Arabic SEO in the UAE: finding the words your customers actually type

A keyword list is a record of how a particular group of people phrase a particular need. Translate it and you keep the need but discard the phrasing, which was the only part carrying information.

Three things move independently between the two languages. The wording: the most natural Arabic query for a service is frequently not the literal translation of the English one, and sometimes the literal translation is a phrase no native speaker would produce. The spelling: Arabic is typed with variations that English does not have, and those variations split your data. And the intent: the same service can attract a different question in each language, because the people asking are often at different stages or in different segments.

You can still use the English list — as a map of topics, not as a source of terms. It tells you that people ask about price, timeline, and who to hire. It does not tell you the words they use to ask.

The practical consequence is that Arabic keyword research is its own project with its own budget line, done by somebody who searches in Arabic themselves. It is not a step at the end of the English one.

Start with the searches you already own. If the site has any Arabic traffic at all, Search Console already holds real queries from real people — the single most reliable source available to you, because it is observed rather than estimated. Filter to the Arabic pages and read what came in, including the queries that brought no clicks.

Then read your own inbox. WhatsApp messages, contact-form submissions, and the first line of sales calls are a corpus of how customers phrase the problem before anyone has coached them. This is the fastest route to phrasing that no tool will suggest, and for a UAE service business it is usually sitting unused.

Then use autocomplete properly. Type the Arabic seed into Google and read the suggestions, then type it again with each letter of the alphabet appended, and again as a question — كم، كيف، أفضل، أين، هل. The suggestions are built from what people actually typed, which makes them evidence. Do the same on YouTube, where Arabic search behaviour differs noticeably from web search.

Then read the competition that is already ranking in Arabic — not the English agencies with an Arabic tab, the pages that genuinely rank for Arabic commercial queries. Note their headings and their phrasing, not their meta keywords. And finally, read Arabic questions in the wild: forums, Facebook groups, the comments under relevant Arabic YouTube videos. Buyers phrase things there without an audience.

Arabic is typed with a set of variations that have no English equivalent, and every one of them splits a keyword into several. The hamza is the biggest: أبوظبي is frequently typed ابوظبي, الإمارات as الامارات, إعلانات as اعلانات. Final ة is often typed ه — شركه for شركة. Final ى and ي are swapped constantly. Diacritics are essentially never typed at all. And the definite article ال may be present or absent in the same query from the same person on different days.

Search engines normalise some of this some of the time, and you should not build a strategy on an assumption about which. Treat it as an open question and answer it from your own data: look in Search Console at whether variants of the same term appear as separate queries for your pages. That answer is specific to your market and worth more than a general claim.

What this does not mean is spelling your page wrong to match. A page written in sloppy Arabic reads as sloppy to the customer you are trying to win, and that cost is immediate and certain while the ranking benefit is neither. Write correct Arabic.

What it does mean is that your research must merge the variants before you judge demand, or you will read one healthy term as four weak ones and drop it. And it means the natural home for the variants is your own observed query data, not the page copy.

Modern Standard Arabic is what serious commercial pages are written in, and it is what most people type when searching for a professional service. That is the safe default for a B2B or institutional page and you should not be argued out of it.

But a large share of UAE commercial vocabulary is borrowed English written in Arabic script, and buyers type it that way. ديجيتال ماركتنج sits alongside التسويق الرقمي; ديزاين alongside تصميم; سيو alongside تحسين محركات البحث. These are not mistakes to be corrected, they are queries. The usual answer is to write the page in proper Arabic and let the borrowed form appear naturally where it genuinely would — in a heading phrased as the question, in an FAQ, in a parenthetical — rather than either banning it or stuffing it.

Mixed-script queries are common too: the service in Arabic and the city in English, or the reverse. A bilingual site handles these better than a monolingual one simply by existing in both languages with correct hreflang, which is one of the quieter arguments for building both halves properly.

Dialect matters less in search than people expect — the UAE market searches for services largely in MSA or in borrowed terms — but it matters in the questions. “كم يكلف” will outrun a more formal construction. Write headings the way a person would ask, and reserve the formal register for the body.

Keyword tools report Arabic volumes that are frequently wrong, usually low, and sometimes zero for terms that visibly return competitive results. The causes are structural: the spelling variants above fragment counts, Arabic sampling is thinner, and several tools infer Arabic figures rather than measuring them. Treat the numbers as directional and never as a threshold.

What to trust instead, in order. Your own Search Console query data, because it is observed. Autocomplete depth — a seed that produces a long, specific suggestion list has demand behind it whatever a tool says. Whether the results page is commercially contested: if ads are running against an Arabic query, somebody is paying for that traffic and has measured it. And the presence of a real answer in the results: Google rarely builds rich results for questions nobody asks.

This changes the shape of the work. In English you can often rank a list of terms by volume and work down it. In Arabic you are better off picking topics you know your buyers care about — from your inbox, from sales calls, from the questions you answer weekly — and letting observed data from your own pages correct you over the following months.

That is slower to justify in a spreadsheet and more accurate in practice. It also means the first Arabic pages you publish are partly instruments: they exist to generate the query data that makes the next ones better targeted.

Title, meta description and H1 are written for the Arabic query, not translated from the English ones. This is where most bilingual sites give themselves away: an Arabic title that is a word-for-word rendering of an English title optimised for a different phrase. Write them last, after the research, in Arabic, from the query.

Each language needs its own indexable URL with correct reciprocal hreflang and a self-referencing canonical. Whether the Arabic slug is in Arabic script or shared ASCII is a genuine trade-off, not a rule: Arabic-script URLs are readable to the reader and become percent-encoded gibberish the moment they are copied into a message, while shared ASCII slugs keep the language pair trivially easy to match and share cleanly. We use shared ASCII slugs for that second reason. Either is defensible; drifting between them is not.

Right-to-left is a layout the page is drawn in, not a filter applied to a finished English page — and that includes the direction of icons, progress steps, carousels and anything with an implied reading order. Set the document lang and dir correctly, because assistive technology and search engines both read them.

On numerals, pick one convention and hold it. In the UAE, Western digits are the safer default in digital contexts even in Arabic copy, and prices in particular should be written in the numerals your reader scans without thinking. Mixing ٣٬٠٠٠ and 3,000 across a site looks like two people wrote it, because two people did.

Site-wide averages are useless to you here. An English half with thousands of impressions at position 80 and an Arabic half with hundreds at position 6 average into a number that describes neither.

The reliable separation in Search Console is by page: filter to URLs containing /ar/ and you have the Arabic half of the site by itself, with its own queries, positions and click-through. Do the same for /en/ and compare them as two properties that happen to share a domain. This filter is exact and needs no assumptions.

Inside that Arabic view, the numbers worth watching are different from the English ones. Position matters more than impressions early on, because the Arabic pools are smaller and a page can be genuinely winning on low volume. Click-through tells you whether your Arabic title and description are written for the query or translated from the English ones — a strong position with weak CTR is almost always a translated title.

And read the query list itself, not just the totals. It is the only place the spelling variants, the borrowed terms and the mixed queries show up as facts rather than as theories, and it is what makes the next round of research better than this one.

The translated title. The single most common and the easiest to fix: a page that ranks and gets no clicks because its title answers an English phrase in Arabic words.

The Arabic page that is not a page. Content swapped in by a script at the same URL, or hidden behind a language toggle that never changes the address, leaves nothing for a search engine to index or rank. If it does not have its own URL, it does not exist.

The machine-translated site with hreflang. The tags are correct and the content is thin, which produces a technically flawless second-class page. Search engines are not the only ones who notice; the customer notices first, and they simply trust it less without being able to say why.

Abandoning it at six weeks. Arabic pools are smaller and move on their own schedule; a page can sit unread and then start earning position steadily. Judge it on the Arabic view in Search Console at three months, not on site-wide traffic at six weeks.

And the reverse error, worth naming because it is the fashionable one: publishing volumes of generated Arabic to cover more terms. Thin Arabic is the most abundant thing in this market already — it is precisely what makes genuine Arabic content rank here. Adding to the pile competes with yourself. The same logic applies to being cited by AI answers, where specificity is the whole of the selection criterion.