Introducing Weglot's AI Language Model: Smarter Translations, Tailored to Your Brand

At Weglot, we're constantly rethinking what going multilingual could look like. That journey just took a big step forward.

We're excited to introduce the AI Language Model: a new feature that gives you more control, better customizations, and faster improvements to your website translations. All powered by generative AI and enriched with your brand's unique data.

Why We Built It

Until now, every Weglot user worked from the same high-quality base: a top-tier translation engine like DeepL, Microsoft, or Google. But we wanted to take things further. Could we offer our users translations out-of-the-box that already sound like them?

As Justine, our VP of Product, put it: "We tested a generative AI MVP and immediately saw stronger results: it lets our users shape how their translations sound and behave."

In short, it doesn’t need to end with generic quality. With the new AI Language Model, you get translations that are context-aware, brand-aware, and fully customizable.

How It Works

The model learns from:

• Your glossary
• Your manual translation history
• Your custom instructions, like:
  
  • "Use a professional tone"
  • "Keep headlines short"
  • "Sound friendly but formal"
  • “Keep proper nouns and brand terms in English for all languages”

And best of all, it starts working from the first translation. You don’t need to train it or upload anything. Just add your instructions, and the model adapts instantly.

Designed for Flexibility

Besides offering better quality translations, we also wanted to offer more control. Whether you're localizing blog posts, product pages, or UI elements, you can:

• Apply your model settings across multiple pages or projects
• Refine outputs in bulk (up to 10,000 translations at a time)

All of this is part of a much larger shift.

How This Fits Into Weglot’s Long-Term Vision

Weglot as a product is always evolving to make website translation as easy and hands-off as possible. Each new feature is driven by real-world feedback from users managing multilingual content at scale. So as technology rapidly changes the way people search and brands are viewed, our product continues to evolve in ways that support that shift.

And with every release, we aim to remove more friction from the translation workflow without compromising on quality or control.

That brings us to our long-term vision: a Weglot that’s natively built around AI. This means:

• A more intuitive user experience
• A review workflow for collaborative editing
• Smarter automation (like automatic shortening or content scoring)
• Brand-aware translation behavior out of the box

As Rémy, our co-founder and CPO, shared: "We want to give teams the tools to shape translations around their brand. This feature is part of that shift."

How to Get Started With Your AI Language Model

To find it, go to your Weglot Dashboard > scroll down to Settings > click on “Language model”.

Then, to see it in action, go to Translations > Languages > choose your desired language pair.  

Choose a translation string, then select the sparkle icon on the menu at the bottom to apply your AI language model to it.

You’ll see a prompt reminding you of your AI quota and confirming that you’d like to enhance your translation with AI. Click “Yes, apply AI” to proceed.

And there you go! Your translation string has been improved with AI.

You can also rephrase and shorten the text until you’re satisfied with the output.

What’s Coming Next?

The AI Language Model is now available directly in your Weglot Dashboard for all Weglot users. Enhanced workflow capabilities are in development, and as always, we’ll continue evolving the feature based on your feedback.

This marks an important move toward more adaptive, brand-aware translation — and a meaningful step forward for managing multilingual websites with precision and control.

Ready to try it? Log in to your Weglot Dashboard and start customizing your translations today. And if you have any suggestions, we’d love to hear them, which you can leave on our roadmap.