The difference between an amateur translation and a professional localization isn’t grammar. It is terminology.
If you are using AI to translate your website, product catalog, or app interface, the AI will provide grammatically perfect translations. But grammar isn’t enough. The AI needs to know how your specific company speaks.
If your software has a feature called “Smart-Sync,” you don’t want the AI translating it literally into Spanish on one page and French on another. You want it locked in as “Smart-Sync” globally.
To achieve this, you need a Translation Glossary. This guide will show you how to build a lean, highly effective glossary that forces any AI translation engine to respect your brand voice while following a professional CSV translation workflow.
Rule 1: Keep it Lean (The 50-Word Limit)
The biggest mistake teams make when building a glossary is treating it like a dictionary. They add hundreds of common nouns and verbs.
Do not do this.
When you overload an AI with generic rules, you restrict its ability to construct natural, flowing sentences. A glossary is not for common words; it is for exceptions. A world-class corporate glossary rarely exceeds 50 terms.
You only want to add a word if the AI’s default translation would be actively incorrect for your brand. There is a mechanical reason too: every term you save is sent along with every matching translation request, so a bloated list is paid for on each file rather than once when you write it.
- Hundreds of common nouns and verbs
- Every term travels with every matching request
- The engine is fenced in where it should be fluent
- 20 to 50 terms the AI would get wrong
- Brand names, interface labels, house jargon
- Everything else is left to the engine
Category 1: The “Do Not Translate” List (DNT)
Start by listing the words that should never change, regardless of the target language. These are the easiest to identify and the most critical to protect.
- Company Name: AI Glot stays AI Glot.
- Trademarked Product Lines: Nike Air Max does not become Nike Air Maximum in French.
- Proprietary Technologies: If your vacuum uses “Cyclone-V Tech”, keep it that way.
A “do not translate” entry does not freeze the row it appears in. It freezes the term, and the sentence built around it is translated as normal.
Nike Air MaxSmart-SyncRuns Smart-Sync every hour.Lance Smart-Sync toutes les heures.Category 2: UI and Navigation Elements (For Software/SaaS)
If you are translating an app interface, consistency is a matter of user experience. If a user clicks a button called “Repository,” they expect the corresponding page title to also say “Repository.”
Identify your core interface nouns and verbs:
- Nouns: Dashboard, Workspace, Settings, Repository, Profile.
- Verbs: Deploy, Commit, Sync, Subscribe, Cancel.
Decide exactly how these should be translated in your target language and lock them into the glossary.
Category 3: Industry Jargon and Acronyms
Every industry has specialized terminology that general AI models might misunderstand.
- Marketing: SEO, ROAS, CTR, Landing Page.
- E-commerce: SKU, Dropshipping, Checkout, Cart. Specialized product catalog translation often requires keeping these terms in English.
- Tech: SaaS, API, Endpoint, Webhook.
Decide if you want these acronyms localized (e.g., in French, CEO is often translated to PDG) or if you want to force the AI to keep the English acronym because it’s the industry standard.
Category 4: Brand Tone Preferences
Sometimes, a word has multiple valid translations, but your brand prefers one over the other.
For example, the English word “Customer.”
- A luxury brand might want it translated to the equivalent of “Client.”
- A tech startup might prefer “User.”
- A retail store might prefer “Shopper.”
If a core noun defines your relationship with your audience, put your preferred translation in the glossary to override the AI’s generic choice. This is where a glossary earns its place: not on words with one obvious translation, but on words where two answers are both defensible and only one is yours.
How to implement your Glossary
Once you have your list of 20 to 50 terms, it’s time to apply it to your workflow.
If you are using a dedicated file translation platform like AI Glot, implementing the glossary is built into the system. It is the best way to translate Shopify products while keeping your terminology locked.
- Navigate to the Glossary section.
- Select your language pair (e.g., English to French).
- Input your Source Terms and Target Terms (If it’s a “Do Not Translate” term, just put the same English word in the Target field).
- Run your translation batch.
The platform will parse your CSV, and whenever the AI encounters a term from your glossary, it will strictly apply your rule before translating the surrounding text.
You do this once, not once per batch. The glossary lives in the workspace, attached to its language pair, and it is picked up automatically when the plan for a new file is built. It is read at that moment and held for the whole run, so editing a term while a batch is translating changes the next batch and not the one in flight.
- 1Add the termOnce, for one language pairIn your workspace
- 2Upload any later fileNext week, next quarterAny supported format
- 3Approve the planThe glossary is read hereSnapshot taken
- 4Every string obeys itNothing to re-enterIn the output
By taking 15 minutes to define your core terms, you guarantee that your brand voice remains consistent across thousands of pages and products. Sign up to AI Glot to see how you can apply this to your next batch.