Key takeaways
- Let your data pick the languages. Look at orders by country and the languages customers already write in, then support the few that matter properly before adding more.
- Not every language needs a native speaker. A tiered model works for most stores, with native agents for the biggest markets and an AI agent plus translated replies for the rest.
- Translate the help content customers need most (shipping, returns, payment), and adapt it to local rules and carriers instead of translating word for word.
- Have a native speaker check your saved replies and AI answers in each language once, then spot-check monthly. Machine translation is good, but tone and policy terms still go wrong.
- Tidio is our top pick for stores selling across borders because Lyro answers natively in more than 45 languages and agents can translate customer messages in one click, all in the same inbox.
You switched on international shipping in spring, and by August it's showing up in the inbox. A German customer asks where the return label is. A Spanish one wants to know whether the size runs small. Someone in the Netherlands writes in Dutch, and nobody on the team speaks Dutch.
Selling across borders has never been easier: your store platform handles currencies, payments and taxes. Support is what usually lags behind. Customers who can't get help in their own language buy less, return more and leave worse reviews.
The good news is that multilingual support no longer means hiring an agent per language. With the right mix of AI, translation and a little local knowledge, a small team can answer well in five or ten languages. This guide shows how.
The problem: your store speaks more languages than your team
The signs are familiar:
- Slow replies to foreign customers, because each message goes through a copy-paste translation.
- Wrong or awkward answers, especially on returns, refunds and legal terms.
- Help pages in one language, so foreign customers can't help themselves and write to you instead.
- Local expectations missed: different carriers, delivery times, payment methods and consumer rules.
- Time zones, so customers on the other side of the world always wait a night for an answer.
Fixing this is mostly about deciding where to invest and where good-enough automation will do.
Step 1: Let your data choose the languages
Before translating anything, look at two numbers for the last six to twelve months:
- Orders and revenue by country. Where does foreign revenue come from, and which markets are growing?
- Languages customers write in. Tag or sample conversations by language. Many customers write in English even when it isn't their first language, so this list is often shorter than the order list.
Combine them. A market with a lot of orders and many messages in its own language is a priority. A market with few orders and English-writing customers can wait.
Step 2: Choose a support model per language
Treating every language the same is expensive. Most stores do better with tiers:
| Tier | Which markets | Who answers | Help content |
|---|---|---|---|
| 1. Native | Your home market and your biggest foreign market | Fluent agents, with AI for repeat questions | Fully localised |
| 2. AI-first | Growing markets with steady volume | AI agent in the customer's language; agents reply with translation | Key pages translated and adapted |
| 3. Translated | Occasional orders | Agents reply with AI translation | English, with the AI answering in the customer's language |
Review the tiers twice a year. When a tier 2 market keeps growing, it's time for a native speaker, even a part-time one.
Step 3: Localise the content customers need most
Your AI agent and your team can only answer as well as the content behind them. Prioritise:
- Shipping: delivery times and carriers for each country, customs and duties for non-EU destinations.
- Returns: how to return from abroad, who pays, and local rules. In the EU, for example, consumers generally have 14 days to withdraw from an online purchase.
- Payment: local methods and currencies.
- Sizing and product details: units, size conversions and local naming.
Adapt rather than translate word for word. "Free returns within 30 days" means something different if the customer has to pay international postage.
Translate the answer, not the page. A short, local shipping answer is more useful than a perfect translation of your home-market policy.
Step 4: Let the AI agent answer in the customer's language
Modern AI agents detect the language a customer writes in and answer in it, even when your help content is only in English. For tier 2 and tier 3 markets this one feature does most of the work.
Set it up carefully:
- Test in each language with 15 to 20 real questions from your inbox, ideally checked by someone fluent.
- Watch policy terms, such as refund, exchange, store credit and withdrawal, which translate inconsistently.
- Keep the handoff clear. When a person takes over, the customer shouldn't suddenly get replies in another language.
In Tidio, Lyro answers natively in more than 45 languages from the same knowledge base, so one set of content covers every market.
Step 5: Give agents translation they can trust
When a person takes over, AI translation in the inbox lets an English-speaking agent read the message and reply in the customer's language. To keep quality high:
- Write saved replies in your main language, then have them reviewed once by a native speaker in each tier 1 and tier 2 language.
- Keep sentences short and plain in replies that will be translated. Idioms and jokes rarely survive.
- Escalate sensitive cases, like complaints, legal threats and large refunds, to someone fluent, or translate with extra care.
Step 6: Cover the time zones
Customers in other time zones message while you sleep. Options, from cheapest:
- AI agent around the clock for common questions in every language.
- Honest out-of-hours messages with the time your team will reply, in the customer's time zone if your tool supports it.
- Shifted hours for one team member if a large market sits several hours away.
- A part-time native agent in your biggest foreign market, which also covers the language.
Step 7: Meet customers on local channels
Channel habits differ by country. WhatsApp is everyday infrastructure in much of Europe and Latin America, while in other markets email or website chat is still the main channel. Check where your foreign customers already write, and connect those channels to the same inbox rather than adding separate tools.
Common mistakes
- Translating everything at once. Spend on the markets that bring revenue.
- Auto-switching the whole site to a language you can't support in conversations.
- Using different policies per language by accident, because one translation is outdated.
- Never checking quality. A monthly spot check of ten AI answers per language catches most problems.
- Forgetting returns. International returns are where foreign customers most often need help, and where bad answers cost the most.
The best tools for multilingual e-commerce support
We compared how each tool handles AI answers in other languages, translation for agents, multilingual help content and setup effort for a small team.
Tidio covers both sides of multilingual support in one inbox. Lyro detects the customer's language and answers natively in more than 45 languages from your existing help content, and when a person takes over, agents can translate customer messages with AI in one click. Chat, email, Messenger, Instagram and WhatsApp all work the same way, with store data from Shopify, WooCommerce or BigCommerce in each conversation.
Why we like it
- AI agent answers natively in 45+ languages
- One-click AI translation for agents in the inbox
- Works across chat, email and social channels
- Quick to set up without a localisation team
Watch out for
- AI conversations are billed by volume, which adds up across many markets
- Fully localised, per-language help centers need more work than in Zendesk
Zendesk offers live conversation translation in its agent workspace for chat, messaging and social channels, plus a multilingual help center with per-language article versions. It suits retailers with regional teams and a dedicated admin.
Why we like it
- Mature multilingual help center
- Live conversation translation in the agent workspace
Watch out for
- Setup and maintenance need a dedicated owner
- Several translation and AI features sit on higher plans or add-ons
Gorgias can translate inbound and outbound ticket messages automatically, with each agent choosing their own language preferences. A good option if you already run Gorgias and sell into a handful of European markets.
Why we like it
- Automatic ticket translation, set per agent
- Deep Shopify order data
Watch out for
- Help center covers fewer languages natively; more need third-party apps
- Ticket-based pricing rises with volume
Crisp's LiveTranslate translates conversations in real time across chat, email and messaging channels in its shared inbox, and the chat widget adapts to the visitor's browser language. A solid low-cost choice for small teams.
Why we like it
- Real-time translation across all inbox channels
- Flat workspace pricing
Watch out for
- Fewer e-commerce automations and store actions
- AI agent features are less mature than the leaders
A plan for the first month
- Week 1: pull orders by country and conversation languages, and set your tiers.
- Week 2: localise shipping, returns and payment pages for tier 1 and tier 2 markets.
- Week 3: switch on an AI agent that answers in the customer's language, and test it with real questions in each language.
- Week 4: turn on agent translation, have a native speaker review your top saved replies, and add out-of-hours messages for each region.
Start with Tidio if you want the AI agent and translation in one inbox. Brands with high volume across many markets can have Tidio's services team set up and maintain Lyro and its knowledge on the Premium plan, while Zendesk remains the stronger choice if you need fully localised help centers for every language. For the rest of your setup, see our e-commerce customer service best practices.
Frequently asked questions
Is machine translation good enough for customer service?
For most everyday questions, yes. Modern AI translation handles shipping, order and returns conversations well. It is less reliable for legal terms, product-specific vocabulary and tone, so have a native speaker review your saved replies and policies once, and keep sensitive cases with a fluent person.
Which languages should an online store support?
Start with the languages of the countries that make up most of your orders and the languages customers already use when they write to you. For many European stores that means English plus German, French or Spanish. Add more only when you can answer in them as well as in your main language.
Should I hire native speakers for customer support?
For your one or two biggest foreign markets, it's worth it, at least part time. For smaller markets, an AI agent answering in the customer's language plus agent-side translation covers most needs at a fraction of the cost.
Do I need to translate my whole help center?
No. Translate the pages customers need before and after buying (shipping, returns, payment, sizing, contact), and adapt them to local carriers, delivery times and consumer rules. Tools like Tidio can then let the AI agent answer in the customer's language from that content.