Arabic, Hebrew, Persian and Urdu PDFs to InDesign, in reading order

How much of it is backwards

Across 48 PDFs with Arabic, Hebrew, Persian, Pashto, Urdu, Kurdish and Yiddish in our test set, 4,021 of 10,336 right-to-left lines were held in another order from how they read, in 42 of the files.

A translator or a translation model gets those lines as nonsense, so this is the first thing to get right. It is easy to miss, too: a PDF viewer draws the page correctly whatever order the file holds, and the fault shows only once the text is copied out.

What we found in real files

What we do with a line

• Words and letters go in the order they are read, worked out from where they stand on the page.
• Latin words and numbers inside the line keep their own direction, so a name in Latin letters or a date reads properly.
• Marks stay on their letter, so Hebrew points and Arabic vowel signs do not wander.
• A line already in reading order is left exactly as it was.

Paragraphs, not just lines

Getting the order right also lets us tell paragraphs apart properly. A line in a right-to-left block starts on the right and ends on the left, and that is the edge that shows whether it ended early. Measured from the wrong side, headings ran into the text under them and last words stood alone on a line.

In InDesign

• Paragraphs and stories are set right to left, in the World-Ready composer as InDesign's own files do.
• A document that opens right to left gets right-to-left pages.
• Text in a font lacking Arabic or Hebrew is set in the Noto face for its script, named in your font list.
• Spaces between right-to-left words stay in the run with the words, so a CAT tool does not get a tag at every space. In one Urdu leaflet that took a paragraph from 33 runs to 2.

2021 and 2022 set it differently

In our comparison of InDesign versions, right-to-left lines stand on 2024 and 2025 where 2026 on Windows sets them, and on a Mac all but 0.1% do. On 2021 and 2022 about 12% stand elsewhere, so go through the pages again if a file moves to or from one of them.

Where to look twice

A PDF in an old custom-encoded font, where the layer holds Latin letters drawn as Arabic ones, has no reading to recover. A line mostly in Latin with a right-to-left word in it is left as the PDF holds it. A bracket alone on a line, with its partner on another, gives no way to tell how it was named. A table printed without rules may be read as lines, and a few ragged lines still part into paragraphs of their own. Check those by eye.

How to check a document in a minute

1. Copy a sentence out of the Word copy into a plain text editor and read it. Reading order shows there, whatever the page looks like.
2. Find a line with a Latin word or a number in it and see that it stands where the PDF prints it.
3. Find a line with brackets: each should open towards the text it holds.
4. Open the IDML and compare a page with the PDF. Text in another face than you expected is named under Type > Find Font.