Arabic Keyword Research for B2B: Dialects, Spellings and English Searches

Ask a sales team in Riyadh or Cairo how customers describe a product and you will hear several versions of the same request: an English model name, an Arabic generic term, the Arabic word spelled two ways, and a local word the marketing team never uses on the website. Keyword research for Arabic B2B markets is mostly the work of finding those versions, deciding which ones carry real demand, and placing them on the right pages without building a separate page for every spelling.

This guide is written for marketing managers and business owners selling technical products and services in Saudi Arabia, the UAE, Qatar and Egypt. It covers where the variation comes from, which data sources can and cannot separate it, and how to map it to a bilingual site safely. Once the vocabulary is settled, using it on the site is a separate job, which an SEO programme should also cover: our guide to writing Arabic service pages for Saudi buyers picks up where this one ends.

Four layers of variation in Arabic B2B searches

Treat every important topic as a cluster with four layers, and research each one separately.

1. Language choice: Arabic, English or both

Technical buyers often use English product names, standards and model numbers, while describing the problem or the service in Arabic. Engineers, procurement officers and business owners may also search differently for the same purchase. Do not assume one language dominates in your category; confirm it in your own query data, described below.

2. Transliteration

Many technical terms are written as English words in Arabic script alongside a translated Arabic term. In fire protection, as a hypothetical example, a team might find both a translated term for sprinklers (رشاشات) and a transliterated one, which itself appears in more than one spelling (سبرينكلر, سبرنكلر). The same happens with system names such as فاير الارم (fire alarm) and with model names: اف ام ٢٠٠ spells the FM-200 suppression agent in Arabic letters and writes the number in Arabic-Indic digits, while another buyer types FM200 or اف ام 200. Transliteration and digit form vary independently, so the research sheet needs a separate digits column (see the template below).

3. Spelling and orthography

Arabic users frequently type the same word in different ways, particularly on mobile keyboards:

• ة / ه at the end of a word, as in طفاية and طفايه (extinguisher) or مضخة and مضخه (pump)
• أ / إ / ا at the start of a word, as in إنذار and انذار (alarm)
• ى / ي at the end of a word, as in مستشفى and مستشفي (hospital)
• Attached prefixes and the definite article: صيانة, الصيانة, للصيانة

Search engines handle some of these variations better than others, and behaviour can change. Rather than relying on a general rule, test your own priority terms (see the method below).

4. Dialect and local vocabulary

Gulf and Egyptian buyers sometimes use different everyday words for the same thing, and trade vocabulary differs between markets and even between sectors. Generic word lists are unreliable here: a pair that holds for household products may not hold for a trade buyer, and published lists rarely say where their words came from. The reliable source is your own sales conversations, the WhatsApp messages, calls and quotation requests your team receives from each country, logged in a simple table before any word goes on a page.

A quick first signal, before the inquiry log, is Google’s autocomplete. On 31 October 2026 we requested Google’s suggestions for the same kinds of seed with the country set to Saudi Arabia (gl=sa) and then Egypt (gl=eg) and recorded the suggestions. The clearest contrasts are below; every Arabic term in the table is quoted exactly as Google suggested it.

What Google suggested on 31 October 2026: Saudi vs Egypt

How to record your own: type each seed into the Google search box with the search location set to the target country, in a private window so your own search history has less influence, and copy every suggestion with the date and the country into the sheet. Keyword Planner, with location targeting set to each country, adds term ideas as a second source. Autocomplete then tells you which forms to look for in the step that decides: your inquiry log.

Here is the table for one non-fire sector, HVAC supply and maintenance. Fill one row per word exactly as a buyer typed it, from anonymised inquiries, with no names, phone numbers or quantities:

The hypothetical row shows the format only; it is not client data and says nothing about how often either market uses the phrase. Pair rows for the same seed across markets. When Saudi and Egyptian buyers consistently use different words for one product, that is the clearest case for different headings on ar-SA and ar-EG pages. When the words match, one Arabic page may serve both.

Where the data comes from, and what each source can tell you

Search Console: group variants with regex

Google’s Performance report groups data by the query users typed and can be broken down by country. Since 2021 it has also offered custom regular-expression filters on queries and pages, using RE2 syntax (Google Search Central). That makes it practical to pull every spelling of a term into one view. A filter such as طفا(ية|يه) returns both endings of the extinguisher term; adding a country filter for Saudi Arabia and then Egypt shows whether the split differs by market.

Keyword Planner: read it as ranges

Google defines “Avg. monthly searches” as the average number of searches for a keyword and its close variants, and states that the figures are rounded (Google Ads Help). In Google Ads, close variants include misspellings, singular and plural forms, stemming and same-intent wording (Google Ads Help). In practice this means Keyword Planner may not separate ة from ه spellings cleanly, so do not use it to decide which spelling to write. Accounts with lower ad spend may also see broad buckets such as 1K–10K rather than averages, a restriction Google confirmed when it was introduced in 2016 (Search Engine Land). Report these numbers as ranges, never as precise volumes.

A five-step method for a B2B product or service line

1. Collect seed terms from sales. Export three to six months of inquiries per country, remove names and numbers, and list every way buyers named the product, the problem and the service. Note the language and spelling exactly as typed.
2. Expand each seed into its four layers. For every seed, write the English term, the Arabic term, transliterations, the common spelling variants and any local alternatives.
3. Check demand. Run the expanded list through Search Console (with regex grouping) and Keyword Planner for each target country. Mark each term: confirmed demand, low or no data, or not yet visible.
4. Test the results pages. Search each important variant in each target country and compare the top results. If two spellings return essentially the same pages, they are one target. If they return different types of result, for example one shows suppliers and the other shows definitions, they may reflect different intent.
5. Assign one target page per intent. Map the cluster, not the spelling, to a URL. Record the primary term, the variants to include in the text and the page’s Arabic and English counterparts.

The keyword sheet: one row per variant

Keep the research in one sheet. Each row is one variant of one seed term, so every spelling choice can be traced back to evidence. The filled row below is a hypothetical example showing the format only; it is not search data or a real inquiry:

Two columns stop most bad decisions. “Inquiry source” shows which words buyers really used. “Checked on” stops a six-month-old results-page test from being treated as current. The finished sheet is the input for the page work itself: titles, headings, FAQs and anchors are written from its confirmed terms.

Mapping variants without creating duplicate content

The common mistake (hypothetical example) is building “طفاية حريق”, “طفايه حريق” and “طفايات الحريق” as three thin pages. Google’s spam policies describe doorway abuse as pages created to rank for specific, similar search queries, and scaled content abuse as mass-producing pages mainly to manipulate rankings. Pages built only for spelling variants fit that description closely. Instead:

• Use the most common correct form in titles and headings, chosen from your Search Console and sales data.
• Include variants naturally where a reader would expect them: a short FAQ answer, a product table, a sentence that names both the Arabic and the English term, image alt text.
• Keep one page per intent and language. A Saudi page and an Egyptian page in Arabic are justified only when the content genuinely differs, such as different product ranges, approvals, delivery terms, contact details or, from the table above, consistently different buyer vocabulary (Google Search Central covers how same-language versions are handled).
• Connect the language and country versions correctly. Codes such as ar-SA and ar-EG, reciprocal hreflang and visible-language rules are set out in our guide to Arabic and English website structure.

Worked example: a food importer’s nuts and coffee range

Hypothetical example. A food importer sells roasted nuts and coffee beans to retailers, cafés and hotels in Saudi Arabia and Egypt. Suppose its anonymised inquiries show buyers writing مكسرات بالجملة and مكسرات بالجمله, the English “specialty coffee” alongside the Arabic قهوة مختصة and the transliteration سبيشالتي, and pack sizes typed with both Western and Arabic-Indic digits (1 كيلو, ١ كيلو). A Search Console filter such as مكسرات بالجمل(ة|ه) shows whether both spellings appear in both countries, and a separate filter on سبيشالتي|specialty|مختصة shows which form of the coffee term brings impressions in each market. Results-page tests then show whether the two nut spellings return the same supplier pages, and whether the coffee query returns suppliers or café listings. The terms and outcomes here are hypothetical; they are not search-volume data or real inquiries.

If the nut spellings behave as one search, the plan is one Arabic wholesale-nuts page and one English page, with مكسرات بالجملة in the heading and the variant left to the body. If the coffee terms return different kinds of result, the importer may need one page for trade buyers of beans and leave café-intent searches alone. Separate Saudi and Egyptian pages are added only if the range, pack sizes, delivery terms or the vocabulary table differ by market. Social and ad copy reuse the same confirmed vocabulary, and in paid search the same low-volume trade terms are handled as described in Google Search Ads for low-volume B2B queries.

Checklist before you publish

• Every target term traced to a source: Search Console, Keyword Planner range or anonymised sales inquiry
• Variants grouped into clusters, with one URL per intent and language
• Results pages tested per country for the main variants, with the date recorded
• Arabic pages fully written in Arabic, not translated menus around English text
• Digits column filled for every term that contains a number or model code
• AI-drafted Arabic checked against the sheet’s confirmed vocabulary, using the review in where AI falls short in Arabic marketing copy
• No near-identical pages built for spellings or cities
• Ad, social and page copy using the same confirmed vocabulary

Start with one product or service line, not the whole site: pull its last three months of Search Console queries and sales inquiries, group the variants and test the main ones in each country. Once that page is live, confirm it is indexed and canonicalised as planned, using the checks in our technical SEO audit checklist.