Translated website not ranking in Google? Eight checks for Arabic and English sites in the Gulf

A translated website usually fails to rank for one of eight reasons: Google cannot reach the translated pages, has not indexed them, sees only a translated menu, gets a broken hreflang pair or a canonical to the other language, is redirected away, finds mixed languages on one page, or the words are ones nobody types. Check them in this order.

This guide is for owners of Saudi and Gulf stores, clinics and travel companies whose Arabic and English versions do not both bring customers. On 30 October 2026 we read Google’s pages on localized versions, multilingual sites, canonicals and spam, 17 guides that Google showed on page 1 in Riyadh, and the language setup of 47 Saudi websites that appeared on page 1 in our earlier Riyadh audits. We sell Arabic SEO, so we have a reason to say “get help”. Every check below costs nothing, and a developer can run most of them in an afternoon.

Free 5-point SEO audit for your website. Send us your website address. Within 48 hours a senior specialist sends you 5 prioritized fixes with the evidence, including a look at whether your Arabic and English versions are set up the way Google expects. Get your free SEO audit.

The short answer: eight checks, in this order

Check	How to test it	A failure looks like
1. One address per language	Open the Arabic page in a private window, switch language, watch the address bar	The address does not change, or changes only through a cookie or a ?lang= link
2. Indexed	URL Inspection in Search Console on the translated address	“URL is unknown to Google”, or Google chose another page as canonical
3. Really translated	Read the title, headline and body of the translated page, not only the menu	English menu, Arabic page
4. hreflang pair	View source on both pages: each lists itself and its twin	Only one page names the other, or the address redirects
5. Canonical	Find rel=“canonical” on each language page	The English page names the Arabic page as canonical
6. No forced redirect	Request the page with and without a language header	Visitors or crawlers are sent to one language
7. One language per page	Read the text of one page	Two languages side by side
8. Words people type	Compare autocomplete and Search Console queries in each language	A literal translation of a phrase nobody searches

The first three checks decide whether Google has a translated page to rank at all. Numbers 4 to 6 are the plumbing between the languages, and the last two decide whether the page deserves to rank. Stop at the first failure and fix it before you read on, because a broken pair makes the later checks hard to judge.

Check 1: does each language have its own address?

Google recommends a different address for each language version of a page, rather than cookies or browser settings that change the language on the same address. Its reason is practical: Googlebot usually crawls from the USA and sends its requests without an Accept-Language header, so a site that reroutes visitors by language settings may never show Google its other version. A subdirectory such as /ar/ is the simplest choice; Google does not recommend a parameter such as ?lang=en.

One clinic site in our audit shows the failure. Its hreflang names both ar-SA and en, but both tags carry the same address. The English link on the page is a ?lang=en address that answers 302 to the home page, so English exists only as a cookie state. There is nothing for Google to index in English, whatever the tags say.

Test it: in a private window, change the language and reload. The address bar must change and stay changed. Then check that the switch is a real link: Google finds links only when they are an <a> element with an href, so a button that runs a script is invisible to it.

Check 2: is the translated page indexed?

Inspect the translated address in Search Console. A page that is “unknown to Google” has not been found; a page Google calls “Duplicate without user-selected canonical” was judged a copy of another page, and Google will not serve it in search. Search Console’s own page indexing help adds that alternate language pages are not detected by it, so the report cannot tell you the pair is working. Inspect the pages one by one, and compare the Arabic and English results for the same page.

A new language folder is crawled over weeks. Do not change the setup every few days while you wait: our guide to how long SEO takes explains the four clocks (found, seen, chosen, paid) and what to judge in the first month.

Check 3: is the page translated, or only the menu?

Google says localized versions are only treated as duplicates when the main content of the page remains untranslated. It names the case of translating only the template, such as the navigation and footer, as one where bad results appear: the same content shows up several times in search with different menu languages.

Two sites in our audit match that description. On one clinic site, hreflang lists /en/ as English, but its title, headline, description and body are Arabic, and only the menu is English. On a cleaning company’s site, the /en/ page has an Arabic title, and its canonical names the Arabic home page, with no hreflang at all. In Google’s terms both are Arabic pages on a second address.

We found the same on page 1 of the Arabic search. Of the 8 results we could read for the Arabic search, two have an Arabic-language address (/ar/ and /ar-sa/) and returned an English title, headline and text to our request. A reader who clicks one expects Arabic.

Test it: read the title, the headline and the first paragraph of the translated page. If any of them is in the source language, the page is not translated yet.

Check 4: does the hreflang pair work?

Google’s rules are short. Each language version must list itself and every other version. The addresses must be full, with https and the host. If two pages do not point to each other, Google ignores the tags. Codes start with the language (ar, en), optionally followed by a region (ar-SA); a region alone is not valid. An x-default is the fallback for visitors whose language you do not serve, and it was designed for language selectors. The tags can sit in the HTML head, in an HTTP header or in the XML sitemap, and Google treats the three as equivalent.

Return links, the mistake that 9 of the 17 guides we read name, were not the problem in our audit. Every one of the 22 alternate addresses we could read answered 200, pointed back to the page that named it, and named itself as canonical. Several of those sites are Salla, Zid or Shopify stores, where the platform prints the tags. The faults were elsewhere:

• Tags that point at an address that redirects: 4 sites. Three Salla category pages print their own Arabic address with spaces where the real address has hyphens, and the spaced address answers 301. A multi-country retailer lists one country address that answers 302.
• Tags that point at an old domain: 1 travel site. Its English and Arabic tags name another host, and that host’s https address fails the certificate name check, although plain http to its home page answers 200. This is what a half-finished domain move leaves behind: see store migration SEO.
• Switchers that send visitors to the other language’s home page: 2 sites. From an inner page, the link to the other language lands on its home page, not on the twin of the page you were reading.

Test it: View source on the Arabic page and on its English twin and search for “hreflang”. You should see the same set of tags on both, each tag with a full address that opens without redirecting.

Check 5: does each page name itself as canonical?

Google asks that, when you use hreflang, each page names a canonical in the same language, or the best possible substitute if none exists. It also says that for canonical choice it prefers addresses that belong to a working hreflang group. A canonical that names the other language tells Google the page is a copy, and the page can drop out of results. The cleaning company’s English page above is the example in our audit. Everything else we could read, 22 of 22 alternates, named itself.

Check 6: is anything redirecting by IP or browser language?

Google advises against automatically redirecting users from one language version to another on a guess about their language, and against adapting content by IP address, because most of its crawls come from the USA and it does not vary the location to find variations. It suggests links to the other versions instead. Five of the 17 guides we read give the same warning.

Test it from a connection outside Saudi Arabia:

curl -sI -H "Accept-Language: ar-SA" https://example.com/
curl -sI https://example.com/

Both should answer 200 with no Location line. In our own audit, one of the 54 pages answered our connection from Egypt with a redirect to another host with /ar/ in the address, so we left it out of the counts. It is a reminder that routing by connection exists.

Check 7: is there one language per page?

Google says it uses the visible content of a page to find its language. It does not use code-level language information such as lang attributes, or the URL, and it advises one language for the content and navigation of each page, with no side-by-side translations. Three sites in our audit print lang="en" on Arabic text, and one of the 17 guides tells its readers that search engines use the attribute. Google says they do not. Fix the attribute for other tools if you like, but do not count it as a ranking fix: put your effort into the text readers see.

Check 8: are the translated words the ones people type?

A page can be correct Arabic and still target a phrase few Gulf customers search. Our Arabic keyword research guide shows how dialect, spelling and English mixing change the list, so research each language on its own and treat the translated list only as a starting point. Demand for this problem is small and arrives under technical words: our autocomplete pull of 69 seeds returned nothing for any English or Arabic description of the problem, and returned suggestions only for terms such as “multilingual seo” and “hreflang”.

Machine translation needs care at scale. Google’s spam policy names generating many pages through automated transformations such as translating, where little value is provided to users, as scaled content abuse. Its guidance on generative AI asks you to check generated text, titles and descriptions by hand. Translation by a tool is fine as a draft; a native reviewer and the right search words make it a page. If you are hiring someone to write the Arabic, run the ten checks in our guide to Arabic content writing services first.

What we found on 47 Saudi sites

We re-read the language setup of every page that answered 200 in our four earlier audits of Riyadh page-1 results: online stores, service companies, clinics and travel companies. Of 54 pages, 51 were readable, from 47 sites. This is what the raw HTML, and for the sites with no pair there, the HTTP header and the sitemap, say.

What we looked at	Sites	What it tells you
hreflang names a second address, in HTML or header	19	18 language pairs and 1 English-only regional set
hreflang lists only the page itself, or two tags on one address	8	Arabic only, or a language with no address of its own
No hreflang in HTML or header, none in the readable sitemaps	20	Valid for an Arabic-only site; a gap where an English page exists
Language pairs we could read	17	One pair named a host that did not open
Second address really in the other language	16 of 17	One English address serves Arabic
Clean pair, end to end	16 of 47	About one site in three

An Arabic-only site is a valid choice, and 20 of the 47 may be exactly that. The finding that matters is the other side: of the 18 sites that set out to serve two languages, only 16 gave Google a pair that worked, and the ones that failed did so in ways the usual checklists do not test. Return links are in every checklist; an English address that serves Arabic text is in none we read.

Platform notes: Salla, Zid, Shopify and WordPress

We read these help pages on 21 and 24 October 2026 for our platform guides, and they may have changed since.

• Salla: its languages FAQ says only the Growth and Raed themes support multilingual stores, instant translation is paid per character, and instant translation can overwrite manual edits. Its own hreflang guide adds the tags through a Google Tag Manager tag. See Salla or Zid for SEO.
• Zid: English is switched on in settings and is not automatic translation; product and store data must be written again in English.
• Shopify: hreflang tags are created automatically for every published language. Arabic can be added, but it is not on Shopify’s list of languages with professional checkout and theme translations. See Shopify SEO in Saudi Arabia.
• WordPress: bilingual sites usually run a translation plugin, so check the page source of one Arabic page and its twin after every plugin update. Our WordPress SEO guide covers the hreflang settings.

What the fix can be worth

For a travel agency serving the French market, total organic clicks grew 991% in 3 months after we fixed how its language versions were set up, from a small base (monthly clicks from France went from 25 to 455). That is one site’s result, not a promise. It shows why the setup deserves a day of checking before another page is translated.

A brief you can paste to your developer

• Addresses: each language on its own address, with a visible link to the other language on every page, pointing to the twin of that page.
• hreflang: each page lists itself and its twin, full addresses, correct codes, x-default for the fallback. Every address must answer 200 without redirecting.
• Canonical: each page names itself.
• No forced redirects by IP, location or browser language.
• Translation: title, headline, description and body in one language per page.
• Report back with the page source of one Arabic page and its English twin.

How we checked, and the limits

On 30 October 2026 we fetched the 54 pages of our four earlier audits (stores, service companies, clinics, travel) with a phone browser header and without following redirects, read the HTML as it is before scripts run, then fetched up to four alternate addresses from each tag set. For sites with no pair in the HTML we read the HTTP header and the first sitemap files (26 of 29 sites had a readable sitemap, and none printed an hreflang entry). We also read 17 of the 22 guides on page 1 of three Riyadh searches, two in English and one in Arabic, coded by one reader. The other 5 did not return text (3 answered 403, one a redirect, one a navigation shell). Limits: the sample is page 1 of commercial searches, not random, and includes a few large multi-country brands; one page was read per site for the counts; hreflang added by script after load would not show; several regexes were checked by reading the evidence, which changed three counts before publishing; and we did not measure rankings, so no fault here is proven to cost traffic. A sample of this kind shows what is common, not what is typical of the Gulf.

Sources

Google Search Central, opened on 30 October 2026 without following redirects: Tell Google about localized versions of your page, Managing multi-regional and multilingual sites, How to specify a canonical URL, Spam policies for Google web search, Guidance on using generative AI content, Understand the JavaScript SEO basics. Search Console Help: Page indexing report. Platform pages read on 21 and 24 October 2026 are listed in our guides to Salla and Zid and Shopify.

Is your Arabic site set up the way Google expects? Our free 5-point audit checks the addresses, the hreflang pair and the canonicals of both language versions, and tells you which fixes need a developer. It is the first step of our Arabic SEO service and our SEO work in Saudi Arabia. No obligation, delivered within 48 hours. Request your free SEO audit.