Arabic AI readability checker
Can AI read your Arabic, or do tatweel, hidden characters and images break it?
- Checks your page
- Reads:HTML
- 1 rule
- Last updated
What we keep
The result appears here
Enter your page’s address and press “Check the page”. This is what the tool runs on it:
- Arabic text that machines read badlynot started
What it checks
- Tatweel (ـ) between two letters:
الـعـروضis not «العروض» to a search index. - Invisible characters inside words: zero-width space, word joiner, a stray byte-order mark, a direction mark between letters.
- Numbers that mix Arabic-Indic and Latin digits:
٢٠٢4. - Paragraphs with a mark on nearly every letter (fully vowelled), and letters in presentation forms (U+FB50 to U+FEFF).
- Pages with little Arabic text whose images carry Arabic sentences in their alt text.
Example
<h1>الـعـروض الـخـاصـة</h1>
<h1>العروض الخاصة</h1>
How to fix
- Write words plainly, without tatweel; if you want a stretched look, get it from a font or CSS, not from characters in the text (
letter-spacingbreaks Arabic letters apart: never use it). - Search the source for U+200B, U+2060 and U+FEFF and delete them; keep a direction mark between words only.
- Use one digit set in a number:
٢٠٢٤or «2024». - Keep full vowelling for the Qur'an and teaching texts; for ordinary pages write plain letters.
- Copy Arabic from a word processor or a PDF with care: they can paste presentation-form letters. Retype or normalise to plain letters.
- Put text in the page, not in images: an image can have an
alt, but a paragraph in a picture is invisible to a machine.
For example:
grep -rnP "\x{200B}|\x{2060}|\x{FEFF}" src/ templates/FAQ
Not as a penalty. It affects what the machines read: a word with tatweel or a hidden character may not match the query that people type, and an assistant cannot quote what it cannot read as text.
To the eye. To a tokenizer or an index, الـعـروض is a different string from «العروض». Use CSS or a font for the look.
No, for the Qur'an, poetry and teaching texts it is right. The check reports paragraphs where nearly every letter carries a mark, so you can decide.
Only whether a page with little Arabic text has Arabic sentences in its images' alt text, a sign that the text is in pictures. It does not read the pictures.
What this tool runs
- Arabic text that machines read badlyminor
ar-ai-readability
How we measure
We read the visible text of an Arabic page's HTML, leaving out code, and count each kind: words with tatweel, words with an invisible character, numbers that mix digit sets, paragraphs of 20 Arabic letters or more where marks are a quarter of the letters or more, words in presentation forms, and the pictures case (fewer than 60 Arabic letters and alt texts that are Arabic sentences of eight words or more). Each kind is reported once, with how many and the first. See the rule's page for the details.