Shopify Markets: why a good translation can still contain the wrong facts

Translation and market adaptation are two different jobs

A translation carries content into another language. A market adaptation decides which content applies to a specific sales market. Not only terms can differ, but also technical data and business statements.

A literal translation of the German text in the example would carry over 230 V and CE. For the assigned US version those facts would be wrong. The correct US description therefore needs more than a linguistic translation: it needs the binding data of the actual US version.

Note: 230 V to 120 V is not a unit conversion. Translation software must not turn it into a different product version on its own or infer a certification. Whether that version exists and which facts apply has to come from reliable product data.

What happens when the general product description changes?

Say the merchant adds a paragraph on how to operate the machine. The US description should get that addition but keep stating the US voltage and the facts documented for that version. A full retranslation from the general German text would not be enough.

Shopify marks a translation as outdated when the original content has changed since the last update. That status does not tell you which regional facts are still right or which parts of the text should be updated.

So the rule for maintenance is: a changed source means reassess, not replace everything automatically. First decide which changes can be taken over linguistically and which market information remains binding.

It works the other way round too. If only the US free shipping threshold changes in the data source, the US description may need an update although the German text is unchanged. Reliable maintenance follows the dependencies of the target text, not just its linguistic source.

Which market information should be protected?

Worth protecting is anything that does not follow from the target language alone. An English text, for example, does not say whether it is meant for the United States, the United Kingdom or another market.

1. Technical equipment: voltage, frequency, plug, model variants and documented certifications.
2. Shipping and delivery: delivery areas, delivery times, thresholds, customs and import notes.
3. Service and contract information: warranty statements, return terms and regional service addresses.
4. Assortment: available variants, pack sizes, ingredients and product specific notes.
5. Offers and payment: regional promotions, payment methods and delivery restrictions.

A checklist, not a mandate for automation

The list is a checklist, not a mandate for automatic adaptation. Technical and contractual statements in particular need confirmed data. An AI model should not fill in missing market information from the language or from typical national habits.

How do Shopify translation apps support this workflow?

When choosing an app, "compatible with Shopify Markets" is not enough. What matters is whether it supports language assignment, separate market texts and their later updates. The following is based on publicly available vendor documentation, checked on 4 October 2026. We did not run a hands-on test of the apps for this article.

Shopify Translate & Adapt: Shopify's own app allows adapted content for a selected market and optionally a selected language. The documentation names regional terms, promotions and market specific metafield values among its use cases. When automatic translation runs again, empty and outdated automatically translated content is processed. According to Shopify's documentation, manually added or edited translations stay unchanged. Protecting existing content is therefore already an important benchmark, not a fundamentally new idea.

T Lab: T Lab documents AI translation, a glossary, your own API keys and an autopilot for new or updated content. These features help with translating and continuously updating general language versions. According to its own support documentation, T Lab does not support market specific translations. The vendor recommends combining it with Shopify Translate & Adapt: standard translations in T Lab, market adaptations in Translate & Adapt.

Langify: By default, Langify leaves existing translations unchanged in individual machine translation. Retranslating existing versions has to be selected explicitly, and fields can be excluded from bulk translation. Bulk translation fills in missing content by default. For versions that became outdated after source changes there is an additional "Translate outdated" option. That is a different way of working than continuous updating without user action.

Weglot: Weglot documents Shopify Markets compatibility, a glossary, exclusions and a visual translation editor. The integration uses its own translation and delivery path, so setting it up for Shopify Markets is not the same as a purely native Shopify translation. When the original text changes, Weglot treats the content as new and creates a new translation. A previous manual translation remains in the translation list, but according to the support documentation the new automatic version is displayed. If you want to keep using the old manual version, you can copy it into the new entry. For market texts this matters: a manual version can be stored without being shown.

LangShop: LangShop supports different translations of the same language per market, for example British English for the United Kingdom and American English for the United States. The feature is documented in the changelog and explained in a separate use case. LangShop also documents exclusion rules and detection of outdated translations. Market variants, exclusions or change indicators alone are therefore not unique selling points for a new translation app.

Four decisions for maintaining market texts

For a traceable workflow it pays to decide explicitly how future changes are handled. Each decision should first apply to one specific piece of content, one field, one language and one market. Extending it to whole content groups should be a deliberate step.

1. Exclude from translation: keep the existing text, no AI update.
2. Always submit for approval: propose changes, show the differences and wait for a decision.
3. Maintain manually: no AI translation, create a maintenance task when relevant changes occur.
4. Translate by explicit instruction: generate new text while respecting explicitly assigned market values and text exclusions.

An instruction has to be specific

A specific instruction could read: "Take the voltage from the US metafield of the matching variant. Do not carry over the European warranty paragraph." The instruction "keep market information" would be too vague if it is not clear which facts are meant and where they come from.

If the data source is missing or contradictory, the new text should not be published automatically. The existing content stays, but the conflict has to be visible. Protecting existing content and checking it professionally are two different states.

The special cases that are often overlooked

A good workflow does not only protect existing texts from being overwritten. It also notices when there is no protected market version yet. According to Shopify, general translations are visible across markets as long as no corresponding market translation exists. Simply skipping can therefore leave the wrong general version in place.

Deliberate omissions matter just as much. If a merchant removed a paragraph for one market, a later retranslation should not put it back without asking. That decision belongs in the instruction too.

Finally, something can change after approval: the target text, the product data or the rule itself. Publishing should therefore only happen if the approved proposal still fits the current data. Otherwise a new review makes more sense than silently overwriting.

What we are working on at Glotari

Glotari translates with your own AI key and has every translation checked by a second model that returns findings and a Translation Score. A different model is no guarantee of correctness, and our product page says so.

For market texts we are working on saved market rules. Existing market adaptations are meant to stay protected by default, joined by a market overview, saved decisions and partial translations based on clearly assigned market values. This is not available yet. The difference is not meant to be that different texts per market become possible for the first time, but in how later changes are handled traceably: what may be updated, which data is binding and when a new approval is needed?

Checklist before your next automatic translation

Use these questions to check the workflow for your shop. Test it first on a single product with an existing market adaptation before you extend automation to larger content groups.

1. Market: Is it clear which market and which product version the text applies to?
2. Binding data: Are voltage, shipping values and other market facts assigned to a clear source?
3. Updates: What happens when the general description changes?
4. Data changes: What happens when only a market value that is used changes?
5. Protection: Is a manual version not only stored but also the version that is actually shown?
6. Missing version: Is it detected when no market version exists yet?
7. Approval: Are changes made in the meantime detected before writing?

Can Shopify show different English texts for different markets?

Yes. Shopify supports market specific translations of the same language as well as general translations without a market. Separate English versions can be stored for different markets.

Does a glossary automatically protect technical market facts?

A glossary defines how words or terms are handled. For a fact such as 120 V the workflow additionally needs a binding assignment to the matching product version and market. A fixed term alone does not replace that assignment.

May AI adapt voltage or certifications to a country?

Not by guessing. Such facts should only be taken from confirmed product data for the specific version. Translating a text changes neither the technical equipment nor the documented certification of a product.