Canonicalization issues: conflicting canonicals and og:url

The edges mean different things

A redirect changes where the next request goes. A canonical annotation nominates a preferred representation. A sitemap proposes a URL for discovery, and an internal link offers a path through the publication. These edges can disagree, but they cannot all be interpreted as the same relation. A normal pair of articles linking to each other is no evidence of a canonical loop.

For each URL, preserve requested and final responses, directives, source and rendered declarations, and the content evidence used to assess equivalence. A source canonical naming A and a rendered one naming B must remain two observations. A fingerprint can suggest similarity; it cannot decide whether the distinctive paragraph, language or application state matters to the reader.

RFC 6596 makes the content test explicit: a canonical target should contain duplicate content or a superset. A later paginated component therefore shouldn’t nominate an incomplete first page. A clean graph can still conceal lost material.

Keep discovery edges when reviewing how a crawler or person reaches the pages. Restrict identity-conflict traversals to the relations being tested, such as redirects and canonical declarations. Otherwise ordinary navigation cycles will pollute the findings. A template sending every essay to the homepage is the complementary failure: no conflicting declarations, but an equivalence decision the content cannot support.

To locate a conflicting declaration, compare the raw response with the rendered DOM before choosing which URL the route should own.

Every target can exist and the system can still loop

In a worked example, /print/guide declares /guide as canonical. That address redirects to /research/guide, whose canonical points back to /print/guide. Every target exists. Three successful existence checks miss the loop because none follows the references back to their starting point.

Follow the redirect and canonical edges to expose the loop; a strongly connected component identifies its participants. A self-canonical is a valid endpoint. Several terminal representatives indicate another defect. Ordinary navigation cycles, by contrast, can help readers circulate through related pages.

Choose the document that preserves the cluster’s content and purpose. Redirect appropriate retired aliases directly there and align surviving duplicate declarations. Pagination, translations, print views, and PDFs each need an equivalence decision. When the target omits material, keep the document or find a sufficient representative.

A route record can own the public identity

The page registry should own the canonical path, its aliases, indexability and significant modification date. Redirects then derive from retired aliases, the canonical from the maintained path, the sitemap from eligible pages and navigation from the same route keys. Google describes redirects and rel=canonical as stronger signals than sitemap inclusion and recommends consistent internal links. Agreement gives those signals a coherent request to make. Canonical guidance.

A blocked duplicate may remain known even when the crawler cannot retrieve its canonical. That is why robots.txt cannot reconcile contradictory identities. Decide which documents should consolidate before choosing the directives that express that decision. Noindex requests removal. Canonicalization requests consolidation.

type PublicRoute = {
  key: string;
  canonicalPath: `/${string}` | "/";
  aliases: `/${string}`[];
  indexable: boolean;
  lastSignificantChange: string;
};

const sitemapRoutes = routes.filter((route) => route.indexable);
const redirects = routes.flatMap((route) =>
  route.aliases.map((from) => ({ from, to: route.canonicalPath, permanent: true })),
);

Several canonical generators can disagree. The multiple canonical tags capture locates them; the canonical and noindex final responses cover the indexing decision when exclusion directives are involved.

Consistency enforces the content decision

Review equivalence before testing projections. Compare each proposed representative with the source documents and record why they consolidate or stay independent. Then reject competing alias owners, canonical conflicts, redirect/canonical cycles, and sitemap rows naming unusable or ineligible endpoints. The graph enforces that content decision.

Compare visible headings and dates with JSON-LD, initial declarations with rendered declarations, and built routes with public responses through middleware and caches. Google’s selected canonical is a dated observation that can differ from both the current preference and an older crawl.

A small site may need only a route table and representative fixtures, while a larger graph earns its cost when aliases and templates obscure the cause of a conflict. I would test either approach with a perfectly consistent wrong target, asking whether it catches the lost content before accepting agreement as a result.

A domain change adds timing and rollback to these identity checks. The technical SEO migration gates follow the URL manifest from prelaunch validation to public readback.

Metadata and duplicate-page repairs

Page metadata and date precision follows the document inside the head and JSON-LD. The OG URL comparison examines the social-sharing case.

Duplicate title tags and missing canonical tags now have focused repair notes.

When og:url does not match the canonical

Read the focused comparison and browser inspector.

JSON-LD is another projection of the document

Read the focused note.

A day should not acquire an invented instant

Read the focused note.

How to fix duplicate title tags

Read the focused note.

How to fix a missing canonical tag

Read the focused note.

I would ship the URL family when its pages preserve the intended content and every identity signal points to the appropriate representative. Graph agreement earns its value by carrying the right page to the reader; a perfectly consistent wrong destination still needs repair.