SEO · Arabic and English · Gulf
SEO Agency for the Gulf, in Arabic and English
We will not promise you a ranking. We will show you the method, put every technical fix in writing, and report only numbers you can open yourself.
Two languages
one technical base
Arabic and English share the crawl and the hreflang set, then split for research and writing
6–12 mo
for contested terms
Technical faults take weeks. Competitive rankings do not, and faster promises are guesses
Zero
ranking guarantees
We commit to the work, the method and the reporting. Nobody controls the result
Your accounts
every figure we quote
Search Console, Analytics and PageSpeed stay in your name. Re-run any number yourself
What this page commits to
Key takeaways
- Gulf SEO is a bilingual problem before it is a keyword problem. Most sites we audit lose in Arabic for technical reasons.
- Arabic is the cheaper half of the opportunity: the supply of genuinely written Arabic pages is thin, the demand is not.
- Technical work pays out early and once. Content pays out slowly and repeatedly. The wrong order wastes three months.
- We publish no client results, traffic figures or ranking claims. Where a competitor shows a number, we show the method.
- Every figure we report comes from an account you own. If you cannot verify it, we do not report it.
- Filweb is a studio in Istanbul, with no Gulf office, address or landline.
What is on this page
A bilingual problem, not a keyword problem
Almost every Gulf business we look at competes in two search markets and treats them as one. The English site gets the budget and the writer. The Arabic version was produced at the end of the project, translated in bulk, given the same titles, and wired up with hreflang tags that do not validate. One market is being contested. The other is on the table.
That is structural, not editorial. Search engines treat your Arabic pages as separate pages that must earn their own relevance, so they need their own research, headings and links. Mirrored pages compete with nothing.
So the work divides. A shared technical foundation works for both languages or fails for both, and above it everything is done twice. We start at the foundation, because invisible content is the most expensive thing an agency can sell you. If the site itself is the limit rather than its visibility, that is our web design or UI and UX work instead.
What we will and will not promise
We do not guarantee rankings. Not first position, not top three, not a count of keywords by a date. The answer does not change with the size of the budget.
A ranking comes from a system we do not control, judged against competitors who are also working, on a results page that changes without notice. An agency guaranteeing a position is either promising something worthless, such as a phrase nobody searches, or describing an outcome it cannot cause. What we put in writing instead:
- The technical faults we found, by URL, ranked by what each costs you.
- The pages we will produce, in which language, the queries each answers, and when.
- The exact reports, filters and date ranges, named in advance so the measurement cannot move.
- What we will not do: bought links, spun translations, invented local addresses.
- The point at which we would tell you to stop, agreed in month one.
The honest failure mode in SEO is not that nothing happens. It is that something happens slowly enough that nobody wants to be the one who calls it. For a shorter feedback loop, Google Ads is the better instrument.
Month by month, for six months
What happens, what you should see, and what would tell you it is not working. The last column is the one agencies leave out.
| Month | What we do | What you should expect to see | What would tell you it is not working |
|---|---|---|---|
| Month 1 | Crawl both languages. Audit indexation, hreflang, canonicals, speed. Research keywords. Record the baseline. | A fault list by URL, a developer specification, a baseline you could re-run today. | A score out of 100 instead of a fault list. Nothing you can open without us. |
| Month 2 | Ship or hand over the fixes. Correct templates, titles, internal links. Google Business Profile data. | Pages leaving the not-indexed reports, crawl errors falling, fixes visible in the page source. | Fixes reported as done that are not in the live source. No list of who blocks what. |
| Month 3 | First pages live in both languages, written rather than translated. Structured data added. | New pages indexed in days, and impressions on queries you had no page for. | Pages unindexed after three weeks, or Arabic that reads as translated. |
| Month 4 | Expand around whatever showed demand. Resolve competing pages. Begin earned coverage. | A wider set of queries in the Performance report, and movement on the tracked list. | Publishing continuing on a schedule that ignores what month three showed. |
| Month 5 | Conversion work: the pages receiving visits, the enquiry routes, the tracking that proves one. | Organic enquiries in your own analytics, reconciled against your form log or inbox. | Wider visibility with no change in enquiries, and nobody asking why. |
| Month 6 | Re-run the month one baseline identically. Decide what continues and what stops. | A written comparison against the recorded baseline, and a recommendation that includes stopping. | An automatic renewal, a fresh set of metrics, or a baseline quietly redefined. |
The honest checkpoint is month three: by then the technical work should be verifiable and the first content indexed.
How the two kinds of work behave
An argument for sequencing, not a projection. No numbers on either axis, deliberately.
Read it as order of operations. Technical fixes are largely one-off: a blocked page becomes crawlable, and does not become crawlable again next month. Content behaves the opposite way, since each page widens the queries you can appear for. Front-loading content while indexation is broken wastes six months.
Why Arabic SEO is the cheapest opportunity in the Gulf
The section we would keep if we had to delete the rest.
In most Gulf categories the Arabic results page is less contested than the English one. Not because Arabic demand is small or the queries are less commercial, but because the supply of pages that genuinely answer an Arabic query is thin. Confirm it yourself in an afternoon: search your main service in Arabic, open the first ten results, and count how many are original Arabic pages written for that query. Then do the same in English. The difference is your opportunity, in your category, rather than a statistic we could quote at you.
Why the Arabic side stays thin
Budget order comes first. The English site is built first because the people commissioning it work in English, so Arabic arrives when the money is spent and the deadline is close, which is when bulk translation gets approved.
Second, a translated page is structurally wrong even when the prose is acceptable. It answers the English question in Arabic, carrying the English title word for word. The Arabic query worth having often uses a different term, because the colloquial name for a service is rarely the literal translation of the English one.
Third, the technical layer. Arabic sections sit in the least tested part of the site. We routinely find Arabic pages whose canonical points at the English page, or an hreflang set missing its self-reference. The pages exist, someone paid for them, and they cannot rank. Nobody notices, because the reports being read are English.
Fourth, asymmetry. International competitors with large content teams frequently have no Arabic content at all, so you compete against absence rather than strong pages.
What cheapest does and does not mean
Cheapest means cheapest per unit of visibility gained, not cheap in absolute terms. Genuine Arabic costs more per page than translation, because you pay a writer who reads the results page rather than a tool that substitutes words. The arithmetic still favours it, because the bar is set by the pages already there, and in Arabic that bar is lower.
It does not mean Arabic is always the answer. An expatriate procurement manager in Dubai may search in English whatever their first language is, and in some technical categories the Arabic queries that exist bring readers rather than enquiries.
What it means for the budget split
There is no correct ratio, and an even split is usually wrong. The split follows the buyer, so these are recommendations with reasoning attached, not measurements:
- Consumer services in one city, or a Saudi retail brand: Arabic majority, since residents search in Arabic against thin listings.
- Government and semi-government buyers: Arabic for discovery, English for the documents.
- Multinationals and Dubai procurement teams: English majority, Arabic for reach, since no Arabic version reads as no local presence.
- Hospitality and inbound tourism: English and source-market languages first, Arabic for domestic demand.
- Expatriate-facing professional services: English first, Arabic on regulated pages.
The technical foundation is never split: both languages share one crawl, one hreflang set and one speed budget, which is why we quote technical work once and content per language. To size the gap from your own data, filter the Search Console Performance report to your Arabic URL folder and compare it with the English one. Impressions on Arabic queries for pages never written to answer them means demand exists and you are the accidental answer. An Arabic folder with almost no impressions is a technical fault, and the checks below say which. The same thinking drives our Salla and Zid pages, and a multilingual platform where Arabic came first rather than last.
Local search, and when it beats everything else
For a business with a physical address in Dubai, Riyadh or Doha, the map results sitting above the normal listings are often worth more than page one itself. Someone searching for a clinic, a showroom or an office is close to deciding, and the three businesses shown on the map take most of that traffic. This is also the one part of search where a small local business can outrank a large national competitor, because proximity and profile completeness count for more than domain strength.
The work is unglamorous and mostly sits outside your website. A Google Business Profile has to be claimed, verified, categorised correctly and kept accurate, and the same name, address and phone number has to appear identically everywhere else it is listed. Inconsistency is the usual problem we find: an old branch address on a directory, a landline that changed two years ago, or two competing profiles for the same office created by different staff.
- Claim and verify the profile. An unverified or unclaimed profile can be edited by the public. Verification is also what unlocks the categories, service areas and messaging settings that decide when you appear.
- Choose the primary category deliberately. It does more to determine which searches you show up for than anything you write, and the obvious choice is often not the one your competitors rank under.
- Make the name, address and phone number identical everywhere. Directories, your own site's footer, your schema markup and your profile must agree exactly, including how the building and area are written.
- Publish the Arabic version of the profile properly. A profile in English only is invisible to a large share of the people searching for you in Saudi Arabia and Kuwait.
- Answer reviews, do not buy them. Purchased reviews get filtered and can cost the profile its standing. We will help you ask customers properly; we will not fabricate them.
- Keep a location page per branch on your own site, with the address, opening hours and a map, so the profile has something consistent to point at.
Local work suits you if customers come to you, or if you serve a defined area from an address. It does very little for a business selling remotely across the whole Gulf from one office, which is why it is not on every plan we write. If your enquiries mostly start with someone searching a city name, say so early and we will weight the first two months towards this rather than towards content. The Dubai, Riyadh and Jeddah pages cover what changes city by city.
The technical work nobody shows you
Technical SEO has a presentation problem. It is the work most likely to change your results and the least likely to look like anything in a monthly summary, because a correct canonical tag looks like nothing at all. Hence the habit of reporting activity instead: pages audited, issues found, tools run. We deal with that by making every item verifiable by you: each fix names the URL, the fault, the change and how to confirm it.
Two faults deserve naming, because we find them constantly on bilingual Gulf sites. An hreflang set must be reciprocal and complete, every page listing every page including itself, with one nominated as the fallback through x-default; the self-reference is the piece most often missing, and its absence is not a partial failure, because the whole set is then ignored. Google states this in its guidance on localised versions. The second is a canonical tag copied across a template, so every Arabic page declares the English page as canonical: an explicit instruction to drop the Arabic version.
Speed sits alongside both, measured from real visitors against the thresholds in Google’s Core Web Vitals documentation. Structured data decides which result formats you are eligible for, with the vocabularies public at schema.org.
Technical checks, and how to verify them
What we check, what each fault breaks, and how you confirm the fix yourself. Nothing in the third column needs a tool we own.
| Check | What it breaks when wrong | How you verify the fix yourself |
|---|---|---|
| hreflang set | The wrong language is served, or the two versions read as duplicates and one is dropped | View source on both URLs: each must list every language, itself included, plus one x-default |
| Canonical tags | The wrong URL is indexed, or a whole Arabic section is excluded | URL Inspection shows the declared and the selected canonical. On the Arabic URL, both should be the Arabic URL |
| Indexation: robots, noindex, sitemaps | The page cannot rank at all, whatever is written on it | Open /robots.txt, then the Pages report for excluded URLs and the stated reason |
| Core Web Vitals | Mobile visitors on cellular data leave before the page is usable | PageSpeed Insights field data for that URL, plus the Core Web Vitals report. Allow 28 days |
| Structured data | You lose eligibility for the result formats that take up space on the page | The Rich Results Test on the live URL, then the enhancement reports once recrawled |
| Redirect map after a rebuild | Old addresses return errors and what those pages earned is lost | Sample the old-URL list: one permanent redirect each, not a chain or the homepage |
| Language and direction attributes | Arabic renders in the wrong direction and fonts, and the Arabic title is inherited from English | The Arabic page source should open with lang="ar" and dir="rtl", and carry a title written in Arabic |
You get this as a priority-ordered list with URLs, so a developer can act without a meeting, or we implement it on a staging copy. If the platform itself is the obstacle, we say so, which leads to e-commerce development rather than more invoices.
What correct hreflang looks like
One English page and its Arabic counterpart. The pieces most often missing are the self-reference and x-default, and either omission invalidates the set.
Test your own site in a minute. Open your Arabic page, view the source, and look for a declaration pointing at the Arabic page itself. If it is missing, the set is incomplete and nobody checked it.
Translate, or write fresh?
This decision sets the cost of the Arabic half of the work, so we are blunt about it. Translate the pages that are records: legal text, policies, specifications, product attributes. They need to be accurate and identical in both languages, and nobody arrives at a privacy policy from a search result.
Write fresh Arabic for anything meant to be found. The Arabic page has to answer an Arabic query, and that query is often not the translation of the English one: the colloquial name of a service differs from the formal one, and brand names get transliterated several ways. A translated page inherits the English target and misses all of it.
There is a middle path, and it costs less than it sounds: keep the English page as the source of facts rather than of sentences, restructured around the Arabic query. You pay for the research once. What we advise against is machine translation with a light edit, which native readers spot and which carries the English targeting into the Arabic page. Fresh Arabic cannot be scheduled like translation, because research does not compress, which is why first content lands in month three.
What we report, and where it comes from
Every line names its source. The test each one must pass: could you open this yourself, in an account you own, and get the same number?
| What we report | Where the number comes from | How you check it yourself |
|---|---|---|
| Indexed pages by language | Search Console, Pages report, filtered by URL folder | Open the same report in your property and apply the same filter |
| Impressions and clicks by query | Search Console Performance, with the date range stated | Reproduce it with the filters we name. The data settles over a few days |
| Position on the tracked query list | Search Console Performance, on the list agreed in month one | Same report, same list. It is an average, not a position you occupy |
| Arabic against English visibility | Two folder-filtered views of the same report | Compare the views yourself. The gap argues for where the next page goes |
| Core Web Vitals | Search Console Core Web Vitals, and PageSpeed Insights field data | Enter your own URL into PageSpeed Insights: 28 days of real visitors |
| Enquiries from organic search | Your analytics, reconciled against your form log or inbox | Both sources are yours. Where they disagree we say so rather than pick the higher |
| Technical faults open and closed | Our ticket list, plus the live page source | Open the URL and view the source. Anything open names who it waits on |
Three things are deliberately absent. A traffic figure from another client, because client results are not ours to publish. A proprietary visibility score, because a score we compute cannot be checked by you. And a rank tracker as the headline, since rankings vary by device and location.
When not to hire us
Cases where we say no
- You need enquiries this month. Search cannot be brought forward by paying more. Paid search can.
- Your developers cannot ship technical changes and you cannot give us access. Nothing else we do will matter.
- You will not give read access to Search Console and Analytics. Then nobody can verify anything, including you.
- You have a five-page brochure site with nothing to optimise. The budget belongs in the site first.
- Another agency has guaranteed you a position. We cannot match a promise we do not believe.
“We will not sell you a position. We will sell you the work that makes one possible, and a way to check that the work happened.”
Related reading
Where to go next, depending on what the constraint actually is:
- Web design and UI and UX design — when the site is the limit, not its visibility
- Google Ads — for demand you need this quarter, with a feedback loop in days
- Digital marketing — how search fits alongside social and paid media
- E-commerce development, Salla and Zid — search work specific to stores
- City views: Dubai, Abu Dhabi, Riyadh, Jeddah, Doha, Kuwait City
- Costs in Dubai, Saudi Arabia, Qatar, Kuwait, and our published prices
- A Dubai case study, the blog, or contact us
One last point. Filweb is a studio in Istanbul working Gulf hours. There is no Gulf office, local address or landline, and we will not invent one. Every account stays in your name.
SEO questions
How long does SEO take?
Technical faults are fixable in weeks. Competitive commercial terms in Dubai or Riyadh take six to twelve months. Faster promises on a contested term are guesses.
Can you guarantee first position?
No, and we would not trust an agency that does. Rankings come from a system none of us control. We commit in writing to the work, the method and the reporting.
What does Arabic SEO actually involve?
Modern Standard Arabic with Gulf vocabulary, research carried out in Arabic rather than translated from an English list, and pages structured around the Arabic query.
Do we need a new website first?
Usually not. Most sites can be fixed rather than replaced. We recommend a rebuild only when the platform caps what is achievable, and we name the limit we hit.
Can you do Arabic and English at the same time?
Yes, the normal case. One technical foundation, one crawl, one correct hreflang set. Above that they separate: their own research, writing and internal linking.
How do you build links?
Earned coverage, content worth citing, and genuine local citations such as UAE directories and chambers of commerce. We do not buy link packages: the risk is real and the gains short-lived.
Can you work with our existing developers?
Yes, often the best arrangement. We write a prioritised specification, review the work before it ships, and confirm the fix on the live page. Your team keeps the code.
What if our developers will not implement the technical fixes?
It happens, usually a frozen roadmap or an outside agency guarding the codebase. We cut the list to the items that genuinely block indexation, so the ask is small and specific. We state the cost of skipping each one, so the decision sits with whoever owns the roadmap rather than with us. Then we offer to make the changes ourselves on a staging copy for their review, which removes the workload objection. If all three fail we say the engagement cannot work, because we will not publish onto a foundation we know is broken.
Should we translate our English content into Arabic, or write fresh Arabic?
Translate the pages that are records: legal text, policies, specifications, product attributes. Write fresh Arabic for anything meant to be found, because the Arabic query is often not the translation of the English one. The middle path we use most keeps the English page as the source of facts rather than sentences, restructured around the Arabic query. We advise against machine translation with a light edit: native readers spot it, and it carries English targeting into the Arabic page.
Do you have an office in Dubai or the Gulf?
No. We are a studio in Istanbul working Gulf hours and will not pretend otherwise. Your domain, hosting, Search Console and Analytics stay in your name.
What do you need from us to start?
Read access to Search Console and Analytics, enough platform access to see how pages are generated, and a conversation about who your buyer is and which language they use.
Can we verify your reporting ourselves?
That is the design. Every figure comes from an account you own, and the report names the source, filter and date range. If a number cannot be checked without us, we do not report it.
Should we do SEO or Google Ads first?
Ads if you need enquiries this quarter, because search cannot be bought forward. Search if you want something that keeps working when you stop spending. Your cash position decides.
Is there a minimum commitment?
Realistically six months: the first month produces documents, and technical work must land before content can matter. The proposal states what would make us recommend stopping.
Want to know what is holding you back?
Send us the address. We will tell you what we find, in writing, including the case where the answer is that SEO is not your problem.