E-commerce
July 1, 2026
Many customers change languages during a conversation. They start in French, write a sentence in English, use a foreign technical word, or paste an error message in another language.
The chatbot must follow without losing context. It must understand the intent, respond in the most helpful language, and avoid asking the customer to rephrase everything.
This guide explains how to manage multilingual conversations and code-switching in e-commerce, with a seamless and respectful experience.
Summary
Why do customers change languages?
A customer may change languages because they speak multiple languages, because an error message is in English, because a product uses international terms, or because they are copying information from an email.
This change does not mean the conversation has to start from scratch. The chatbot must maintain context, recognize the dominant language, and respond naturally.
A good multilingual chatbot follows the customer; it does not force them to choose a perfect language.
Convert over 2,000 customers on average per month with Qstomy.
The world’s 1st Shopify AI dedicated to customer conversion
Empowering 200+ e-commerce merchants
How do I choose the reply language?
The simplest rule is to respond in the customer's primary language. If the customer starts in French and adds an English term, the response can remain in French while using the exact term.
If the customer clearly switches to English, the bot can respond in English or politely ask for their preference if the change seems temporary.
How to handle technical words?
Some words do not need to be translated: checkout, tracking, refund, size guide, warranty, or SKU can appear in French conversations. The bot must understand these terms and explain them if necessary.
It can say: "By refund, do you mean the reimbursement of your order?" This clarification keeps the tone natural without correcting the customer.
How to avoid context errors?
The risk of code-switching is misinterpreting an intent. A customer might write "return" to refer to a product return, or "return" in a technical message. The bot must use context before responding.
If the intent remains ambiguous, it is better to ask a short question than to translate at random.
How do you manage copied and pasted messages?
Customers often paste an error message, a carrier email, or a bank notification in another language. The bot must analyze it, then explain it in the customer's language.
For example, if the banking message is in English but the conversation is in French, the bot can summarize in French what the message means and what action to try.
This capability is particularly useful during checkout and delivery tracking, where external systems sometimes display messages that the customer does not choose.
Which flow to follow?
The flow must preserve the context despite the language change.
Detect the dominant language and occasional changes.
Retain the intent, command, cart, and details already provided.
Respond in the language that is most helpful for the customer.
Clarify only if the word or phrase changes the meaning of the request.
Transfer with the original content if a human must take over.
Which messages should be used?
For an English term in a French conversation: "Yes, I can help you with the refund, so the reimbursement of your order."
For a complete switch: "I can continue in English if you prefer."
For a copied message: "The message indicates that the payment has not been validated. You can check the bank authentication or try another payment method."
When to transfer?
The transfer is useful if the request becomes sensitive, if the customer uses a language not covered by the bot, if a translation could change the legal meaning, or if the customer expresses strong frustration.
The bot must transmit the original message, the summarized translation, the detected language, and the order context.
Which KPIs should be monitored?
Track multilingual conversations, language switches, linguistic clarifications, transfers due to unsupported languages, and satisfaction by language.
This data shows whether the chatbot is truly serving international customers or simply translating without understanding.
Which mistakes should be avoided?
Avoid correcting the customer's language, restarting the conversation after a language change, or literally translating a term that has a specific business meaning.
The chatbot must remain flexible. A multilingual conversation should seem normal, not fragile.
How can Qstomy help?
Qstomy can follow multilingual conversations, maintain context, and respond in the language most suited to the customer.
The chatbot reduces friction for international customers without losing the accuracy of e-commerce answers.
Explore AI support, the AI sales agent or request a demo.
CODESWITCHbot Checklist (8 steps)
Sync MIXLANG-MAP #891: feed code_switch bot_wrong_lang
Policy CODESWITCHBOT-SUP: 6 rules FOLLOW DETECT RESPOND-MATCH
8 intents bot_codeswitch_*: flow CSB-1 to CSB-8
4 templates TPL-CODESWITCHbot-*: CONFIRM FOLLOW BILINGUAL HYBRID
supported_langs FR EN: expand according to market
KB Link #268: customer service answers by language
Red team switch: FR turn 1 EN turn 3 follow test
KPI Dashboard: codeswitch_bot_* section 9 + delta mixlang_
FAQ
Difference #891?
#891 = agent tickets language mixing. #892 = bot follow widget switch.
Difference #16?
#892 = dynamic per message. #16 = global multilingual strategy.
Always confirm on switch?
Config switch_confirm_on_change. Direct follow if progressive switch.
Difference #890?
#890 = FR typos. #892 = language change FR EN.
Going further
This week: enable DETECT-PER-MESSAGE FR EN, test switch turn 3, sync bilingual KB #268, measure codeswitch_bot_mixlang_deflect.
Enzo
July 1, 2026
