Translations LLM

HOW IT WORKS

Recognisable in every market

There is more generic content online than ever, and much of it is written by the same AI tools. AI lets you work faster and smarter, but your product texts still need to carry your value proposition and what makes you different. If they sound like everyone else's, customers have no reason to remember you. They need to recognise your brand in Denmark, Finland and every other market you sell in.

Anyone can give an AI tool a manual or style guide. The difference is in knowing what to put in it, and in testing it until the output is right. With localisation on top, every text fits how customers in each market shop.

We start with one market, translate test batches and let native specialists go through every text. Each time they correct something, that rule goes into the setup. That could be a term the model keeps getting wrong, a size that needs converting or a phrase that sounds translated.

The goal is to keep feeding the model so it knows your brand well enough to do most of the work for you.

The more markets you sell in, the harder quality control gets. One person cannot correct every product text in five languages. With a trained model, every market goes through the same process, and a smaller percentage of your texts that the model is unsure about goes to a native specialist for review.

That saves you money on translation. You can spend it on native review of the pages where customers decide to buy, such as category pages and checkout. Texts that make customers trust you, and that shows in your conversion rate.

Portrait of Max

Translating for several markets?We will help you find the right setup

Max Riis Christensen CEO, MakesYouLocal

Book a meeting → Write us an email →

Localisation
Translation

"“With their extensive knowledge of our brand and products, we feel confident that we are in good hands with MakesYouLocal. They understand our brand and consistently deliver high-quality, well-crafted translations of our product descriptions across our different markets.”"

— Pernille Helsinghoff, eCommerce Specialist, Skechers

Skechers

Four Nordic markets, one translation process

Skechers' product texts are written in the US, but customers in Denmark, Sweden, Norway and Finland expect different measurements and expressions. We trained the model on Danish first, with brand terms, technology names and measurement conversions. Then we rolled the same setup out to the other three markets with test batches.

Native specialists proofread every batch, and each delivery needs fewer corrections than the last.

4 Markets

1 Shared process

Natives Proofread every batch

Read the full case →

Easy Management

Upload and delivery made easy Get them back ready to use.

You get more than just an LLM Trained in your tone-of-voice. You also get a platform that handles the whole process for you. You upload your texts in one place, and the finished translations come back to the same place, so you do not have to coordinate between agencies and tools.

Upload your texts

Log in to ExportStation and submit your files for the market you need.

The model translates with your setup

From the first line, the model uses your brand terms and tone of voice, in the same way across every market.

Native specialists review

Every batch is proofread while the model is learning. Later, only the texts the model is unsure about go to a specialist.

Download the finished files

The translations are delivered back in ExportStation.

Book a demo →

Frequently asked questions

Interested in knowing more about LLM and AI translations

What is an LLM? LLM stands for Large Language Model. It is trained on large amounts of text so it can write and translate every text imaginable. The biggest force is that it can adapt language at scale.

Standard LLMs are trained on general content from the internet. They know the language, but they do not know your brand. A custom LLM is trained on your own content, such as product descriptions, category pages, tone of voice and terminology. Over time, it learns how you write. It knows you use "Children" and not "Kids", that your texts need gender-neutral wording and that your product names follow specific rules.

How does proofreading work?

The goal is to make the LLM translate a big part of your texts. In the beginning more text should go through a specialist to ensure queality and check that your terms are used correctly, or that measurements and product names follow your rules.

Every correction is added to the setup, so the model learns the mistakes and can be more and more precise.

Which texts should be proofread? A good place to start is with the texts your customers read just before they decide to buy. That is usually your category pages, checkout, campaign copy and the product descriptions for your best-sellers. If something sounds off there, it costs you sales, so these should be proofread by a native specialist.

Beyond that, you do not need to check everything by hand. If you use a LLM model it will flag the 5 to 20 percent of texts it is unsure about, and those go to a native specialist for review. That way, the specialists spend their time where it makes a difference, instead of reading through hundreds of product texts that are already right.

Ongoing or one-off: how does it work? Some brands send us new products on an ongoing basis, so translations run alongside their product launches all year. That works well if your catalogue changes often or you sell in many markets.

Others send texts when they need them, for example before a campaign, a new collection or peak season. If you book a meeting with us, we can help you find the setup that will fit your business.

Have more questions? Contact our team →