Arabic Forms with AI: Dual-Script Names, Hijri Dates, and RTL Layouts That Actually Work

Most form-filling tools claim to support Arabic. Read closer and you find a translated UI label and a left-to-right input box. Real Arabic form work is harder. The visa application, the insurance claim, the immigration packet — every one of them encodes data in three modes that English-only tools never had to handle:

1. Two scripts at once. Your name in Arabic and in Latin transliteration, and they must match the passport exactly.
2. Two calendars. Your date of birth in Hijri (which you might have grown up with) and in Gregorian (which the destination wants).
3. Two number systems. Eastern Arabic for display, Hindu-Arabic for parsing.

Get any one of those wrong and the form bounces. This post walks through how AI handles each, what still needs your eyes, and where the line falls between automation and personal review.

Dual-script names: the ICAO 9303 rule

The international standard for travel documents (ICAO Document 9303) defines the romanization rules every passport uses. The MRZ (machine-readable zone, bottom of the photo page) carries the canonical Latin form. The visa text field above it carries the Arabic form. These two together are the name your destination knows you by.

When AI reads a passport image, it pulls both:

• Arabic name from the photo-page text region: محمد عبد الله الفلاسي
• Latin transliteration from the MRZ: MOHAMMED ABDULLAH ALFLASI

Both go onto the DS-160, the Schengen application, or the Arabic CV. The form usually asks for "Full name in native alphabet" and "Surname / given names" as separate fields. AI fills both and shows them side by side for confirmation.

The non-obvious failure case: when your Arabic name has diacritics on the form but your passport photo page does not. The destination system strips diacritics for comparison, so مُحَمَّد and محمد match. But the form designer expects you to mirror what the passport shows. AI normalizes to the passport's form by default and gives you a one-click toggle if you want to add diacritics back.

Hijri-to-Gregorian: which calendar does the form want

There are three Hijri systems you might encounter:

• Umm al-Qura — used by Saudi Arabia and most diplomatic posts. Astronomical-calculation based. The current year is 1447 AH.
• Tabular Hijri — used in many Excel-style forms for simplicity. Can differ from Umm al-Qura by one or two days.
• Civil Hijri (Iran) — different month names and a solar-year basis. Used on Iranian forms only.

Most US, EU, and UN forms want Gregorian. AI converts in the background and keeps the Hijri original visible for your review:

• Hijri: 5 رمضان 1432
• Gregorian: 5 August 2011

When the form accepts Hijri (Saudi Absher, some Pakistani applications, internal forms in the Gulf), AI keeps the Hijri as the input and skips the conversion. The point is to never silently overwrite — the user always sees both, and the right format goes to the right system.

The numeral problem

Eastern Arabic numerals (٠ ١ ٢ ٣ ٤ ٥ ٦ ٧ ٨ ٩) are visually native to Arabic text and look correct on a printed form. They break almost every digital system: US Customs and Border Protection, Schengen visa portals, UNHCR registration, French Pôle Emploi, German Bürgergeld. All of them parse Hindu-Arabic only.

The rule AI applies: store as Hindu-Arabic, render as Hindu-Arabic in system fields, convert to Eastern Arabic only when the form's metadata explicitly asks for the display form (Saudi Absher, some Iraqi and Egyptian government PDFs, classical Arabic publications).

When you paste an Eastern Arabic phone number into a US system field, AI warns:

Detected ٠٥٠١٢٣٤٥٦٧ — converting to 0501234567 for system compatibility.

You see it, you approve it. No silent rewrite.

RTL review UX: what good looks like

A flipped English form is not RTL. A true RTL review layout has:

• Input cursor starting at the right edge of the field
• Tab order moving right-to-left through the form
• Validation error messages appearing to the right of the input, not the left
• Dropdowns opening rightward, not leftward
• Bidirectional text (Arabic name + English transliteration on the same line) re-flowing correctly without manual marker insertion
• Numbers reading left-to-right within an otherwise right-to-left line (the standard Arabic typesetting rule)
• Form sections progressing top-to-bottom, but each section's right-to-left ordering preserved

Most tools get one or two of these right. Few get all eight. FillWizard's review layer renders the filled form in a real RTL DOM (dir="rtl" on the section, not just CSS text-align), which means screen readers, browser shortcuts, and copy-paste all work the way an Arabic-native reader expects.

What still needs a human eye

Five things AI should not decide for you:

1. Which transliteration is yours. If your passport says Yusuf but your bank account says Yousef, AI cannot pick. List them all under "Other Names Used" and let the consular officer match.
2. Whether your Hijri date is Umm al-Qura or tabular. Birth certificates from rural Saudi Arabia and Yemen sometimes use the tabular form. If the year is right and the day is off by one, that is the giveaway. Confirm with the original document before exporting.
3. Religious affiliation on visa forms. Some forms ask for religion, some do not. The Schengen visa form has it as optional in some posts, mandatory in others. Read the post-specific guidance — AI flags it but does not pre-fill.
4. Marital status with Arabic-specific subtleties. "ملاك" (engaged with civil contract but no consummation) is a Gulf-specific status that does not map to Western "single", "married", or "engaged". The form often accepts one of the simpler English labels; pick the one your destination expects, not the one that is technically most accurate.
5. The interview itself. When the consular officer asks "What is your father's name?" in English, you answer in English with the MRZ transliteration. Not the Arabic, not the family-known nickname. AI prepares you, but the moment is yours.

Practical: a DS-160 in Arabic, end to end

A typical Arabic-speaking DS-160 applicant from Cairo, Riyadh, or Casablanca opens FillWizard with three uploads:

1. Passport photo page (in Arabic and English).
2. Prior US visa label (if any).
3. Travel summary covering the last five years.

AI extracts the dual-script name, converts the Hijri birth date to Gregorian, normalizes the Eastern Arabic phone digits to Hindu-Arabic, and pre-fills the 80 percent of DS-160 fields that are pure data. The review screen renders the filled form in true RTL: Arabic on the right, Latin transliteration on the left, validation messages anchored right, dropdowns opening rightward. You read top-to-bottom, right-to-left, in about 12 minutes. Then you copy the field list across to CEAC and answer the security questions personally.

For the broader DS-160 flow, see How AI fills the DS-160 in 4 minutes. For applicants from Spanish-speaking communities on the same form, see Spanish-language US immigration forms.

What to do this week

If you have an upcoming Arabic-script form — a Schengen application, a US visa, a job application packet — pull out your passport and one prior form you filled by hand. Photograph the passport, upload the prior form. Watch AI extract the dual-script name, the Hijri-converted dates, and the normalized numerals. The first time you see the review screen render top-to-bottom in real RTL, with errors anchored to the right and tab order flowing the way your eye actually reads, you stop wondering why filling Arabic forms used to feel like fighting the tool. It was the tool.