Question 1

How does Localizethat decide which company context to send to the AI?

Accepted Answer

For each localization, Localizethat looks at the content type, source text, target language and market, then assembles a context package from the company knowledge that applies: the relevant brand voice, terminology found in or related to the source, claims and market rules that apply to that language and market, and previously approved translations for similar content. Everything else stays stored but is left out of the prompt.

Question 2

Why not just send all of our company context to the model?

Accepted Answer

A large workspace can hold hundreds of claims, thousands of terminology entries and years of approved translations. Sending all of it into every localization adds cost and noise, and increases the chance the model gives weight to information that has nothing to do with the content being localized. A focused context package keeps the relevant guidance visible.

Question 3

Do different types of context get retrieved differently?

Accepted Answer

Yes. A glossary term is not selected the same way as a brand voice, and an approved claim is not treated like a previous translation. Brand voice is chosen by localization and content type, terminology by presence and relevance in the source, claims and market rules by language and market applicability, and Translation Memory by matching against the source segment.

Question 4

What happens when a translation has already been approved?

Accepted Answer

When Localizethat finds a previously approved translation for the same source, that decision can be reused instead of asking the AI to produce another version. Segments with no previous decision are localized with AI. Either way the result goes to human review before it is finalized.

Question 5

Does our Translation Memory grow automatically?

Accepted Answer

Approved language memory grows from finalized work: when a localization is finalized, its source and target pairs become reusable Translation Memory for future localizations, with provenance recorded so you can see where an entry came from.

Question 6

Can AI output change our approved claims or company rules?

Accepted Answer

No. Brand voice, terminology, approved claims, forbidden claims, market rules and content type defaults are controlled company context. They only change when your team explicitly updates or approves them — never because AI generated a sentence or a reviewer edited one.

Question 7

If our knowledge base doubles, does our prompt size double?

Accepted Answer

No. Prompt size follows the content being localized, not the size of the stored library. A workspace with 50,000 approved translations does not need a prompt containing 50,000 translations — Localizethat retrieves the few that matter for the current content, so stored context can grow while the AI context stays focused.

Question 8

Can we see what context was used for a localization?

Accepted Answer

Yes. Each run records which context was retrieved, how much of the available company knowledge it represents, and how the request breaks down between context, source content and instructions.

The right context.
Not all the context.

Localizethat doesn’t give AI more context.
It gives it the right context.

Not every type of company knowledge works the same way.

More company knowledge shouldn’t mean more AI cost.

If it’s already approved, why solve it again?

Your context can grow without your rules changing themselves.

Translation Memory

Powerful underneath.
Simple where you work.

Context files are useful.
Structured context goes further.

Common questions about how context works.

Put the right context to work.

The right context.Not all the context.

Localizethat doesn’t give AI more context.It gives it the right context.