The Rutba Stack

How Arabic brands get found, retrieved, and recommended

Someone in Dubai opens an AI assistant and asks, in Arabic, for the best dental clinic in the city. They get back one answer with a handful of clinic names in it. Your brand is in that answer or it isn't. If it is, you get a sale, if it isn't, you do not.

That answer is not random. When a model decides which businesses to name, it runs the same checks every time. It works out who you are, looks for other sources that agree, hunts for facts it can retrieve cleanly, and confirms those facts don't contradict each other. Only then will it say your name out loud.

Those checks are predictable. In Arabic, they are also almost entirely unmet. We mapped them into six layers and eighteen signals, and we call it the Rutba Stack.

SEO, GEO and AEO are not the same thing

Most agencies use these words interchangeably. They describe different surfaces, and work that moves one will not automatically move the others.

All three draw on the same underlying material. A site a crawler cannot read is a site a model cannot recommend. This is why we don't sell them separately. The six layers below feed all three surfaces. What changes between clients is the order we work in.

The six layers

The order matters. Schema markup on a site that crawlers can't reach is a second floor built on nothing. We fix from the bottom up.

1. Foundation

Most Arabic websites fail here and never find out. Sitemaps carry dead links. Pages ship with no H1 tag at all, or with several competing ones. Security plugins block AI crawlers by default, which means the systems deciding what to recommend have never actually read the site. A surprising amount of Arabic copy exists only inside images, where no retrieval system can see it. RTL declarations and hreflang between the Arabic and English versions are wrong more often than they are right.

These are the cheapest problems in the stack to fix, and the first ones to show movement in Google Search Console.

2. Language

For search purposes, Arabic is not one language. It is Modern Standard Arabic, plus a set of live dialects that differ market to market, plus Latin-script transliteration, plus English words dropped into the middle of Arabic queries. A brand optimised for MSA alone is optimised for the way almost nobody types.

This is where translated websites quietly die. A translated page is comprehensible. That is a different thing from being retrievable. We also check whether your Arabic name and your English name resolve to the same business in the eyes of a model, because when they don't, your authority is split in half.

3. Structure

This is the highest-leverage layer available to Arabic brands right now. Implementation is cheap, competitor adoption is close to zero, and results appear in weeks rather than months.

We check whether your facts live in formats machines store: valid schema carrying Arabic values, business hours and locations exposed as data rather than as a styled image, presence as a node in public graphs. We check whether your pages are built in extractable units, with question-shaped headings answered directly underneath. The test we apply to every page is simple. Could a forty-word excerpt stand alone, make sense, and name you correctly? If not, the page will never be quoted.

4. Corroboration

A model will not confidently assert something on the strength of your own marketing copy. It looks for agreement. When your positioning shows up consistently in places you don't control, it stops being a claim and becomes a fact.

Listing consistency is where Arabic service businesses fail most often. Three live profiles carrying three different phone numbers doesn't only confuse customers, it makes you unquotable. Reviews matter here too, but only the ones with text in them. A five-star rating with no Arabic writing attached carries almost no weight. Forty Arabic reviews naming the specific service, in the specific city, is a signal that gets learned.

5. Distribution

Models learn from what is public, free and plentiful. The Arabic pool is much smaller than the English one, so a handful of sources end up carrying disproportionate weight. Every category has them: a regional publication, a professional body, a long-running forum, an Arabic Wikipedia entry, a YouTube channel nobody thinks of as competition.

Finding yours is research, not guesswork, and it forms part of every diagnostic we run. We also look at whether you exist in the formats that get scraped heavily, including captioned Arabic video, published PDFs and guides, recorded talks, and Arabic alt text. Arabic video in particular is badly undersupplied relative to demand.

6. Durability

Correct information that shifts, disappears, or contradicts itself produces hedging and hallucination. You can win every layer above this one and still get skipped, simply because the model isn't confident your facts will hold.

The biggest single culprit in this market is bilingual contradiction. Arabic and English versions of the same website routinely disagree about opening hours, service lists, locations, team members and prices, usually because one side was updated and the other wasn't. When a model finds two versions of you disagreeing, it doesn't choose between them. It becomes cautious about both. Almost nobody audits for this, and it is very fixable.

What it comes down to

Six judgments get made about your brand before anything gets recommended.

Fail one and the answer goes to somebody else. Pass all six in Arabic, where very few businesses have even started, and you become the default answer in your category.

The Rutba Index

Every diagnostic produces a score for each signal, a score for each layer, and one composite figure out of 100.

We always report your score alongside the median for your category, because the gap between the two is where the opportunity sits. Findings are then ranked by impact against implementation cost, and that ranked list becomes the plan. Nobody has to argue about what to do first.

Where the free audit fits

The audit on this site is the first layer of the stack, run automatically. Twenty-three checks covering Foundation, parts of Language, and the machine-readable half of Structure — crawlability, headings, Arabic rendering, hreflang, schema validity, AI-crawler access, answer blocks. It scores in under a minute, costs nothing, and needs no call.

What it cannot do on its own is the research half: dialect coverage for your category, listing corroboration, the Arabic sources your market actually learns from, bilingual contradictions between two versions of your site. That work is human, and it is what the full diagnostic adds.

So the honest version is this. Run the free audit first. It will tell you whether you have a Foundation problem, and most Arabic sites do. Fix that, and the rest of the stack becomes worth paying for.

Arabic service businesses are about to find out that being invisible to AI is much the same as being closed. The ones that fix it over the next eighteen months will hold their categories for years, because the systems learning them now will keep repeating them.