Limits and reproduction
Re-run automation/evidence_collector.py against the public origin and compare the resulting JSON fields. A difference is evidence of a changed response, not automatically an improvement.
- This audit records emitted HTML; it does not claim which canonical Google ultimately selects.
- Search Console validation must be checked after recrawl.
Primary references
— Baseline captured from the public site, limitations documented, and remediation decision published.
Prepared by SiteSignal Hub Field Lab. Evidence collection and structural checks are automated; no personal use, customer result, or human test is claimed.