Most PDFs on government and enterprise websites are visually fine but structurally empty.

Read the document

We take the original PDF apart and pull out everything in it — the words, the images, where each thing sits on the page. This is raw material: all the content, no structure yet. The original file isn't changed; this is just the input for the work that follows.

Rebuild the structure

This is the main repair. We work out what every piece of the document actually is — which lines are headings and at what level, which text is a plain paragraph, which items form a list, which cells form a table — and write that out as the invisible structure layer. We also set the reading order: the sequence the content should be spoken in, which is often not the order the marks happen to sit in the file. After this step a screen reader can move through the document the way a sighted reader's eye does — jump heading to heading, know when it has entered a table, hear a list as a list.

Fix the text itself

A PDF doesn't always store text as text. Often it stores only the drawing of each letter — the lines and curves of its appearance — without recording which letter it actually is. A separate hidden table is supposed to connect each drawing back to a real character, and that table is frequently missing or wrong. When it is, the page still looks perfectly readable to you, but a screen reader trying to turn the drawings back into letters produces gibberish, or reads nothing. We rebuild that drawing-to-character table for every font in the document, so the words a screen reader speaks are the words actually on the page.

Label the document as a whole

Three facts have to be set on the file itself. Its title — so it doesn't appear in software as "Untitled," leaving the user no idea which document they're in. Its language — so the screen reader uses the right pronunciation, instead of reading English text with, say, Spanish phonetics. And a conformance marker — a flag inside the file that formally states "this PDF was built to the accessibility standard," which is the thing compliance tools and auditors look for. We set all three.

Make the interactive parts usable

Links, form fields, buttons — a sighted user reaches these with a mouse. A screen-reader user reaches them only through the structure layer, moving from one to the next. If those elements aren't labeled in that layer, they are simply invisible to that user: a link they can't follow, a form field they can't fill. We label every link, field, and image so each one is announced and reachable.

Check the result — and prove it

We run the finished PDF through veraPDF, the independent checker, which tests it against the accessibility standards rule by rule and reports exactly what passed and what failed — including the small number of things the automated steps can't fully fix on their own, which it flags for a person to review rather than leaving silently wrong. That report ships with the document. It isn't our assessment of our own work; it's a third party's, and it's the evidence you put in front of an auditor.