A product listing in the wrong language is a closed door. A shopper who cannot read your description, size guide or returns policy will leave, however good the product. Translating the catalogue is one of the most direct ways to reach new customers, and also one of the easiest ways to spend money on pages nobody sees or trusts.
This guide helps you decide where translation pays, choose between machine and human methods, handle the details that go beyond words, and run a workflow that keeps translations current when products change. It builds on product title formulas and the AI description review process. Platform features change, so consult each platform’s documentation.
Decide the market first
Translation without readiness is waste. Before translating anything, check that you can serve the market.
| Question | Why it matters |
|---|---|
| Is there demand? | Look at existing visitors, orders and search interest from that country or language |
| Can you ship there at an acceptable cost and time? | A listing that cannot be delivered is a disappointment |
| Can you handle duties and tax? | Surprise costs at the door drive returns and bad reviews |
| Can you support customers? | Questions will arrive in that language |
| Do you meet local rules? | Product safety, labelling, consumer rights and privacy |
| Can you accept local payment methods? | Many markets prefer particular methods |
Start with one or two markets where several answers are yes. A language and a country are not the same: Spanish may serve Spain, Mexico and the United States differently, with different shipping and legal issues for each.
Machine, human or both
| Method | Strengths | Weaknesses | Best for |
|---|---|---|---|
| Machine translation | Fast, cheap, consistent for plain text | Errors in terminology, tone and ambiguity | Long-tail catalogue, first drafts |
| AI with glossary and instructions | Can follow style, glossary and context | Can invent or smooth over errors; needs checks | Large catalogues with rules |
| Human translator | Natural, accurate, culturally aware | Slower, costlier | Top products, brand pages, legal text |
| Hybrid: machine plus human review | Balance of speed and quality | Needs a process | Most stores |
The hybrid route is practical for most stores. Machine translation or an AI model produces drafts using a glossary; a native speaker reviews the best sellers, the legal pages, and a random sample of the rest. Quality is measured by the review: if the sample shows frequent errors, tighten the glossary and instructions, or move more content to human translation.
Build a glossary
A glossary is the cheapest improvement you can make. List brand names that must not be translated, product terms, materials, sizing words and phrases you use for features. Give the approved translation for each language. Feed it to the translation tool or the AI prompt and use it in review. Without it, the same material can be translated three different ways across the catalogue, which looks careless and hurts search.
| Term | Rule | Example |
|---|---|---|
| Brand and collection names | Do not translate | Northline |
| Material terms | Use the standard local term | Linen, with the local word agreed once |
| Size words | Match the local convention | Local size labels, not a literal translation |
| Feature names | Translate consistently | Same word for “washable” everywhere |
| Do-not-say list | Words that mislead or are restricted | Medical claims or protected terms |
Localise, do not only translate
- Units. Centimetres or inches, kilograms or pounds, and the local way to write dimensions.
- Sizes. Clothing and shoe sizes differ by country. Provide a conversion chart.
- Currency and prices. Show local currency and indicate whether tax is included, as customers expect in many markets.
- Dates and formats. Day-month order, decimal commas, phone formats.
- Imagery. Images with text need translated versions; models, settings and colours should suit the market.
- Tone. Formal or informal address varies by language and category.
- Legal text. Returns, warranty, privacy and terms must reflect local consumer rules; use qualified help.
- Product names and examples. Idioms and references may not travel.
Search in other languages
People do not search in literal translations of your English keywords. A phrase that is common in one language may be rare in its translation, and the local word for a product may differ by country. For each market, research how shoppers search: use local keyword tools, look at competitors’ titles and ask native speakers. Then write titles and descriptions around those terms, as described in our guide to Etsy SEO, where the same principle applies to marketplace search.
Technical setup
Search engines need to know which version of a page serves which audience. Use separate URLs for each language or region, through subfolders, subdomains or country domains, and mark the relationships with hreflang annotations in the page, the sitemap or the headers. Each page should link to its alternates and itself, and canonical tags should point to the right version. Translate titles, meta descriptions, image alt text and structured data too, not only visible body text. Avoid automatically redirecting by IP address in ways that block search engines or users who want another language, and let people switch language easily. See Google’s documentation for details; mistakes here are common.
| Item | What to do |
|---|---|
| URLs | One URL per language or region version |
| hreflang | Annotate all versions, including a default |
| Canonicals | Point to the same-language page |
| Metadata | Translate titles, descriptions, alt text and structured data |
| Language switcher | Visible and remembering the choice |
| Sitemaps | Include all versions |
Marketplaces and feeds
If you sell through marketplaces, translation rules differ again. Some marketplaces let you add listings in each country’s language through a single account, others require separate storefronts, and many have their own category and attribute vocabularies. Google Shopping feeds need the language and country set correctly for each feed, with titles and descriptions in that language and prices in the local currency. Check the policy on machine-translated content for each channel, as some require accurate, human-quality text, and keep a record of which listings are translated and when they were last reviewed. The data discipline in Google Shopping feed errors applies to every language version.
Customer support and policies
A translated product page raises the expectation that someone can answer questions in that language. Translate the support pages that matter most: shipping, returns, sizing, contact and order tracking. Provide an email address or form where replies can be written in the customer’s language, using translation tools with review for the first months. Translate automated emails too, because an order confirmation in the wrong language undermines the whole experience. Our customer support chatbot can answer in several languages from your policy text, with handoff to a person where needed.
A workflow that scales
- Source of truth. Keep product data in one language and one structured format.
- Mark translatable fields. Titles, descriptions, attributes, alt text, metadata.
- Translate with the glossary, by machine or AI, in batches.
- Check automatically. Placeholders intact, numbers and units unchanged, no left-over source language, length limits respected.
- Human review. All top products and legal text, plus a random sample.
- Publish through the platform’s translation system or bulk upload.
- Detect changes. When the source changes, flag the translations as outdated and retranslate.
- Measure. Track traffic, conversion and returns by market.
Step seven is the one that gets forgotten. A translated catalogue that does not update when prices, sizes or materials change quickly becomes inaccurate. Our product data enrichment, bulk product upload and product description writing services include change detection and translation workflows.
A worked example
A store with 1,500 products wants a German version. A professional translation at about $0.10 per word for 150 words per product would cost around $22,500. A hybrid process, with AI translation using a glossary, automated checks and human review of the top 150 products and a 5 percent sample of the rest, costs perhaps $3,000 to $4,000 in review time and tooling. If the German site brings in 400 orders a month at a $60 average order and a $20 margin, that is $8,000 of monthly margin once established, so the project pays back within the first months, assuming the market is served well. Without readiness, shipping and support, the same translations could produce nothing. The figures are illustrative.
Mistakes to avoid
- Translating before the market is ready.
- No glossary, so terms vary.
- Ignoring size and unit differences.
- Translating only body text, and leaving metadata in the source language.
- Forgetting hreflang and language URLs.
- Auto-redirecting visitors by IP.
- Letting translations go stale.
Sources and further reading
Figures and rules in this guide come from the primary pages below. Platform limits and prices change, so open the page before you build on a number.
- Google Search Central: localized versions of your pages, hreflang and multi-regional guidance
- Shopify Help Center: Translate & Adapt, managing store languages
- W3C: language tags in HTML and XML, marking up language correctly