Last updated: August 2026
A second language looks like a project. It behaves like a subscription.
The build is the small part. What happens afterwards is that every page, every offer and every price now exists twice. Then three times. Then four. Your English site keeps moving, because that is where the business actually runs. The other versions do not, because nobody owns them. Eighteen months later a customer lands on your Russian pricing page and reads a number you stopped charging last year.
That is the honest case for publishing fewer languages properly rather than many badly, and it is where this guide starts.
The second thing worth your attention is hreflang, the most commonly broken piece of technical setup on multilingual sites. Google's rule is that the annotations have to point both ways, and when they do not, the whole set gets ignored [2].
Arabic is its own subject and we have written about it three times already. This guide sits above Arabic. It is about how many languages, which ones, how to structure them, and what it costs to keep them honest.
The cost that shows up after launch
Think about what a second language does to your operating model rather than to your build budget.
Every new service page is now two pages. Every price change is two edits. Every campaign landing page, every seasonal offer, every policy update, every new post. Add a third language and the multiplier is three. None of it is difficult. It is just relentless, and it competes with everything else your team has to do that week.
Here is where an annoyance turns into a search problem. Google's hreflang rule requires every version of a page to point at every other version, and if two pages do not both point to each other the tags are ignored [2]. So a new English page with no Russian counterpart is not part of your multilingual set at all. Google falls back on its own judgement about which version to show which searcher [2]. The set you paid to build no longer covers your newest and most commercial content.
The commercial damage lands first, though. A stale price is a stale promise. A visitor who finds one does not think the page was not updated. They think you quoted one number and charge another.
Real talk: the question is not whether you can afford to translate your site. It is whether you can afford to keep translating it every week for the next five years. If the answer is no for four languages but yes for two, publish two.
What the UAE government actually publishes about languages
Very little, and that is useful to know before somebody sells you a statistic.
Arabic is the official language of the UAE [8]. The government's own fact sheet also states that five other languages have official status in the courts: Chinese, English, French, Hindi and Russian [8]. Separately, the same fact sheet lists other widely spoken languages as Bengali, Farsi, Malayalam, Turkish and Urdu [8]. Road signs and commercial outlet signage appear in Arabic and English [8].
Read that for what it is. Court status is a legal recognition. The widely spoken list describes prevalence. Neither measures your customers, and the fact sheet attaches no percentage, share or ranking to any language on either list [8].
We are not going to quote you a population figure or a language share, and it is worth saying why. Figures of that kind circulate constantly, and almost all trace back to the federal statistics authority through several layers of retelling. We could not reach that authority's site directly to verify anything while researching this piece [9]. So we are not publishing numbers we could not confirm at source, and you should treat any language usage percentage you see elsewhere the same way.
The two official lists are good for one thing: a shortlist. Guessing from them is more defensible than guessing from a marketing blog.
The only language data that matters is yours
So stop guessing.
Your analytics already records the browser language of every visitor. Your CRM knows which enquiries close. Your sales team knows which calls get harder because of language. Your inbox and your WhatsApp know which questions arrive in which one.
Ask three questions of that data:
- Where do enquiries that already close come from? Not traffic. Revenue.
- Where do enquiries stall? A language barrier usually appears as a healthy enquiry rate with a poor close rate from one segment.
- What would the second language actually change? If your buyer is a procurement manager who works in English all day, another version of your site is a courtesy, not a conversion lever.
A language earns its place when it maps to people who transact with you. Not people who pass through the country, and not people inside an unsourceable statistic.
Where the language versions live on your domain
Google names three URL patterns and publishes the trade offs for each [1].
- A country code domain. Google calls this clear geotargeting, with server location irrelevant, but notes it is expensive, can have limited availability, and sometimes carries strict requirements [1].
- A subdomain, such as
ru.example.com. Easy to set up and allows different server locations, with the trade off that users might not recognise the geotargeting from the URL alone [1]. - A subdirectory, such as
example.com/ru/. Easy to set up and low maintenance because it is on the same host, with the trade off of a single server location and harder separation between sites [1].
Google does not recommend URL parameters for this kind of segmentation [1].
For most UAE businesses already running one domain, the subdirectory is the sensible default. It is the lowest maintenance option Google names, it sits on the domain you already have, and it does not create new infrastructure for somebody to forget about. A subdomain earns its place when a separate team or content system genuinely owns that language version. A dedicated country domain is the heavy option, and Google itself flags the cost [1].
One more line from the same page matters more than it looks. Google states it does not vary the crawler it uses to discover locale variations of a single site, and that you must explicitly tell Google about any locale or language variation [1].
hreflang, and why it is broken on so many sites
hreflang tells Google that several pages are localised variations of the same content, so the right one can be served to the right searcher [2]. It does not tell Google what language a page is written in. Google determines that algorithmically from the content itself [2].
Three rules matter more than the rest.
It has to be reciprocal. Google's own wording is that if two pages do not both point to each other, the tags will be ignored, so that someone on another site cannot arbitrarily name their page as an alternative version of yours [2]. Every version of a page must reference every other version, including itself.
x-default is the fallback. It is used when no other language or region version matches the visitor's browser setting [2]. Point it at a language selector or at your default version, which for most UAE businesses is English.
Placement is your choice, but pick one. Google supports three equivalent methods: link elements in the page head, HTTP Link headers, useful for non HTML files such as PDFs, and entries inside an XML sitemap [2]. Mixing methods is how sets end up half declared and nobody notices.
Google also names the common mistakes directly: missing return links, incorrect language codes, and invalid region codes such as EU, UN or UK, where the correct code for the United Kingdom is GB [2]. The values follow BCP 47, the standard behind language tags in HTML and XML, which is what makes short codes and region variants such as en-AE valid. W3C also recommends declaring the language on the html element of every page [7].
The failure we see most often is one directional. The new Russian page correctly points at the English page. The English page, built two years earlier and rarely revisited, never gets updated to point back. Because the set is invalid the moment one direction is missing, that single omission takes out the whole cluster rather than one pair.
Google's stated behaviour when annotations are broken is to ignore them, not to penalise the site [2]. That sounds mild. The practical result is the same as having no hreflang at all, and Google's own guidance says it is usually best to explicitly indicate your language or region specific pages [2].
So build the check into your release process. If a page ships in one language, either its counterparts ship with it or somebody records that the set is incomplete. Nobody finds a broken hreflang set by accident.
Do not guess the visitor's language for them
This is the most common well intentioned mistake on a multilingual site.
Google's guidance is explicit: avoid automatically redirecting users from one language version of a site to a different language version [1]. Google also states that IP location analysis is difficult and generally not reliable, and that it may not be able to crawl variations of your site properly when content is served that way [1].
The search consequence is worse than the user consequence. Googlebot's crawl requests are not reliably localised, and much of its crawling comes from one region, so a site that redirects on perceived location can end up showing Google only one version of itself [1][3]. Your other languages then do not merely rank badly. They may never be discovered.
Google's recommended alternative is plain: add hyperlinks to other language versions of a page, so users can click to choose a different language version [1]. A visible, persistent switcher. Not a pop-up that guesses.
One related detail catches teams out during phased rollouts. Robots directives have to be applied consistently across every locale variant of a page [3]. Blocking one version and not another creates crawl problems specific to multilingual sites, and a temporary block left in place after launch is an easy way to get there.
Translation quality, and where Google draws the line
Duplicate content is not automatically a problem. Google states that some duplicate content on a site is normal and is not a violation of its spam policies [4].
The multilingual rule is specific and worth memorising: different language versions of a single page are considered duplicates only if the primary content is in the same language [4]. A genuinely translated page is therefore not a duplicate of its source. A page where only the navigation and footer were translated, while the body stayed in English, still is. That is why the translate-the-template shortcut fails. Those pages never earn credit as distinct localised pages.
On machine translation, Google's spam policies name the practice directly. Under scaled content abuse, they list generating many pages through automated transformations, including synonymising and translating, as content produced at scale without adding value for users [5]. Note where the line sits. It is at scale and without review, not at the existence of machine translation. A machine first draft that a competent human then edits is a different practice from publishing thousands of unreviewed pages.
Nor should you assume Google's own translation covers you. Google can translate a result's title and snippet and let a user open a translated version of a page, but describes this as no different from using Google Translate or in browser translation, and states that it does not host translated pages [6]. You can opt out per page with a notranslate tag or header [6]. It is a convenience layer sitting on your unchanged page, not a localised page of your own.
What this costs
No government body, standards organisation or statistics authority publishes translation or localisation rates. This is a commercial rate, not a regulated one, so the figures below are ours and not a market survey [10].
Adding a second language to an existing site starts from around AED 2,500, where the pages already exist and only the content and the technical wiring are new [10]. Final pricing depends on scope. The two things that move it most are page count and whether your current setup can hold multiple locales at all.
For context against our other published figures:
- A focused business website build starts from around AED 3,500. Our website cost guide breaks that down by site type [10].
- Arabic costs more than the languages discussed here, because right-to-left work is layout engineering rather than translation. Our RTL design guide and our Arabic search guide cover that ground, and Urdu carries a similar right-to-left cost profile [10].
- Video does not translate itself. Subtitles start from around AED 300, a professional voiceover track from around AED 550, and a full localisation with rebuilt on-screen text from around AED 1,500. See our voiceover and subtitling guide [10].
- If a language rollout forces a URL restructure, that is a migration, starting from around AED 5,000 [10]. Our migration guide explains what has to be preserved.
Now the part that actually decides this. Compare the one off build against the recurring cost of keeping the second language alive. Content production is the honest benchmark: a single article starts from around AED 500, and an ongoing content retainer starts from around AED 2,000 per month [10]. Our content marketing cost guide sets out the bands.
Run that forward on a site that publishes regularly. Run that forward on a site that publishes regularly and the recurring cost overtakes the one off build faster than most teams expect. Do the sum with your own publishing rate before you agree to a fourth language, because that is the figure to show whoever is asking for it. That is the figure to show whoever is asking for a fourth language.
Real client stories
Situations from our own work. Details changed for privacy.
A Dubai services company with a Russian site nobody owned. The Russian version was built as part of a launch push and handed to nobody afterwards. Two years later it still carried the previous fee structure and a decommissioned phone number. A prospect quoted the old fee back at them during a negotiation and did not accept the correction. That conversation cost more than the translation ever had. We cut the Russian version back to a small, deliberately maintained set of pages with one named owner.
A free zone e-commerce brand with a one way hreflang set. Four languages, all correctly pointing back at English, and the English pages pointing at none of them. Because the annotations have to be reciprocal, the entire set was being ignored [2]. The team had assumed the tags worked because they could see them in the page source. Fixing the return links took an afternoon. Nobody had checked for eighteen months, because nothing on the site looked broken.
A UAE clinic redirecting visitors by IP. Visitors were sent automatically to a language version based on detected location, with no way back except editing the URL by hand. Residents whose location did not match their language were stuck, and the other versions were barely indexed, which is the crawl outcome Google's guidance warns about [1][3]. We replaced the redirect with a persistent switcher, the alternative Google recommends [1].
How SKIMBOX approaches multilingual sites
We usually start by talking you out of languages.
Not all of them. Just the ones you will not maintain. The first conversation is about your own analytics and sales records rather than a statistic, and it tends to end with a shorter list than the one we were handed.
After that the work is ordinary and specific. Choose a URL structure and commit to it [1]. Build hreflang as a reciprocal set, with x-default, in one placement method, and put a check in the release process so it stays that way [2]. Replace any automatic redirect with a visible switcher [1]. Make sure the body content is genuinely translated, not just the template [4]. Then name an owner for each language version, because a version with no owner goes stale.
If search visibility is the point, read our Google UAE ranking guide and our SEO cost guide before you commit budget, because a second language multiplies the search work as well as the content work.
Adding a second language to an existing site starts from around AED 2,500, and a focused business site build starts from around AED 3,500. These are our own figures, not a market survey. Final pricing depends on scope.
If you want this looked at properly, see our web development services or our content marketing services, or contact us.
References
[1] Google Search Central - Managing multi-regional and multilingual sites, covering the country code domain, subdomain and subdirectory trade offs, the locale signals Google uses, the guidance against automatic redirection, the unreliability of IP location analysis, and the recommendation to add hyperlinks to other language versions. developers.google.com [2] Google Search Central - Tell Google about localized versions of your page, covering what hreflang does, the reciprocity requirement, x-default, the three placement methods, and the common mistakes Google names. developers.google.com [3] Google Search Central - How Google crawls locale-adaptive pages, covering the crawl consequences of serving content by perceived locale and the requirement for consistent robots directives across locale variants. developers.google.com [4] Google Search Central - Consolidate duplicate URLs, covering duplicate content not being a spam violation and the rule that different language versions are duplicates only if the primary content is in the same language. developers.google.com [5] Google Search Central - Spam policies for Google Web Search, covering the scaled content abuse policy and its naming of automated transformations including translating. developers.google.com [6] Google Search Central - Translated results in Google Search, covering how Google translates result titles and snippets, its statement that it does not host translated pages, and the notranslate opt out. developers.google.com [7] W3C - Language tags in HTML and XML, covering the BCP 47 standard behind hreflang and lang attribute values and the recommendation to declare language on the html element. w3.org [8] The UAE Government Portal - Fact sheet, covering Arabic as the official language, the five languages with official status in the courts, the separately listed widely spoken languages, and Arabic and English signage. u.ae [9] Federal Competitiveness and Statistics Centre - the UAE's federal statistics authority, which we could not access directly while researching this article, and the reason no population or language usage figure appears here. fcsc.gov.ae [10] SKIMBOX - our own published pricing for website builds, Arabic subtitling and voiceover, video localisation, website migration and content production, and our own commercial figure for adding a second language to an existing site, 2026. skimbox.co
