Canonical URLs

When the same content is reachable at multiple addresses, search engines split ranking signals across duplicates. Canonical URLs consolidate indexing onto a preferred version through the HTTP Link header with rel="canonical", delivering the signal at the protocol level before the HTML body arrives.

Table of Contents

Usage

The same page often exists at several URLs. Protocol variants (http:// vs https://), www and non-www hostnames, trailing slashes, query parameters for tracking or sorting, print-friendly versions, and syndicated copies all create duplicate URLs pointing to identical or near-identical content. Without a canonical declaration, search engines decide which URL to index on their own, splitting ranking signals across the duplicates.

The rel="canonical" link relation names the preferred URL. Search engines treat the declaration as a strong signal for consolidating indexing, link equity, and ranking signals onto the canonical URL.

Link: <https://example.re/page>; rel="canonical"

Three delivery methods exist: the HTTP Link header, the HTML <link rel="canonical"> element, and inclusion in an XML sitemap. All three carry weight with Google. The HTTP Link header is the focus here because the header operates at the server level and reaches crawlers before any HTML parsing begins.

Early signal delivery

The HTTP Link header travels with the response headers, so the canonical URL is available before any HTML parsing begins. Header-level delivery keeps the signal consistent across page templates and covers resources without an HTML head.

Consolidation itself is where the crawl efficiency lives. Pages behind query parameters, session IDs, or tracking codes often return the same content as the canonical. Once search engines settle on the canonical, they concentrate crawling there instead of recrawling every duplicate variant.

A server can also place the canonical Link header in a 103 Early Hints response, sent before the final response is ready.

HTTP/1.1 103 Early Hints
Link: <https://example.re/page>; rel="canonical"

Crawlers gain nothing from this today. Per public statements from Google's search relations team, Googlebot ignores the 103 interim response and waits for the subsequent final response. A canonical in an Early Hints response is a forward-looking option at best. The canonical needs to appear in the final response headers or the HTML to count.

Non-HTML resources

PDFs, images, downloadable files, and API responses have no HTML <head> element. The HTTP Link header is the only method for declaring a canonical URL on non-HTML resources.

A PDF accessible at multiple URLs uses the Link header to point crawlers to the preferred version.

HTTP/1.1 200 OK
Content-Type: application/pdf
Link: <https://example.re/report.pdf>; rel="canonical"

A common pattern is canonicalizing a PDF to its dedicated download page. The PDF itself is the raw file, but the download page provides context, metadata, and internal links. Pointing the PDF's canonical to the HTML download page consolidates indexing signals onto the page with richer content.

HTTP/1.1 200 OK
Content-Type: application/pdf
Link: <https://example.re/reports/annual>; rel="canonical"

The download page at /reports/annual becomes the indexed URL. The PDF stays accessible at its direct URL but drops out of search results in favor of the HTML page.

Server configuration handles this without modifying the files themselves. Nginx add_header, Apache Header set, and CDN edge rules inject the Link header at the infrastructure level.

Cross-domain canonical

The rel="canonical" link relation works across domains. The canonical URL exists on a different hostname. Syndicated content, white-label pages, and content distributed across partner sites use cross-domain canonical to consolidate indexing signals back to the original publisher.

Link: <https://original.example.re/article>; rel="canonical"

Cross-domain canonical is a strong consolidation signal. The target domain accumulates the ranking value from all syndication URLs pointing to the canonical. The syndicated copies still appear online but drop out of search results in favor of the canonical.

Conflict resolution

When the HTTP Link header and the HTML <link rel="canonical"> element declare different canonical URLs for the same page, the outcome is unpredictable. Both methods carry equal weight as signals. Search engines may follow one declaration, the other, or ignore both entirely and select a different URL based on other signals like redirects, internal link patterns, sitemap URLs, HTTPS preference, and hreflang cluster membership.

Using one canonical method per page avoids ambiguity entirely. If server-level configuration sets a Link header canonical and the CMS injects a different HTML canonical, the mismatch hands the decision to the search engine with no guarantee about which URL, if either, wins.

Common mistakes

Missing self-reference. Every page benefits from a self-referencing canonical pointing to its own preferred URL. Without one, search engines rely entirely on other signals to pick the canonical.

Canonical pointing to a non-200 page. A canonical URL returning a redirect, 404, or 410 invalidates the declaration. The canonical target must return 200.

Canonical on paginated content. Each page in a paginated series self-canonicalizes to itself. Pointing all pages to page one hides pages two and beyond from the index.

Canonical combined with noindex. A page with noindex and rel="canonical" sends conflicting signals. The noindex tells search engines to drop the page. The canonical tells them to consolidate onto the page. Pick one.

Relative URLs. Google recommends absolute canonical URLs. Relative references are valid Link header syntax, but a relative path invites parsing ambiguity and risks the canonical resolving to the wrong URL.

Canonicalizing to unrelated content. The target URL must contain content identical or nearly identical to the source. Pointing a product page canonical to the homepage is a signal search engines are likely to ignore, selecting their own canonical instead.

Example

A product page accessible with and without query parameters. The canonical Link header consolidates signals onto the clean URL.

HTTP/1.1 200 OK
Content-Type: text/html
Link: <https://example.re/products/widget>; rel="canonical"

Both https://example.re/products/widget?ref=email and https://example.re/products/widget?sort=price return this same canonical header, pointing search engines to the parameter-free URL.

A 103 Early Hints response carrying the canonical early, with the final response repeating the header for crawlers.

HTTP/1.1 103 Early Hints
Link: <https://example.re/products/widget>; rel="canonical"

HTTP/1.1 200 OK
Content-Type: text/html
Link: <https://example.re/products/widget>; rel="canonical"

A PDF hosted at multiple URLs with a canonical Link header declaring the preferred version.

HTTP/1.1 200 OK
Content-Type: application/pdf
Content-Disposition: inline
Link: <https://example.re/docs/guide.pdf>; rel="canonical"

A response combining canonical with resource hints in a single Link header.

HTTP/1.1 200 OK
Content-Type: text/html
Link: <https://example.re/page>; rel="canonical", </css/main.css>; rel="preload"; as="style", <https://cdn.example.re>; rel="preconnect"

See also

Last updated: August 17, 2026