AreaAudit

Which languages each generator says it supports, and where

What the shape of a declared set says about a company

A declared set of site languages is a map of where a company expects buyers. Translating pages costs money per market and is maintained forever, so the set tends to be more candid than a feature page. As of 2026-09-12.

Reading a declared set as commercial evidence, and where to stopThe size of a set suggests how much has been spent on markets. Which languages come first suggests where revenue is expected. Whether codes carry a region suggests a policy. None of it is evidence about what a model writes or speaks.What the shape suggestsWhat it cannot reachSize of the setSpending on markets so farAnything about the outputWhich codes come firstWhere revenue is expectedAudience size in those placesRegion subtagsA policy about marketsTranslation qualityRegions absentWhere a company is not sellingWhere it could not sellThe reading stops at the edge of the website
Fig. 1 A declared set is strong evidence about a company and weak evidence about a product, and keeping those two readings apart is most of the work.
What each property of a declared set does and does not indicate. Recorded 2026-09-12.
Property of the setWhat it suggestsWhat it does not
Its sizeHow much has been spent on marketsAnything about the product
Which languages come firstWhere the revenue is expectedAudience size
Whether codes carry a regionA policy about marketsTranslation quality
Which regions are absentWhere the company is not sellingWhere it cannot sell

Inclusion rule. Properties of a declared set of interface languages, read as commercial evidence. None of them is evidence about generated output. Order. Most obvious property first.

1The ordering is more informative than the total

Sets in this category tend to nest: the first languages added are the same across companies of very different sizes, and what differs is how far down the sequence a company has gone. So two sets of the same size can describe the same plan at the same depth.

Where a set breaks the ordering, that is the interesting part. A set that skipped a language almost everybody else added first has made a decision worth noticing, even though nothing published explains it.

2The first languages track software spending, not speakers

The languages added earliest are not the largest by population. They are the languages of markets that buy business software, pay for it and expect a page in their own language.

That is why several very large language communities are declared far less often than small wealthy ones. A set is a map of revenue expectations rather than of audiences, and reading it as the second produces nonsense.

3Absences are the readable part for a buyer

A reader in a region can use a column like this immediately: it narrows the field to the companies that have paid to speak to them. That is a shortlist rather than an assessment.

What it cannot say is whether the missing companies would serve the market anyway. Declaring nothing is not a refusal, and several vendors with no declared set publish the most detailed output documentation in the field.

4Where the reading stops

At the edge of the website. A declared set says nothing about what a model writes or speaks, and among the vendors read here the companies with the longest sets and the companies with the longest voice lists barely overlap.

So the shape of a set is good evidence about a company and poor evidence about a product. Keeping those two readings apart is most of the work in reading a language claim at all.

An explainer about method and structure, not a count. No vendor number appears here that is not also in the register. The sourced material is on the register. Related: Why no single score, Check one yourself.