A Localization Review Checks Every Locale Is Heard Right

What We Keep Finding

Here is what a wrong language declaration sounds like. Your German site, marked as English, read aloud with English pronunciation rules. Every word comes out mangled at speed, the way a GPS mangles street names, except it is the entire page. To a sighted reviewer it looks perfect, which is why it survives so long.

Fair warning about that one, though, because it is the part of this job a machine does best. Whether a page declares a language, and whether the tag is well-formed, are among the small handful of things automated rules settle outright rather than flagging for a human to finish. Our free scan checks it. So if that is your whole worry, run the scan and keep your money.

The value here is everything the machine cannot reach, and it starts with a requirement most people have never read. Accessibility support is not a global property. The standard's test for it requires the interoperability testing to be done in the human language of the content. So a component proved to work with JAWS on your English pages is not thereby proved for your Arabic ones. That is not a technicality. It is the reason each locale needs its own listen, written into the standard.

Then there is translation drift, which is invisible by design. The visible text gets translated and the invisible layer does not. Alt text still in English. Button labels announcing in the wrong language. A right-to-left layout that mirrors the columns and not the reading order, which is a straightforward failure of two rules about sequence. And the screen reader itself changes by region, so a locale may not just be a different voice, it may be a different product doing the reading.

One thing this review is not, and it is worth saying because the temptation is strong. Subtitles in another language are not an accessibility accommodation. Captions are. Translating your video is a good idea and it is not this.