Written in one Arabic. Spoken in many.

We write in Modern Standard Arabic by default, or in Gulf, Egyptian, Levantine, Maghrebi, Iraqi, or Sudanese Arabic when your audience speaks it. Every report names the variety and the level that ran.

One standard, six major dialect groups

Modern Standard Arabic (MSA) is the shared written standard in every Arabic market, and our default. When you choose a dialect for content aimed at one market, we write in that market’s dialect group.

VarietyWhere it is spokenContent we suggest it for
Modern Standard ArabicAcross the Arabic-speaking world, as the shared written standardDocumentation, legal, medical, financial, product UI, help content, and text that serves several markets at once
GulfSaudi Arabia, the UAE, Kuwait, Qatar, Bahrain, and OmanConsumer marketing, social media, app microcopy, customer-service scripts, and conversational chatbot replies
EgyptianEgypt, and widely understood across the region through film and TVMarketing, social media, video, and entertainment
LevantineLebanon, Syria, Jordan, and PalestineMarketing, social media, and video
MaghrebiMorocco, Algeria, and TunisiaConsumer content for one named country, where Arabic often mixes with French
IraqiIraqConsumer content, customer service, and assistant testing
SudaneseSudanConsumer content, public information, and customer service

These are dialect groups, and usage varies inside each one: Saudi, Emirati, and Kuwaiti Arabic differ, and so do Moroccan, Algerian, and Tunisian. So your brief names the country for each content type, and the region or audience where usage differs inside it. The report names the variety we wrote in.

Need a variety we don’t list, such as Yemeni, Libyan, or Hassaniya Arabic? Ask us, and we’ll confirm whether we work in it.

One sentence, seven ways

The same short sentence changes its words for “want” and “now” across MSA and the six dialect groups. An AI that defaults to MSA, or mixes two of them, is the first thing a local reader notices.

MSA or dialect? A rule of thumb

If in doubt, MSA. Choose dialect on purpose, per market and per content type, and name the country in your brief.

Use MSA whenUse dialect when
One text must serve several Arabic marketsYou target one market and want to sound local
The content is legal, medical, financial, regulatory, safety, or technical documentationThe content is social media, ads, video, or influencer and community content
The content is product UI or help content (most brands; some consumer apps choose dialect)The content is a customer-service script or a chatbot reply meant to sound conversational
Your brand voice is formal: press, investor, and corporate contentThe content is entertainment, game dialogue, or character voices
You are unsureYour own local team already writes in that dialect

A mixed approach is common: MSA for the interface, dialect for marketing. Your glossary then holds an approved term for each variety.

Where AI goes wrong in Arabic

We know where AI breaks in dialect, and the cycle checks for exactly that: the AI is prompted for your variety, a dedicated check catches MSA drift and mixed dialects, and the devil’s advocate challenges every flag and asks for its evidence before the judge rules.

  • It slips back into MSA. Ask for a dialect, such as Egyptian Arabic, and AI tends to answer in MSA. Public evaluations, such as the WMT25 shared task, report the same tendency, so we treat it as a known weak spot. Source: Findings of the WMT25 General Machine Translation Shared Task (external site), section 7.4.
  • It mixes two registers in one sentence. When your brief sets one dialect voice, an MSA verb next to a dialect word reads as a slip to a local reader; see the example below. Mixing that your style guide allows on purpose is not flagged.
  • One word, two meanings. In Moroccan Arabic, العافية commonly means “fire.” In most other varieties, it means “good health.” A wellness campaign that keeps the word unchanged across markets can land very differently in Morocco.

Example

Brief: Egyptian Arabic, a social postText
SourceGet it now before the offer ends.
AI output, mixedاحصل عليه دلوقتي قبل انتهاء العرض.
Consistent MSAاحصل عليه الآن قبل انتهاء العرض.
Consistent Egyptianالحق اطلبه قبل ما العرض يخلص.

The mixed line pairs an MSA verb and noun phrase with the Egyptian word for “now.” MSA itself is where most into-Arabic errors land: gender agreement, register, numbers, and right-to-left order.

How the cycle handles Arabic

The same AI cycle runs on every Arabic file, at every level. This is how each step handles the variety:

  1. The brief names the variety. We fix the variety for each content type at intake: “Gulf (Saudi Arabia),” not just “Gulf.”
  2. A glossary entry per variety. The same concept can carry a different approved term in each variety.
  3. AI translation, prompted for that variety.
  4. A dialect-consistency check. It flags MSA drift, words from another dialect, and register mixing.
  5. The devil’s advocate and the judge. The devil’s advocate challenges every flagged segment, and the judge rules on it and records the reason.
  6. The report names the variety, and its Human review field shows none, check, edit, or full review, as in every report.

Human review is an optional add-on outside these steps, after the judge’s ruling and before the quality gate: + Human check, + Human edit, or + Full review, by a native linguist for that dialect.

Example

FieldEntry
Arabic varietyGulf (Saudi Arabia)
LevelAI
Human reviewNone
Quality gatePassed

Right-to-left, numbers, and mixed text

Arabic reads right to left, but product names, numbers, and placeholders often sit inside an Arabic sentence and read left to right. Automatic checks cover tags, placeholders, and numbers in every Arabic file, and locale checks cover Arabic punctuation, such as the Arabic comma (،) and question mark (؟). Every file comes back in your format, with its tags and placeholders intact.

Example: حدّثوا تطبيق Zarvelo إلى الإصدار 2.4، ثم افتحوا {settings_link}. Zarvelo is a fictional app.

Arabic across our services

Tell us which Arabic your markets need

Tell us your markets and content types, and your quote names MSA or the dialect for each one. Already translating into Arabic with AI or machine translation? A free check shows what the cycle would fix.