Most Arabic SEO advice available in English stops at "translate your pages and add hreflang." That is not architecture, it is a task. The real work in a Gulf bilingual site is structural: choosing between subfolder, subdomain and ccTLD before you build; handling the fact that ar-AE and ar-SA share a language code but do not share a SERP, a dialect, or a buying context; making right-to-left rendering a presentation concern rather than a crawlability one; and accounting for the messy reality that a large share of Gulf users search in English for Arabic-language businesses, and in Arabic for English-language ones. Get the architecture right and translation becomes a content problem. Get it wrong and no amount of good Arabic copy rescues it.
Language is the easy half. Structure is the half that determines whether the Arabic pages ever get found.
Key Takeaways
- Choose your URL structure before you translate anything. Migrating a bilingual site later is far more expensive than getting it right at build.
ar-AEandar-SAare different targets: different SERPs, different dialect expectations, different competitor sets. Treating "Arabic" as one market is the most common structural error.- RTL is a CSS and
dirattribute concern. If your Arabic content only renders correctly through JavaScript, you have made a rendering choice into a crawling problem. - Arabic search queries arrive with and without diacritics, with multiple valid spellings, and frequently in Latin transliteration. Plan for the variant set, not the canonical form.
- Cross-language search behaviour is real and heavy in the Gulf: English queries for Arabic businesses, Arabic queries for international brands. Both directions need coverage.
- Arabic copywriting needs a native speaker. Machine translation with a light edit reads as machine translation with a light edit, and users notice immediately.
Start With the Decision You Cannot Easily Reverse
The three structures
Subfolder (example.com/ar/) keeps everything on one domain and one authority pool. Subdomain (ar.example.com) separates the property while sharing the root domain. ccTLD (example.ae, example.sa) is a fully separate domain with its own authority and a strong geographic signal.
What actually drives the choice
It is rarely an SEO question in isolation. It is a question about whether you are one business serving a region or several businesses that happen to share a brand, and whether you have the resources to build authority in more than one place.
The Decision Table
| Structure | Authority handling | Geo/language signal | Best when | Real cost | Watch out for |
|---|---|---|---|---|---|
Subfolder /ar/, /ar-ae/ | Single pooled domain authority: strongest default for most brands | Moderate; relies on hreflang and Search Console settings | One brand, one entity, regional site, limited link-building capacity | Lowest. Same CMS, same hosting, same infrastructure | Locale folders proliferating without content to justify them |
Subdomain ar.example.com | Treated as related but partially separate; authority does not pool as cleanly | Moderate | Genuinely separate teams, separate CMS, or separate tech stack per language | Medium. Duplicated infrastructure and often duplicated SEO effort | Two properties to maintain, and one usually gets neglected |
ccTLD example.ae / example.sa | Fully separate. Each domain builds authority from zero | Strongest geographic signal available | Separate legal entities per market, market-specific pricing, local trust matters commercially | Highest. Separate domains, separate link building, separate everything | Underestimating that you now run two or three SEO programmes, not one |
Subfolder + per-country locale /ar-ae/, /ar-sa/ | Pooled, with country separation inside it | Good, when hreflang is correct and content genuinely differs | You need ar-AE and ar-SA differentiation without ccTLD cost | Medium. Content duplication risk is the main tax | Near-identical ar-AE and ar-SA pages: if content does not differ, do not split |
For most edtech and startup brands entering the Gulf, subfolder with locale-level splits only where content genuinely differs is the right default. ccTLDs are for businesses with local entities and local commercial reasons, not for businesses that want to look local.
The ar-AE vs ar-SA Problem
This is the part that generic international SEO guidance handles badly.
Same language code, different everything else
Modern Standard Arabic is the written baseline across both. But search behaviour is not written in MSA-neutral form. Query phrasing carries dialect. Gulf Arabic differs meaningfully between the Emirates and the Kingdom, and it differs again from Levantine and Egyptian Arabic, which dominate a lot of the general Arabic web content that models and search engines have absorbed.
Different SERPs, genuinely
The results a user sees in Riyadh and the results a user sees in Dubai for the same Arabic query are not the same. Local businesses differ, local publishers differ, and the competitive set differs. If you optimise once and assume coverage, you are optimising for whichever market your data skewed toward.
When to split and when not to
Split when the content genuinely differs: pricing, regulation, service availability, local proof. Do not split to create a second URL with the same content and a different flag. That produces duplication cost with no targeting benefit.
hreflang discipline
If you split, hreflang must be complete and reciprocal across ar-AE, ar-SA, en-AE, en-SA and your x-default. Partial hreflang is worse than none because it sends confident wrong signals.
RTL Rendering vs Crawlability
Keep them separate concerns
Right-to-left is handled with dir="rtl" and CSS logical properties. That is it. It is a presentation layer decision.
Where it goes wrong
Sites that implement RTL by swapping content client-side, or that serve Arabic through a JavaScript layer that assembles text after load, turn a styling decision into a rendering-dependency problem. Server-render the Arabic content. Test what a crawler actually receives, not what your browser displays.
Practical checks
- Does the Arabic page return Arabic content in the raw HTML response?
- Is
lang="ar"anddir="rtl"set on thehtmlelement for Arabic pages? - Do Arabic URLs work, and are they consistent? Decide between Arabic-script slugs and transliterated Latin slugs and hold one convention.
- Are titles, meta descriptions and structured data localised, not just body copy?
Render your Arabic server-side and verify what the crawler receives. What your browser shows is not evidence.
Diacritics, Spelling Variants and Transliteration
The variant problem
Arabic queries arrive in many surface forms for the same intent: with and without diacritics, with alef variants written differently, with and without the hamza, with hybrid Arabic-English phrasing. A keyword list built from one canonical spelling will systematically undercount demand.
Transliteration queries are real volume
Gulf users type brand names, place names and category terms in Latin characters constantly. If your brand or category has a common Latin transliteration, that is a query set you should be covering with English-language pages, not ignoring because it "isn't Arabic."
Practical handling
- Build your keyword set with variants grouped by intent rather than treated as separate keywords.
- Do not stuff variants into a page. Cover the dominant form naturally and let the search engine's own normalisation do most of the work.
- Cover genuinely distinct transliterations with distinct content only when the intent differs.
Cross-Language Search Behaviour
This is the observation most bilingual site plans miss entirely: language of query does not reliably indicate language of preference.
English queries, Arabic businesses
A user perfectly comfortable in Arabic will often search in English: particularly for technical, professional, software and B2B categories, where English is the working vocabulary. Your English pages need to rank for those queries and then offer a clear, discoverable Arabic path.
Arabic queries, international brands
The inverse happens for consumer, government-adjacent, and everyday-life categories. An Arabic query landing on an English-only page is a lost visit.
The design implication
Language switching must be visible, persistent and non-destructive: switching language should keep the user on the equivalent page, not dump them on the homepage. That single behaviour probably costs more Gulf conversions than any ranking factor.
Content Production Reality
Machine translation is a draft, not a deliverable
Translated Arabic that has not been through a native speaker reads wrong in ways that are hard to articulate but instantly obvious to a reader. In a market where trust drives conversion heavily, that is a commercial problem, not a quality nicety.
What I do and do not do
I build the architecture, the keyword and intent model, the technical implementation spec, the internal linking structure and the measurement frame. I do not write the Arabic copy. That needs a native speaker, ideally one from the specific market you are targeting. Saying otherwise would be selling you something I cannot deliver well.
Structured data and AEO
Localise your structured data alongside your content. As assistant-driven search grows, cleanly structured bilingual content is disproportionately valuable: Search Engine Land tracks this shift closely, and HubSpot publishes useful ongoing research on content structure and discovery behaviour.
Measurement
Separate your reporting by locale from day one
Aggregate Gulf reporting hides everything. Split ar-AE, ar-SA, en-AE, en-SA in Search Console and analytics before you need the data, not after.
Watch the cross-language funnel
Track language switches. If a large share of users landing on English pages immediately switch to Arabic, your acquisition targeting and your content language are misaligned.
Give Arabic pages time
New Arabic content on a domain with no Arabic history takes longer to establish than the equivalent English content on the same domain. Budget patience accordingly.
FAQ
Should I use a subfolder or a ccTLD for Arabic? Subfolder for most brands: it pools authority and costs least. ccTLD when you have separate legal entities, market-specific commercial terms, or a genuine local-trust requirement.
Is ar-AE really different from ar-SA for SEO? Yes. Same language code, different SERPs, different dialect expectations, different competitors. Split them only when your content genuinely differs.
Does RTL hurt SEO? No, when implemented as a CSS and dir attribute concern. It hurts when Arabic content depends on client-side assembly to appear at all.
Should my URLs be in Arabic script? Either convention works. Pick one and apply it consistently. Mixed conventions across a site cause more problems than the choice itself.
How do I handle diacritics in keyword research? Group variants by intent rather than treating each spelling as a separate keyword. Write naturally in the dominant form.
Do Gulf users actually search in English? Heavily, especially in professional, technical and B2B categories. Cover both directions rather than assuming query language matches preference.
Can I use machine translation for Arabic pages? As a first draft, with native-speaker editing before publication. Unedited machine Arabic damages trust in a market where trust drives conversion.
Do I need hreflang if I only have Arabic and English? Yes, and it must be reciprocal. Partial hreflang implementation is actively harmful.
How long until Arabic pages rank? Longer than English on the same domain if the domain has no Arabic history. Plan in quarters.
What is the most common mistake you see? Treating "Arabic" as one market and one audience, then wondering why performance in one Gulf country is fine and the other is invisible.
If you are planning a bilingual Gulf site and want the architecture decided properly before anyone starts translating, that is the stage where I add the most value. More on how I work at younusfardeen.com, bring your current structure and target markets and I will tell you what I would change and what I would leave alone.