Questions clients ask

A corporate bilingual site is typically six to ten weeks from kickoff to launch. Roughly a third of that is content: Arabic copy written natively rather than translated is the single most common cause of a slipped launch date, so we agree the content plan in week one.

We design both directions at once. Arabic sets denser than Latin, needs more line height, and cannot be letter-spaced without breaking glyph joining. Layouts built with CSS logical properties mirror correctly with no separate stylesheet, which is why we build them that way from the first component.

Yes, and it is usually the right call. Rankings attach to URLs. We map every existing address to a new template rather than renaming anything, so no redirects are needed and no equity is lost. Where a redirect is genuinely unavoidable we issue a single permanent 301 with no chaining.

Because of what they emit. A builder wraps every element in several presentational containers, ships a large runtime, and stores content in a proprietary format that is difficult to migrate away from. Hand-written templates produce a fraction of the markup, load faster, and leave your content in plain post fields where any future developer can reach it.

Usually, yes. In most cases we replace the theme rather than the installation — the database, media library, users, URLs and admin all stay exactly as they are. That keeps the migration risk close to zero and preserves search equity.

We can. For Saudi government data, all data must remain in the Kingdom, and financial-sector data hosted offshore requires prior SAMA approval. Where either applies we select a CST-registered provider of the appropriate class and confirm the physical hosting location at city level, including the CDN, which is part of the subcontracting chain a buyer must verify.

More than mirroring. The primary action moves to the left of a button pair, progress indicators run right to left, icons with direction have to flip while logos and media controls must not, and numerals stay left-to-right inside right-to-left sentences. Arabic also sets denser, so line height and measure need adjusting rather than inheriting.

Five to eight per user type per language surfaces the large majority of severe usability problems. Beyond that the same findings repeat. We would rather run three small rounds across a project than one large round at the end.

It depends on catalogue complexity and how much you need to control. Salla and Zid are Saudi platforms with local payment and logistics built in, which shortens launch time considerably. Shopify is strong operationally but less flexible on bespoke bilingual layouts. WooCommerce gives full control at the cost of owning maintenance. We put the trade-off in writing before you commit.

Not universally. Under the E-Commerce Law, registration with an authenticated e-shop entity such as Maroof is an alternative to publishing your full address. In practice it is close to mandatory for stores taking online payment, because gateways and shipping providers ask for verification before enabling full features.

When a configured commercial product covers eighty per cent of the requirement and the remaining twenty per cent is not a competitive advantage. Custom software carries a permanent maintenance obligation. If the process is ordinary, buy the product and spend the difference on adoption.

You do. The repository, the source, the deployment scripts and the documentation transfer to you. We do not build on a proprietary framework you would have to license from us afterwards.

Sometimes, and the honest answer is that it varies by model and by dialect. Modern Standard Arabic is handled well by the leading models; Gulf dialect and mixed Arabic-English input are weaker, and retrieval over Arabic documents degrades if the source PDFs were scanned rather than born digital. We benchmark on your actual content and report the number before you commit budget.

If the assistant processes personal data of Saudi residents, the Personal Data Protection Law applies — including to processing that happens outside the Kingdom. That means a privacy notice covering the purpose, a lawful basis, honouring the five data-subject rights within thirty days, and a documented position on cross-border transfer. We design the data flow around that from the start rather than retrofitting it.

Almost always because configuration started before the process was agreed. The second most common cause is data: legacy records are dirtier than anyone expects, and reconciliation gets discovered late. We insist on a signed process model and a data-quality assessment before configuration, which is less comfortable at the start and much cheaper at the end.

Only if you need something the browser cannot give you: reliable offline use, push notifications, camera or sensor access, biometric login, or payment credentials stored on the device. Loyalty, ordering and account management often qualify. Marketing content does not — that is a website, and building it as an app usually means paying more for something fewer people install.

Yes. For government agency data, all data must remain within the Kingdom. That means a CST-registered cloud provider of the right class, physical hosting confirmed at city level rather than country level, and every subcontractor in the chain accounted for — including the CDN, which is the part most agencies forget. We also ask the contracting authority for a written data classification before selecting a host, because that classification determines what is permissible.

Technical fixes can move things within weeks. Content and authority take longer: three to five months for meaningful movement on mid-competition commercial terms, longer for the most contested ones. Any agency promising page one in thirty days is either targeting queries nobody searches or planning to do something that will eventually cost you the domain.

We earn mentions. Unlinked brand mentions in third-party text correlate more strongly with visibility in AI-generated answers than backlink counts do — in one analysis of 75,000 brands, branded web mentions correlated at 0.664 against 0.218 for backlinks. We pursue digital PR, review profiles and accurate directory data. We do not buy links; Google now lists inauthentic mentions as an enumerated spam category.

Yes, and not by translating. Arabic queries are shorter, more colloquial and dominated by local operators, while the English SERP for the same intent is full of directories and offshore agencies. The keyword sets barely overlap. Translated pages tend to rank for neither.

Less than the industry claims. Google's own published position is that optimising for its generative features is still SEO, and that no special schema or machine-readable file is required. Three things are genuinely different: AI crawlers other than Google's do not run JavaScript, retrieval is passage-level rather than page-level, and third-party sources take most citation slots. Everything else is SEO done properly.

No. Google states llms.txt has no effect on visibility, and a server-log analysis of 137,000 domains found 97% of published llms.txt files received zero requests — of the requests that did arrive, AI retrieval bots accounted for around one per cent. We will ship one because it costs an hour, but we will not bill it as a visibility service, and you should be sceptical of anyone who does.

Not on the best available evidence. The only quasi-causal study of the question — 1,885 pages that added JSON-LD, matched against controls — found no positive effect on ChatGPT or Google AI Mode citation and a small negative signal for AI Overviews. We implement schema thoroughly anyway, because it is cheap and it drives rich-result eligibility and entity disambiguation. We just will not tell you it is the reason you get cited.

No. A percentage fee rewards us for spending more of your money, which is a poor alignment of interest. We charge a flat retainer based on the work involved, so the recommendation to reduce spend costs us nothing to make.

In our experience yes, and noticeably — but the reason is not the language alone. Arabic-native copy is written around how the offer is actually discussed locally, whereas translated copy carries English sentence structure and English framing. We run both and let the data decide rather than assuming.

Then we do not A/B test. Below roughly ten thousand sessions a month on a template, a split test will not reach significance in a sensible timeframe and you would be acting on noise. We use qualitative research, usability testing and best-practice implementation instead, and revisit testing when volume supports it.

Because it is a second piece of original lettering, not a translation. Arabic letterforms join, sit on a different baseline, and carry their weight differently from Latin ones — setting the name in an Arabic font produces something that reads as a default rather than a mark. Custom Arabic lettering is drawn, tested at size, and balanced against the Latin lockup so neither looks like an afterthought.

It can. A CDN that edge-caches content in Frankfurt or Dublin has moved that data outside the Kingdom, and it forms part of the subcontracting chain a Saudi buyer is expected to verify. For workloads where residency applies we either restrict the CDN to in-Kingdom points of presence or omit it and solve performance at the origin.

On a care plan: we take the site offline behind a maintenance page, restore from the most recent clean backup, identify the entry point, patch it, and give you a written incident summary. The reason care plans exist is that this sequence takes hours when backups are tested and access is documented, and weeks when they are not.

Almost always some combination of consent-denied sessions, cross-device journeys, spam traffic and different definitions of what a conversion is. The fix is not to pick whichever number is higher — it is to define the conversion once, capture it server-side where possible, and reconcile the two systems until the gap is understood and stable.