E-commerce Product Description Translation at Scale

Plan ecommerce translation for large catalogs: build terminology, review machine output, protect Shopify data, and publish product pages and feeds consistently.

Also in: UK RU EN
E-commerce Product Description Translation at Scale

A catalog can contain thousands of product pages, but one inconsistent material name or mistranslated size option can make the whole store feel unreliable. The hard part of ecommerce translation isn’t producing more words. It’s keeping product facts, terminology, storefront content, and feed data aligned as the catalog changes.

A useful plan treats translation as a repeatable content workflow, not a one-off export. Start with clean source data, decide what needs human review, and define how translations move into the store without replacing approved work.

Why catalog translation needs a system

Ecommerce product translation is the process of adapting product titles, descriptions, attributes, options, and related content for shoppers in another language. At scale, a translation workflow also needs to manage catalog updates, review, publishing, and the connections between storefront pages and product feeds.

The commercial case isn’t just about having more language versions online. Clear product information can help shoppers understand what they’re buying and whether it suits their needs. In a survey reported on July 7, 2020, CSA Research said 76% of 8,709 consumers across 29 countries preferred to buy products with information in their native language. The same survey found that 40% said they would never buy from websites in other languages. Those are historical survey findings, not a current market-size estimate, but they show why language can affect a purchase decision.

Translation also touches more than the main description. A shopper may see a title in search results, a material field on the product page, a size label in a variant selector, and shipping information at checkout. If those pieces contradict each other, polished prose won’t save the experience.

EU consumer guidance says traders must give clear, correct, understandable information about a product before purchase, including its main characteristics. For online purchases, the information also includes topics such as delivery restrictions, withdrawal rights, after-sales services, and dispute-resolution mechanisms, according to Your Europe. Product description translation should therefore sit inside a broader localization process, not replace legal and checkout-content review.

What belongs in the translation scope?

Treat the catalog as several connected content groups:

  • Product titles and descriptions.
  • Product type, material, dimensions, care instructions, and other attributes.
  • Variant names and option values, such as color or size.
  • SEO titles, meta descriptions, and URL slugs.
  • Collection names, filters, and navigation labels.
  • Product feed fields, including titles and descriptions sent to advertising platforms.
  • Related purchase information, such as delivery restrictions and after-sales details.

Not every field needs the same kind of translation. A product description may need natural, persuasive language. A material attribute needs a precise equivalent. A SKU or model code should usually remain unchanged. A feed field may also have a platform-specific length or language requirement.

Create a field map before sending content to a translator or translation engine. For each field, record whether it’s translatable, whether it has a character limit, whether it contains product facts, and who approves changes. That map prevents a common handoff problem: a team requests “the product copy” and later discovers that the export omitted options, SEO fields, or attributes.

Choose a workflow before translating the catalog

Large catalogs need a workflow that separates source content, translation, review, and publication. A fast first draft doesn’t help if the team can’t tell which version is current or which rows are safe to import.

A practical workflow has four basic controls:

  1. A stable source. Decide which system owns the approved source text. Product managers shouldn’t edit one spreadsheet while translators work from another export.
  2. A content inventory. Identify the fields and product groups in scope, including variants and feed content.
  3. A review policy. Decide which products need full human review and which can use sampling or targeted checks.
  4. A publishing gate. Keep translated content out of the live store until the language, facts, formatting, and feed consistency checks pass.

Catalog size alone shouldn’t decide review effort. A low-risk accessory with a simple description may need a different process from a product with detailed safety information, care instructions, or technical specifications. Risk depends on what a mistake could change: the shopper’s understanding of a product, the selected variant, or the accuracy of an important claim.

Machine translation, human translation, or a mix?

Machine translation can generate a first draft quickly, but the draft still needs a defined review process. Human translators can interpret context, product voice, and ambiguous source text, but a human-only process can become difficult to coordinate across a large and frequently changing catalog. A mixed workflow uses automation for volume and reserves human attention for decisions and content where errors matter more.

Approach Useful when Main trade-off
Human translation for every field The catalog is small or the copy requires close creative adaptation More coordination and review effort as the catalog changes
Machine translation with human review The team has clear terminology, source data, and QA rules Output still needs fact, language, and context checks
Unreviewed machine translation Rarely appropriate for shopper-facing product copy Errors can reach product pages, options, and feeds unchecked
Internal bilingual team Staff know both the products and the target language Translation work competes with other responsibilities, and terminology can vary between people

A translation vendor, freelance translator, or internal linguist can each fit a different operating model. Compare them by how they handle updates, terminology, review, and handoff - not just by the price per word. The team still needs a content owner who can confirm that a translated product claim matches the source product facts.

Shopify’s automatic translation can add translations for untranslated content and update outdated machine translations, but it doesn’t change translations edited manually. Shopify also tells merchants to review automatic translations before publishing. Check the current Translate & Adapt documentation before setting up a workflow because availability and behavior can change.

Keep terminology and context attached to the work

A glossary is a list of approved terms and their preferred equivalents. For ecommerce, it should cover product names, materials, technical terms, brand language, and terms that must stay unchanged. A style guide sets the tone and explains choices such as formal or informal address, units, capitalization, and how to handle product claims.

Don’t send a translator a bare list of isolated words when context changes the meaning. A size option, for example, may need a different translation from a verb or an instruction. A Shopify Community user described an automatic translation problem involving the Italian word “misura,” intended as “size,” which became German “messen” and English “measurment” in the user’s workflow. The forum post from May 20, 2024 is an individual report, not evidence of a platform-wide error rate. The useful lesson is narrower: short option labels are easy to mistranslate when the system doesn’t have enough context.

The workflow in ChatsControl can help when the team needs to review a document-based batch rather than paste text into a box. You can set a translation brief with a domain, audience, tone, context, and preferred terminology; the translation process uses that guidance throughout. A bilingual view shows source and translation together, and the built-in QA checks issues such as numbers, names, terms, and omissions. The tool translates documents, not a full catalog inside Shopify, so the team still needs to prepare its export and manage storefront import separately.

Prepare Shopify content for bulk translation

Shopify’s CSV translation workflow gives teams a structured way to export and import selected language content. The export includes fields such as resource type, resource ID, field, locale, market, status, source or default content, and translated content, according to Shopify’s CSV documentation. Resource IDs and field names help connect a translation row to the correct store content.

A careful CSV process looks like this:

  1. Export only the languages and content you intend to work on.
  2. Keep resource IDs, field names, locale, market, and status columns intact.
  3. Filter for untranslated or outdated content when refreshing a catalog.
  4. Translate and review the intended fields without changing unrelated rows.
  5. Check delimiters, encoding, and column alignment in the finished file.
  6. Import the file and read Shopify’s import results before publishing.

Shopify’s current help documentation says exports can be filtered for content marked outdated or untranslated, which supports targeted refreshes instead of reprocessing every translated field. The same documentation warns that an import can overwrite existing translations if the operator selects the overwrite option. Review the import settings and keep a copy of the export so the team can identify what changed.

What can go wrong in the CSV?

A CSV file is plain text arranged in rows and columns. Spreadsheet software can silently change how the file is separated, displayed, or saved. Shopify says spreadsheet software should use commas rather than semicolons as the CSV delimiter; semicolons in HTML codes can otherwise cause columns to render incorrectly. Check the current instructions in Shopify’s CSV guide before importing because platform steps can change.

Shopify also says a translation CSV import stops if the file contains errors and lists those errors on the import page. The confirmation email lists translated content that failed or was skipped. Treat that report as part of the workflow, not an optional notification: an apparently successful upload doesn’t mean every row made it into the store.

Some content needs a separate plan. Shopify’s CSV documentation says product tags can’t be translated, and changing a product handle doesn’t support redirects for language-specific URLs. A translated description alone therefore doesn’t solve every localization requirement. Check tags and URL behavior separately before release.

Shopify’s developer documentation describes a GraphQL Admin API flow for translated content: retrieve the resource content and digest with translatableResources, then write a translation with translationsRegister. The digest from the source query must be included when registering the translation, according to the Shopify developer documentation. API teams should check the current documentation before building against it because API details can change.

Review the content shoppers actually see

Translation quality assurance (QA) means checking the result against a defined set of requirements, rather than relying on a general impression that the copy reads well. A reviewer should compare the translation with the source and the product record, then inspect the page where the translation will appear.

Use a review checklist that covers:

Check What to compare
Product facts Materials, dimensions, included parts, compatibility, and care instructions
Terminology Approved product terms, category names, and brand language
Variants Option labels and values shown in selectors
Numbers and units Sizes, quantities, model codes, and any measurements
Claims Benefits and performance statements against the approved source
Readability Natural phrasing and a tone suited to the target audience
Page display Truncation, broken formatting, misplaced text, or awkward line breaks
Cross-channel consistency Product page, feed, ads, and other product content

Review doesn’t mean every product needs the same number of passes. Start by checking a representative set from each content type and category. Add closer review where source text is ambiguous, the product has technical details, or a translation could change the meaning of a variant. Then use recurring QA checks to identify patterns, such as a material term that keeps changing or a number that disappears from translated copy.

A built-in quality check can help catch mechanical differences, but a human still needs to judge meaning and product context. A number check can flag a changed measurement; it can’t decide whether a translated phrase makes a claim the product team never approved. Keep product specifications available to reviewers so they don’t have to guess from the source description alone.

As Shopify’s Help Center puts it: “Before you publish translations to your storefront, review them for accuracy.”

The instruction is direct: automatic output is a draft, not a publication decision. Build the review into the release process so a translation can’t go live just because it exists in the admin.

Who should approve translated product claims?

The translator checks language; the product owner checks whether the copy still describes the product accurately. Assign both responsibilities instead of expecting one reviewer to catch every type of issue. A linguist may not know whether a material, compatibility statement, or included accessory matches the actual item.

For online purchases in the EU, Your Europe says information supplied before purchase should be clear and understandable, and includes the product’s main characteristics. The guidance also covers delivery restrictions, withdrawal rights, after-sales services, and dispute-resolution mechanisms. Read the official consumer guidance alongside local requirements for the markets where you sell.

A helpful approval record names the content owner, language reviewer, source version, and release date. When a product changes, the team can then tell whether the translation is still based on current information. Without that link, an old translation may look complete while describing a discontinued material or outdated product feature.

Scan: Marriage certificate, 1957Translation into English: Marriage certificate, 1957
A real example: Marriage certificate, 1957, Israel, Ministry of Religious Affairs, Hebrew. Scan → translation into English. Exactly what the product returned. Nobody fixed these files by hand. More examples →

Localize product feeds and multilingual SEO

A localized product page and its product feed need to describe the same item. Google Merchant Center says product title and description attributes are required, with maximum lengths of 150 characters for the title and 5,000 characters for the description in its current specification. Check the Google Merchant Center product data specification before implementation because platform requirements can change.

Google also says free-form fields such as title and description should use the same language throughout a data source. Supported enumerated values, including condition values such as new, refurbished, and used, must be submitted in English, according to that specification. Keep those rules in feed QA rather than assuming the storefront translation process will handle them automatically.

Feed and website mismatches can create practical problems. Google lists conflicting product data between a feed and a website among common Merchant Center issues; inaccurate, missing, or conflicting information can lead to disapprovals, limited eligibility, incorrect display, or other issues. Compare feed values with the live localized page before submission. A feed title that says one thing while the page says another can confuse both the platform and the shopper.

How should translated product pages handle hreflang?

Hreflang is an annotation that tells search engines which page versions correspond to different languages or regions. Google recommends distinct URLs for language or regional versions and supports three equivalent ways to identify them: HTML, HTTP headers, or sitemaps. Google says there’s no Search benefit to implementing all three at once; see Google Search Central’s localized-versions guidance.

Each language version should identify itself and all other versions. Alternate URLs need to be fully qualified, and reciprocal linking is required for Google to use the annotations. If maintaining every language link on every page is difficult, Google says some languages can be omitted on some pages while keeping reciprocal links for the ones included.

Markup doesn’t replace translation. Google says hreflang and the HTML lang attribute don’t determine a page’s language; Google uses algorithms to determine it. Publish genuinely localized visible content, then use the annotations to clarify the relationship between versions.

Shopify’s developer documentation says stores can enable and publish up to 20 locales, and newly enabled locales are unpublished by default. Those platform limits and behaviors are version-sensitive, so check the current Shopify documentation before planning a launch. A locale that’s enabled but not published won’t do the same job as a language version shoppers can access.

For more than product copy, see how to localize a store for Germany. Teams working on product listings for specific marketplaces can also use the Amazon product listing translation guide.

Build a repeatable translation and publishing process

A repeatable process makes the next catalog update less risky than the first one. It also gives the team a way to distinguish newly translated content from content that needs review again.

Use this operating sequence:

  1. Define the source of truth. Pick the approved product record and establish who can change the source copy.
  2. Map content fields. Include descriptions, options, attributes, SEO fields, and feed data.
  3. Set terminology and style. Give translators approved terms, product context, and rules for tone and formatting.
  4. Choose the translation method. Match human, machine-assisted, or internal work to the content’s complexity and risk.
  5. Review and correct. Check product facts, terminology, variants, numbers, readability, and page display.
  6. Import or publish in a controlled batch. Protect existing approved translations and inspect platform errors.
  7. Validate the live experience. Compare localized pages with their feeds and confirm language-version links.
  8. Track updates. Mark changed source fields so the team can refresh affected translations instead of treating the entire catalog as new.

Shopify’s CSV export filters for outdated and untranslated content can support a targeted refresh. Shopify’s automatic translation behaves differently for manual translations: it can update outdated machine translations, but it doesn’t change translations edited manually. Keep a record of whether each field was machine-translated, manually edited, or reviewed so future updates don’t silently overwrite a human decision.

What should go into the handoff checklist?

Use this checklist before sending a catalog batch for translation:

  • Source file or export is approved and dated.
  • Product IDs and field names are included.
  • Languages and markets are identified.
  • Required product context and terminology are attached.
  • Non-translatable codes and values are marked.
  • Options and attributes are included, not just descriptions.
  • Target field and feed limits have been checked.
  • Review owners are named.
  • Import settings protect approved translations.
  • The team knows how to handle errors and skipped rows.
  • Live pages and feeds will be checked after release.

Teams with large catalogs may also need a translation management system (TMS), which coordinates assignments, versions, and translation assets across projects. A document translation tool can assist with document batches and review, but it isn’t a full TMS or a catalog integration. Compare tools based on the workflow your team needs, including whether it must manage translation memory, vendor assignments, and direct store publishing. For a broader look at operational growth, read how to scale a translation agency.

Common pitfalls when scaling product translation

The most expensive mistakes often happen between translation and publication. A translation may read well in a spreadsheet but fail as a product option, exceed a feed field limit, or overwrite an existing approved version.

Importing the whole catalog when only a subset changed. Reprocessing every field creates more review work and more chances to overwrite a good translation. Use available filters for outdated or untranslated content and confirm the import behavior before uploading.

Treating each field as interchangeable copy. Product descriptions, attributes, options, and feed values have different roles. A fluent sentence doesn’t guarantee that a size selector or enumerated feed value is correct.

Sending isolated labels without context. Short strings can be ambiguous. Shopify’s CSV documentation also warns that product tags can’t be translated through the workflow, while URL handle changes don’t support redirects for language-specific URLs. Keep those edge cases on a separate checklist.

Assuming a successful import means every row worked. Shopify says an import can stop on errors and its confirmation email identifies content that failed or was skipped. Read the error report and check affected resources rather than trusting the upload alone.

Publishing mismatched product data. Google identifies conflicting information between a feed and website as a common Merchant Center problem. Review the live page and feed together before release, and investigate differences in titles, descriptions, and product facts.

Relying on markup to make an untranslated page understandable. Google uses algorithms to determine page language, and hreflang doesn’t determine it. The visible content needs to be genuinely localized, with annotations used to connect the different versions.

In a Shopify Community post dated May 20, 2024, user kub wrote: “Exporting translations was the first aproach and would be fine but it seems that the option title field doesn’t get exported.”

The post is an individual user’s report, not a general statement about every store or export. It points to a useful check for any workflow: compare the fields in the export with the fields shoppers see, especially option titles and values.

On November 13, 2026, Réka Patik, an English-to-Hungarian translator, tried ChatsControl and wrote that it “created a really good Hungarian translation! Better than most machine or AI output I post-edit”. Her review reflects her own assessment of that translation, not a guarantee that every catalog batch will need the same amount of editing.

On November 9, 2026, a sworn court translator-interpreter from Luxembourg who works in German, English, French, Serbian, and Macedonian told me he had tried CAT tools and wasn’t satisfied because he translates himself and “the most time takes rebuilding the layouts”. After trying ChatsControl on his documents, he wrote “I’m amazed”. His comment speaks to document layout work; a store team still needs to manage catalog exports, product data, and publishing in its own platform.

FAQ

How do you translate thousands of ecommerce product descriptions consistently?

Create an approved glossary and style guide, give translators product context, and keep source content in a stable system. Translate by field and category, then review representative and high-risk content before importing the full batch.

Can I bulk translate Shopify product listings without overwriting existing translations?

Yes. Shopify’s CSV workflow supports structured exports and targeted work on outdated or untranslated content. Check the import settings before uploading because selecting the overwrite option can replace existing translations; see the CSV workflow documentation.

What should ecommerce teams review after machine-translating product descriptions?

Check product facts, measurements, option labels, terminology, tone, feed limits, and consistency with the live product page. Shopify says merchants should review automatic translations for accuracy before publishing.

How should translated product pages be set up for multilingual SEO?

Give each language or regional version a distinct URL and use reciprocal hreflang annotations through HTML, HTTP headers, or a sitemap. Google’s guidance says each version should identify itself and the other versions with fully qualified URLs.

What language must Google Merchant Center product titles and descriptions use?

Google says free-form fields such as titles and descriptions should use the same language throughout a data source. Supported enumerated values, including condition values, must be submitted in English under the current product data specification.

Does machine translation make a product page ready to publish?

No. Machine translation is a draft that needs review for accuracy, product facts, and context. Google also warns that inaccurate, missing, or conflicting product data can cause Merchant Center issues.

Can Shopify automatic translation update translations that our team edited manually?

Shopify says automatic translation doesn’t change translations edited manually, even if they become outdated. Keep a record of manual edits and review the current automatic translation documentation before choosing an update process.

Is translating product descriptions enough for EU online sales?

No. Product descriptions are only part of the information a shopper may need before buying. Your Europe guidance also covers product characteristics, delivery restrictions, withdrawal rights, after-sales services, and dispute-resolution mechanisms; check the official consumer information guidance for the relevant market.

Try ChatsControl

AI platform for professional translators

Try for free →