PDF language

Silktide checks that a PDF says which language it is written in. Software that reads a document aloud has to choose a voice and a set of pronunciation rules before it can say a word, and the document's declared language is what it chooses from.

The declaration is a property of the file rather than something printed on the page. This check covers the document's own language and the language of its bookmarks and metadata entries, all of which can be announced to a reader.

Why this matters

Without a declared language, a reader falls back on whatever the person's system is set to, and the result is a document read in the wrong accent and the wrong pronunciation rules. An English document read by a French voice is not lightly accented, it is largely unintelligible: the letter groups are pronounced by different rules, so words come out as sounds rather than words.

The problem is invisible to most people publishing documents, because it only appears for readers whose system language differs from the document's. That is precisely the audience most likely to be affected, and least likely to report it.

It matters for bookmarks too, which is a case people rarely think about. Bookmarks are the navigation panel of a long document, so they are the first thing a reader listens to, and mispronounced bookmark titles make the whole document harder to navigate before any of the content is reached.

Declaring the language also helps beyond speech. It drives hyphenation and spell-checking, tells search engines who a document is for, and lets translation tools work from the right starting point.

How to fix it

Set the language in the source document and re-export. In Word, use Review > Language > Set Proofing Language and apply it to the whole document; in InDesign, set it on the paragraph styles.
Or set it in Acrobat Pro through File > Properties > Advanced > Reading Options > Language.
Mark passages in other languages where a document quotes or includes another language at length, so a reader switches voice for them rather than reading them with the wrong rules.
Or run Silktide remediation, which detects the document's language and declares it. Note that remediation declares one language for the whole document; see Language coverage for what that means for mixed-language files.

Use the language the document is written in, not the language of the country publishing it. A Welsh-language document published in the UK is Welsh.

How Silktide tests this

Find the PDF documents linked from your website, and analyse each unique file once however many pages link to it.
Look for the document's declared default language.
Check the bookmark entries and metadata entries for a language of their own, where they need one.
Report the document when the declaration is missing, once per document however many entries are affected.

This check tests whether a language is declared, not whether the declaration is correct. A document written in German and declared as English satisfies it and still reads wrongly, which is why the declared language is confirmed against the content during AI review.

Troubleshooting

The document reads correctly for me

It will, if your own system language matches the document. The declaration only changes what happens for readers whose does not, so testing on your own machine is the one situation where the problem cannot appear.

The document contains more than one language

Declare the main language of the document, then mark the passages that differ. Doing the first alone is a large improvement, and is what this check asks for; marking individual passages is the refinement on top. See Language coverage.

The document is mostly numbers or a table

It still needs a language. Headings, labels, and captions are read aloud, and numbers themselves are spoken differently between languages: decimal separators, dates, and currency are all pronounced by the declared language's rules.