DeepL MCP: Keep Product Names and Customer Information Consistent
When the same product has a different Japanese name on every page, or an English notice turns “dispatch” into “arrival,” which version can customers trust? Even fluent translations can leave customers with different understandings of a product and its purchase terms. DeepL now offers connections that let you use its translation tools and company glossaries inside AI assistants such as ChatGPT, Claude, and Microsoft Copilot. To use them well, start by agreeing on the expressions your company needs to use consistently. One product description and 20 frequently used terms are enough to get started.
Product Names That Change in Translation
First, choose a product description you have already published and place the original and its translation side by side. Mark any inconsistent wording in product names, options, and shipping information. A translation tool may translate the same word differently to suit the context, but your company needs a standard wherever customers must recognize the same product or terms.
For example, “dispatched within three business days of ordering” is a different promise from delivery within that time. A translation must not add “free” to an exchange policy that only says exchanges are available. Review translations for consistent product identification and terms, as well as fluency.
For your first task, simply translate the original text in DeepL and compare the result with your company's list of approved expressions. You can establish this standard before connecting an AI assistant. Decide on the target language and the region where your customers are based. If approved wording already exists, check it before creating a new translation.
A Company Glossary: Start with 20 Terms
Create a blank table with these columns: “Korean original / Approved translation / When to use it / Wording to avoid / Reviewer.” Choose five items from each of the four groups below and replace them with terms your company actually uses. That gives you your first 20 entries.
- Product identification: product names, series names, color names, material names, size or specification names
- Purchase choices: purchase options, included items, add-on purchases, subscriptions, preorders
- Shipping information: expected delivery date, dispatch date, business days, shipping charges, delivery or collection method
- Customer support: exchanges, returns, refunds, warranty period, customer support
Decide whether to keep distinctive product names in their original language or use them alongside a local name. Ask the person responsible for the relevant business process to check terms such as “business days” that affect purchase conditions. Keep a separate table for each language and specify the translation direction. Korean-to-English and English-to-Korean each need their own standards.
Keep words and short phrases in the glossary, and style guidance and business rules in separate instructions. A DeepL glossary applies your chosen translations to specified terms. Keep instructions such as “Use polite language with customers” and “Do not add benefits that are absent from the original” alongside the glossary, but separate from the term pairs so they are easier to revise.
Do your product descriptions, quotes, and customer emails use the same wording for the same products and terms?
Have the responsible person approve the terms that most often cause errors first. When changing an expression, check the reason for the change and its effect on translations already in use.
Connect DeepL MCP to Your AI Assistant
Once your reference table is ready, consider a connection that suits your recurring tasks. MCP is a way for an AI assistant to connect to and use external tools. With DeepL's own connection, you can request translations in a conversation and apply glossaries and style rules on supported accounts.
Choose DeepL from your AI assistant's list of apps or connections, then sign in with your DeepL account. In ChatGPT, you can connect by installing the certified DeepL plugin. For an organizational account, check whether administrators allow the connection in both the AI assistant and DeepL. DeepL operates the server, so you do not need to build one yourself.
After connecting, try the following request with one product description that is suitable for public sharing.
Use DeepL to translate the product description below into Japanese. First, show me the available company glossaries and apply the one I select. Tell me which glossary and style rules were actually applied. Preserve the original numbers, units, and shipping terms, and list any wording that needs review separately from the translation. Do not modify the glossary.
The DeepL Free MCP connection does not include glossary or style-rule tools. Even with a paid per-user plan, check team permissions and supported languages. If glossary features are unavailable, start by having someone compare translations against the reference table. Even if you request several languages at once, usage is counted separately for each target language.
If an administrator has assigned a style profile to your account, its glossary and style rules take priority over the settings you request. Check which standards were actually applied alongside the translation result.
Document translation has limits depending on the file and your plan, so test with a short text first. Before entering customer information or material covered by confidentiality obligations, check your company's rules for using external services. Separating permission to request translations from permission to change the company glossary helps reduce unintended changes to your standards.
Check Numbers, Promises, and Formatting
Compare the result with the original in this order.
- Products: Have product names, options, colors, or specifications changed in a way that suggests a different product?
- Numbers: Do prices, currencies, quantities, units, and dates match, with no missing values?
- Promises: Have shipping, exchange, refund, or warranty terms become broader or more restrictive?
- Wording: Was the glossary actually applied, and does the wording sound natural in context?
- Formatting: Are table rows and columns, headings, links, and emphasis associated with the correct content?
DeepL supports preserving document formatting. Even so, longer translations can shift tables or separate product options from their descriptions, so open the final file yourself. Have someone who can read the target language compare important terms with the original.
Once you have established a standard for one product, extend it to shipping information and customer support emails. Record inconsistent terminology, errors in terms and conditions, and human corrections so your company can maintain consistent wording even when people or AI tools change.
Consistent product names and purchase terms depend on aligning your translation tools with your review process. AIB operates an independent AI comparison platform without favoring any particular AI company, helping business owners and professionals choose and apply AI that fits their work and budget. Tell us which translated expressions you keep correcting and how your team handles multilingual customer information. We can explore ways to use AI that suit your team.
Contact AIB about adopting and using AI
Sources and Further Reading
- Official announcement: DeepL is now available in Microsoft Copilot, ChatGPT and Claude via MCP
- Connection and availability: DeepL Remote MCP · Connect DeepL to ChatGPT
- How glossaries work: Managing Glossaries