A multilingual manufacturing website is not a translated brochure

The short answer: a manufacturing export website is not a Chinese site with an English translation bolted on. It needs a separate URL for each language, local units, certifications, cases, contact routes, payment and privacy terms, and a lead-handling process that works in the buyer’s time zone. Translation alone rarely earns search traffic, and it rarely earns trust.

This guide is for manufacturers planning an export site, an overseas brand site, or a multilingual site. Research cut-off: 25 October 2026. It references China Customs data for the first three quarters of 2026 and Google’s official documentation on multilingual sites and AI features. Budget ranges are editorial estimates, not quotes, and we do not promise enquiries, rankings, or launch dates.

Translation, multilingual, and localization are three different things

China’s goods trade reached RMB 39.94 trillion in the first three quarters of 2026, up 18.8% year on year, with exports of RMB 23.2 trillion, up 16.4%, according to China Customs. Demand is expanding. But an overseas buyer will not trust a supplier because the site has an English page. They want to see certifications, delivery evidence, references, after-sales terms, and a way to talk to a human in their time zone.

Subdirectory, subdomain, or country-code domain?

There is no universal best choice. Pick based on the number of markets, team structure, and long-term investment.

Google recommends different URLs for different language versions and hreflang annotations between them. It does not require a specific domain structure. The important part is that the URL is stable, accessible, and has a clear language version.

hreflang: reciprocal, correct, and not a ranking trick

hreflang tells search engines which language or regional version to show. It is not a ranking bonus, and it cannot compensate for thin content or a poor local experience. Most failures come from four mistakes:

• No return links. Every version must reference itself and the other versions, including x-default. A one-way tag from English to Chinese is not enough.
• Wrong codes. Use ISO 639-1 for language and ISO 3166-1 Alpha 2 for region, such as en-GB. Do not use a region code without a language.
• Forced redirects. Do not redirect by IP or browser language. Let users switch languages; otherwise crawlers may only see the default version.
• Content mismatch. The page language, navigation, form, and structured data must all agree.

What localization actually changes

Localization is not a vocabulary exercise. It changes the evidence a buyer needs:

• Units and standards: metric and imperial where relevant; local certifications and test reports.
• Industry terminology: the words buyers use, checked by someone who knows the trade.
• Local proof: nearby customers, distributors, service partners, delivery times, and after-sales commitments. Do not invent a local office.
• Buying path: local phone, email, form, privacy consent, time zone, CRM routing, and response SLA.
• Commercial terms: payment, tax, shipping, returns, and privacy statements appropriate to the market.

Start with one or two priority markets, not ten languages. Build the core pages first: home, products or capabilities, industry solutions, certifications, cases, about, and contact. Depth beats breadth. Every extra language needs content, review, support, and maintenance, or it becomes a thin empty page.

What a three-language, 30-page launch costs

The base site is only one line. The incremental budget for multilingual work is mostly translation, local content, hreflang development, and lead handling. The table below is an editorial estimate for three languages and about 30 core pages, not a quote.

At roughly RMB 7.2 to US$1, that is about US$5,000–50,000 before media spend. The range is wide because the base site and the depth of local proof vary enormously. Translation alone is usually the worst value: it adds pages without adding trust or a reason to choose you.

AI search does not need a separate website

Google’s AI features still depend on core search quality: crawlability, indexability, useful content, and a page that can be shown as a snippet. A multilingual site does not need “another set of content for AI.” It needs independent URLs, clear language signals, citable evidence, and local trust.

What helps: dates, sources, named certifications, specific numbers, case evidence, and pages that answer a question directly. What does not help: forced IP redirects, hidden body copy, machine-translated pages with no review, and batches of near-duplicate pages.

How to accept the launch

• Technical: every language URL returns 200; canonical is self-referencing; hreflang is reciprocal; sitemap is submitted per language; mobile is readable.
• Search: review impressions, clicks, queries, and landing pages by language and country. Do not judge a market by global totals.
• Enquiries: track source language and country, form quality, response time, CRM assignment, and deal cycle — not just form volume.
• Content: check local cases, certifications, product updates, stale pages, broken links, and translation quality on a schedule.

FAQ

Can we just machine-translate the Chinese site?

It can serve as a temporary display, but it rarely supports lead generation. Machine translation often misses terminology, units, compliance language, and local phrasing, and it produces low-differentiation pages. An export site needs independent URLs, local keywords, cases, certifications, contact routes, and at least native review.

Which URL structure should we use?

There is no universal answer. A subdirectory is simplest and concentrates authority. A subdomain helps separate markets or teams. A ccTLD has the strongest local signal but costs more to register, operate, and populate. Choose based on markets, team capacity, and long-term investment, then keep URLs stable.

Does hreflang improve rankings?

No. It helps search engines show the right language or regional version. Missing return links, incorrect codes, and one-way tags can make it ineffective.

How many languages do we need?

Start with the markets that already produce enquiries or where you have distribution. Usually English first, then German, Spanish, French, or Japanese as demand justifies. Each language needs content, review, support, and maintenance.

How long until enquiries arrive?

There is no honest universal answer. It depends on the market, competition, content quality, local proof, paid support, and sales follow-up. Judge direction with indexing, language-level queries, enquiry quality, and deal cycle — not a promise of “results in X months.”

Sources and caveats

Research cut-off: 25 October 2026. China Customs data covers national goods trade in the first three quarters of 2026 and does not represent any single industry or company. Google’s documentation describes multilingual and AI-search behavior; crawling and display can change. Translation rates, page counts, base-site costs, and localization budgets are assumptions, not market prices. We do not promise enquiries, rankings, indexing, or AI citations.

1. General Administration of Customs of China, “2026年前三季度进出口情况新闻发布会,” 14 October 2026. customs.gov.cn.
2. Google Search Central, “Managing multi-regional and multilingual sites,” updated 10 December 2025. developers.google.com.
3. Google Search Central, “Localized versions of your pages,” updated 21 September 2026. developers.google.com.
4. Google Search Central, “AI features and your website,” updated 10 December 2025. developers.google.com.