SEO for multilingual websites: how to do it? Does hreflang work on Baidu?
Chapter 2: Search Fundamentals · Question 32
For multilingual websites, many people's first instinct is "just add an hreflang tag." That is indeed the standard practice within the Google ecosystem, but if your main battleground is Baidu, the situation is completely different.
What is hreflang? How does Google use it?
hreflang is an HTML tag defined by Google, used to tell search engines: “this page has versions in other languages/regions”. The syntax is as follows:
<link rel="alternate" hreflang="zh-CN" href="https://www.example.com/zh/" />
<link rel="alternate" hreflang="en" href="https://www.example.com/en/" />
<link rel="alternate" hreflang="ja" href="https://www.example.jp/" />
Its core function is to have search engines display the corresponding version in search results for different languages/regions and avoid duplicate content determinations.
Baidu's stance on hreflang
Baidu has never officially stated that it supports the hreflang tag. This is a pitfall many teams expanding overseas easily fall into—applying Google's rules directly to Baidu.
Baidu identifies multilingual content in a more "blunt" way:
- Judging by the actual language of the text on the page
- Judging by the regional attributes of the domain/subdomain
- Judging by the "internationalization configuration" settings in Baidu Search Resource Platform
The right approach on Baidu Search Resource Platform
In Baidu Search Resource Platform → Search Appearance → Internationalization Configuration, you can:
- Set the site's language attribute (Chinese/English/Japanese, etc.)
- Set the site's regional attribute (Chinese mainland/Hong Kong, China/Other)
This is more effective for Baidu than the hreflang tag.
Architecture choices for multilingual websites
| Solution | Example | Advantages | Disadvantages |
|---|---|---|---|
| Subdomain | cn.example.com / en.example.com | Independently managed, flexible regional targeting | Authority is diluted; backlinks must be built separately |
| Subdirectory | example.com/zh/ / example.com/en/ | Shares the main domain's authority, simple to maintain | Regional targeting is not precise enough |
| Country-code top-level domain (ccTLD) | example.cn / example.jp | Strongest regional signal, high user trust | High cost, authority is completely independent |
Recommendations for the China market: If your core users are in the Chinese mainland, prioritize a .cn domain or a .com domain paired with a Chinese subdirectory. For the Japan and Southeast Asia markets, you can use the subdomain approach.
Content localization vs. pure translation
This is the most common mistake in multilingual SEO:Just machine-translate the Chinese content directly into English/Japanese and you're done.
True localization requires:
- KeywordRe-research：中文用户搜"手机壳"，日文用户搜"スマホケース"，英文用户搜"phone case"——search intentcompletely different from the competitive landscape
- Content structure adjustment：Japanese users prefer detailed specification tables, while European and American users prefer scenario-based descriptions
- Local case replacement: The Chinese version uses Taobao as an example, so the English version should switch to Amazon
- Compliance check：Different regions have different requirements for advertising language and privacy policies
Cross-language canonical setup
Between multilingual pagesDo notSet each other canonicalEach language version is an independent canonical page:
<!-- 中文版页面 -->
<link rel="canonical" href="https://cn.example.com/product-a" />
<!-- 英文版页面 -->
<link rel="canonical" href="https://en.example.com/product-a" />
Only duplicate pages within the same language (such as URLs with parameters) need a canonical pointing to the canonical version in that same language.
Practical recommendations
- On Baidu's side：Do not rely on hreflang; instead, use the internationalization configuration in Baidu Search Resource Platform plus a separate subdomain/directory approach, so Baidu can naturally identify the language through page language and domain structure
- On Google's side : implement hreflang tags correctly, ensure bidirectional declarations (A points to B, and B must also point back to A), and in sitemap contains hreflang information
- Content strategy：Each language version needs its own independentKeyword researchand content localization; machine-translated content will not rank well on any search engine
Solve with AI tools: multilingual website content and hreflang configuration
Scenario 1: Generating bidirectional hreflang declarations
My website has three language versions: Chinese (https://example.com/zh/) and English (https://example.com/en/), Japanese (https://example.com/ja/)。Please help me generate the hreflang tag code for each language page: each version must point to all other versions (including itself), and use x-default to specify the default language. Also explain how to add the xhtml:link declaration in the sitemap.
Scenario 2: Keyword localization research (not machine translation)
I am a company offering [product/service], and my core Chinese keywords are [a few words]. I want to enter the Japanese market. Please help me: 1) provide the Japanese search terms corresponding to these words (not literal translations, but how Japanese people actually search); 2) provide 2-3 long-tail variants for each term; 3) flag which terms are popular in China but may have very low search volume in Japan. Please output as a table: Chinese term | Japanese search term | Variants | Notes.
Scenario 3: Content localization rewriting (preserve meaning, replace examples)
I have a piece of content aimed at Chinese users (paste the body text) and want to publish it on an English site. Please rewrite it for me: keep the core message and structure, but replace China-specific examples/platforms (e.g., replace Taobao with Amazon), adjust expression habits, and avoid Chinglish. Do not machine translate; write it like a native English speaker.
Scenario 4: Check whether the canonical of multilingual pages is set correctly
Here are HTML snippets from several language versions (paste the head sections of the Chinese and English versions). Please check: 1) whether each version's canonical points to itself; 2) whether there are cross-language errors such as canonicalizing the Chinese version to the English version; 3) whether hreflang is declared bidirectionally. Point out the problems and provide corrected code.
Note: the boundaries of AI
- AI localization ≠ local expert: AI understands language and common patterns, but local habits, sensitive words, and compliance requirements in Japan/Europe/America still need review by a local team or native speaker
- hreflang only serves Google: Baidu does not recognize hreflang, so do not treat AI-generated hreflang as a Baidu multilingual solution; on the Baidu side you still need to go through internationalization configuration
- Don't rely on AI to guess search volume: AI's judgment of search volume in specific regions is only an empirical estimate; actual keyword lists must be verified with tools such as Google Keyword Planner / 5118
