Key takeaway
For complete translations intended to be indexed separately, identify each language’s canonical URL and connect published equivalents with reciprocal hreflang. Use canonical to choose among same-language duplicates and prefer full HTTPS URLs. These signals do not guarantee indexing.

A few months after launching a bilingual corporate website or standalone overseas site, many teams run into a frustrating mismatch: Chinese pages appear in search results as expected, while carefully translated English service and product pages stay invisible—or international brand searches still surface Chinese links. Teams often blame English copywriting first, only to discover the cause sitting inside <head> metadata or XML sitemaps: the English page's rel="canonical" points back to the Chinese source, hreflang links only point one way, or unfinished 404 placeholders sit inside the language map.
To clearly tell search engines which bilingual pages correspond, separate what canonical and hreflang do: each complete translation intended for separate indexing should identify its canonical URL in that language—never cross-language to Chinese—while bidirectional, self-inclusive hreflang annotations connect verified translations and omit untranslated pages.
The examples below concern complete translations intended for separate publication. Google supports relative canonical URLs but recommends absolute ones to reduce configuration ambiguity; hreflang requires fully qualified URLs. Self-canonicalization is a common pattern, not a rule that every duplicate URL should point to itself.
What Canonical and Hreflang Control—and Why Cross-Language Canonicals Backfire
- Chinese canonical page
- English canonical page
- Each identifies its canonical URL
- Link completed language equivalents
Both tags sit in <head> markup, yet they solve two entirely different indexing problems:
rel="canonical"consolidates duplicate URLs for the same content. Per Google Search Central's guide on consolidating duplicate URLs, when identical content appears at multiple URLs—such as tracking parameters (?utm_source=...) orhttpvs.httpsvariants—canonicaldeclares the single preferred URL. Preferred pages themselves should carry an absolute self-referentialrel="canonical".hreflangroutes searchers across localized equivalents. According to Google Search Central's documentation on localized versions, fully translated pages in different languages are not considered duplicates;hreflanglinks localized variations together so Search serves the version matching each user's language and region.
Once those roles are clear, the danger of a cross-language canonical becomes obvious. Some templates treat Chinese as the "master record" and leak its URL into the English canonical tag. When /en/services/website outputs <link rel="canonical" href="https://www.example.com/zh/services/website" />, you instruct Google to treat the English page as a duplicate of the Chinese URL and drop the English version from indexing—undermining your broader SEO and GEO content strategy in overseas markets.
Google's canonical documentation sets three strict rules for keeping both signals compatible:
- Always specify a same-language canonical: When using
hreflang, pointrel="canonical"to a URL in the same language (falling back to a substitute language only if no canonical exists in that language). - Never combine
hreflangandcanonicalon one tag: Google ignoresrel="canonical"elements carryinghreflang,lang, ormediaattributes. Keep canonical and alternate tags separate. - Healthy
hreflangclusters reinforce canonical selection: When choosing among same-language duplicates, Google favors URLs participating in reciprocalhreflangclusters and strongly prefershttps://URLs.
Bidirectional Confirmation, x-default, and Subdirectory vs. Subdomain Mapping
With canonical boundaries settled, four technical rules govern how hreflang behaves across single-domain paths and separate subdomains.
1. Self-Inclusion, Return Links, and Absolute HTTPS URLs
- List every valid variant plus the current page itself: Both Chinese and English pages must emit the exact same set of
hreflanglinks, including a self-referencing entry. - Missing return links break the pair: If Page X links to Page Y via
hreflang, Page Y must link back to Page X—otherwise Google ignores the tags (preventing external domains from hijacking another site's localized cluster). - Use fully-qualified HTTPS URLs: Always write
https://www.example.com/..., never relative/en/...or protocol-relative//example.com/...paths. - Choose one delivery channel: HTML
<head>, HTTPLink:headers (useful for PDFs), and XML Sitemaps (<xhtml:link>) carry equal search weight. Pick one method instead of maintaining duplicate sets that drift out of sync. - Use valid ISO language, script, and region codes: Start with an ISO 639-1 language code (
zh,en), optionally paired with an ISO 15924 script (zh-Hans,zh-Hant) or ISO 3166-1 Alpha 2 region (en-GB,en-US). Never pass a country code alone (bemeans Belarusian, not Belgium) or unsupported codes likeUK.
2. Where x-default Fits
The reserved hreflang="x-default" value specifies a fallback destination when a searcher's browser language matches none of your declared languages. Google designed x-default primarily for language selector pages or default landing homepages (for instance, routing unmatched locales to your global English or neutral entry homepage), rather than tagging unrelated subpages inconsistently.
3. Single-Domain Subdirectories (/zh/ + /en/) vs. Cross-Subdomain (www + en.)
As outlined in Google Search Central's guide to managing multi-regional and multilingual sites, subdirectories (example.com/zh/ + example.com/en/) simplify shared-host maintenance, whereas subdomains (www.example.com + en.example.com, common in a Multilingual & Global Standalone Site Solution) allow separate server locations and cleaner host isolation.
Crucially, hreflang works across different subdomains and separate domains. However, Google does not treat en. in a subdomain, /en/ in a path, or the HTML lang attribute as proof of page language—it detects language from visible body content and relies on explicit hreflang markup to connect cross-host pairs.
Here is how <head> markup looks across both structures using hypothetical domains:
<!-- Example A: Single-domain subdirectories (on https://www.example.com/zh/services/website) -->
<link rel="canonical" href="https://www.example.com/zh/services/website" />
<link rel="alternate" hreflang="zh-Hans" href="https://www.example.com/zh/services/website" />
<link rel="alternate" hreflang="en" href="https://www.example.com/en/services/website" />On https://www.example.com/en/services/website, only the self-referencing canonical switches to English while the hreflang list remains identical. In a cross-subdomain setup (www + en.), the English page outputs:
<!-- Example B: Cross-subdomain deployment (on https://en.example.com/services/website) -->
<link rel="canonical" href="https://en.example.com/services/website" />
<link rel="alternate" hreflang="zh-Hans" href="https://www.example.com/services/website" />
<link rel="alternate" hreflang="en" href="https://en.example.com/services/website" />Front-end language switcher UX and multi-site editorial workflows live in separate architectural layers from these search-engine discovery tags.
Handling "Chinese Live, English Not Ready Yet" Without Breaking Indexing
In live operations, teams regularly publish a Chinese update before its English translation is reviewed—or keep domestic notices Chinese-only. Rushing incomplete pairs causes two failures: pointing hreflang or sitemaps at a 404/redirecting placeholder breaks bidirectional confirmation, while serving untranslated Chinese copy under /en/... violates Google's warning against mixed boilerplate/body pages that create duplicate content.
Google explicitly permits omitting unready languages on individual entries while preserving hreflang where translations exist:
- Keep the live Chinese page self-canonicalized: Output
<link rel="canonical" href="https://www.example.com/zh/..." />. - Omit unready English URLs from
hreflang: Do not emithreflang="en"on the Chinese page; if no other language is published yet, omit cross-localehreflangentirely. - Return true 404 and exclude from English sitemaps: Keep unready English URLs out of
sitemap.xmland never fall back silently to Chinese copy. - Enable symmetric
hreflangat English release: Only after the English page returns HTTP 200 with verified English copy and its own self-canonical, add it to the English sitemap and publish reciprocalhreflangtags on both pages together.
Hypothetical Example: Bilingual Canonical & Hreflang Mapping Reference Table
The matrix below is an explicit hypothetical example using fictional hosts (www.example.com and en.example.com) to illustrate expected outputs across six page states:
Hypothetical Page Scenario | Public Status | Required | Required | Sitemap Rule |
|---|---|---|---|---|
Homepage (Subdirectories) | ZH & EN live (200) | ZH points to | Reciprocal | Include both canonicals |
Service Page (Subdirectories) | ZH & EN live (200) | ZH points to | Self-inclusive, symmetric | Include both canonicals |
Solution Page (Cross-Subdomain) | | ZH points to | Full cross-host | Each host sitemap lists own canonicals (or aligned |
Domestic News (Gap State) | ZH live (200); EN unready | ZH points to | Omit | ZH sitemap only; EN returns 404 & stays out of EN sitemap |
Campaign Query URL | EN with | Clean EN canonical | Clean parameter-free | List clean canonicals only |
On-Site Search Page | | Clean base URL or policy canonical | Omit | Exclude from sitemaps |
Four Pre-Launch Verification Steps
Before opening indexing on a bilingual site, check actual rendered responses across four verification steps:
- Inspect
rel="canonical"on sample English pages: Check View Source (and rendered DOM if client JavaScript runs) on the homepage, a service page, and an article to confirm one singlerel="canonical", written as an absolutehttps://URL pointing to that English page itself—never back to Chinese. - Test
hreflangpairs end-to-end: Request the exact URL insidehreflang="en"from a Chinese page and verify HTTP200status (not 301 or 404), genuine English body content, and a matching returnhreflangpointing back to the canonical Chinese URL alongside its self-link. - Confirm one single source of truth against XML Sitemaps: Ensure HTML, HTTP headers, and XML sitemaps do not emit conflicting
hreflangsets, and verify every sitemap URL returns HTTP200with zeronoindexor untranslated pages. - Spot-check a Chinese-only page and HTTPS host redirects: Confirm untranslated entries emit zero
hreflang="en"links and never serve Chinese body copy at/en/..., whilehttp://variants redirect cleanly tohttps://canonical hosts.
Clear separation between canonical and hreflang removes guesswork for search crawlers: every language stands cleanly on its own URL, verified pairs confirm one another symmetrically, and unfinished translations stay honestly offline until ready. If you are planning a bilingual website or international standalone site, explore our Corporate Website & Standalone Site Services or bring your URL map to our consultation page to verify launch readiness.