The most common way businesses enter Persian- and Arabic-speaking markets online is also the way that fails: take the English site, translate the keywords, publish, wait. The content is grammatically fine, professionally translated — and invisible, because translation produces correct language, not search language. People don't search the way translators write. Having built and ranked content natively in both languages, here's where the literal approach breaks.
People search in speech, not in dictionary terms
A translator renders "affordable web design" into formal, correct Persian or Arabic. A real user types the colloquial phrase they'd say — often a different word entirely, frequently mixed with an English loanword (سایت for website, دیزاین borrowed outright). Keyword research has to start from real query data in the target language, not from an English list run through translation — because the translated term and the searched term are routinely different words with different volumes.
Arabic isn't one search market
Written Arabic content defaults to Modern Standard Arabic, but search behavior is shaped by dialect: an Egyptian, a Saudi, and a Moroccan may reach for different everyday words for the same product. MSA is usually the right writing register, but keyword targeting has to know which regional variants your actual buyers use — "Arabic SEO" done as one undifferentiated market quietly optimizes for nobody.
Persian has a character-level trap: the ZWNJ
Persian compounds are joined with a zero-width non-joiner (ZWNJ) — an invisible character that changes how a word is written and, critically, how it's tokenized. The same compound typed with a ZWNJ, with a space, or fused solid are three different strings, and real users type all three. Add the multiple Unicode forms of common letters (Arabic ي vs Persian ی) that keyboards produce inconsistently, and naive keyword matching misses a large slice of real queries.
ONE WORD, THREE STRINGS
ZWNJ-joined · space-separated · fused solid — three different tokens to a search engine, all typed by real users. Content and metadata need to be written, and sometimes technically normalized, with these variants in mind.
Transliteration is a real query pattern
A meaningful share of Persian speakers type Persian words in Latin characters ("Finglish"), and Arabic speakers do the same ("Arabizi/Franco"), especially on desktop keyboards. These queries won't appear in a translated keyword list at all, but they're how actual customers search — worth knowing, measuring, and occasionally targeting.
The technical layer fails quietly too
Even perfect native content underperforms on a site that isn't built for it: RTL layouts flipped by CSS hacks instead of proper logical properties, Latin slugs on Arabic pages (or broken percent-encoded ones), hreflang tags that don't reciprocate (a single error makes search engines ignore the whole language cluster), and fonts that render Persian text in an Arabic-styled fallback readers instantly find off. This is why we treat multilingual SEO as an engineering discipline: the content and the architecture rank together or not at all.
What actually works
Native keyword research from real query data in each language; content written by (or reviewed by) native speakers who write for the market, not translated at it; technical RTL and hreflang done correctly once; and separate measurement per language, because averages across languages hide which market is actually moving.
We deliver SEO natively in English, Persian, and Arabic.
Hamed Azimi
Founder, NexusWave
Hamed Azimi is the founder of NexusWave Technologies, a Vancouver-based AI-native digital agency and product studio.
