What being cited actually means
Being cited by an assistant is not the same as ranking in search. Depending on the product and the question, the model is retrieving pages live, summarising what it fetched moments ago, or drawing on patterns absorbed during training. Only the first two respond to anything you publish this quarter.
That distinction decides where the money goes. Training data is a lagging record of everything written about you over years, and nothing published today will change it soon. Live retrieval responds to what exists and is corroborated now. If you want movement inside two quarters, you are working on retrieval, which means pages that answer cleanly and sources that repeat you.
Why corroboration beats domain authority
Classical search rewards the strongest single domain. An assistant behaves more like a careful researcher: it prefers a claim several independent sources agree on over one asserted confidently by a single site, however polished that site is. The question it is implicitly asking is not who said this loudest, but who else says it.
The practical consequence is that the work looks more like public relations than like search engine optimisation. A link from a site nobody reads still passes authority in a classical ranking model. A mention in a publication a model has learned to trust does something different: it corroborates. Those are not the same purchase, and most regional budgets are still buying the first one.
“The assistants don’t reward the loudest site. They reward the one everyone else corroborates.”
Why this hits Gulf brands harder
Three things compound in the region. The local publisher ecosystem is thinner than in Europe or North America, so fewer independent sources exist to corroborate anything. Directory and listing sites occupy much of the space that trade press occupies elsewhere. And a large share of genuinely authoritative regional information sits in PDFs, government portals and LinkedIn posts rather than on indexable pages.
The effect is that a Gulf brand can be well known commercially and nearly invisible to an assistant. Everyone in the industry knows who you are. Nothing on the open web says so in a form a model can read, and a model has no way to tell the difference between a quiet market leader and a company that does not exist.
The Arabic gap
Ask the same commercial question in Arabic and the set of sources an assistant draws on thins out sharply. In several categories the answers fall back on translated English sources, or on whichever regional portal happens to have a page. For a brand willing to publish properly structured Arabic answers, this is the least contested surface available anywhere in the discipline.
It is not a translation job. A machine-translated version of an English page inherits the English page’s framing and usually loses the declarative opening that made it quotable in the first place. The answer has to be written in Arabic, by someone who knows how the question is actually asked, which is rarely a literal rendering of how it is asked in English.
Where we would start
Pick the commercial questions your buyers actually ask, check what the assistants currently say, fix the pages that should be cited and are not, then go and earn independent mentions on those exact topics. The order matters: the page is a prerequisite, and the mentions decide the outcome.
Pick the questions before you pick the pages
Start from the questions your sales team answers every week, in the buyer’s wording rather than your category’s. The gap between “AI enablement consultancy” and “how do I get my team to actually use AI” is the whole game. Build pages against the second kind, because that is the question anyone actually types.
Fix the page, then go earn the mentions
The page is a prerequisite, not a strategy. Once it answers cleanly under its heading, the remaining work is getting independent sources to say the same thing: trade press, industry bodies, partner sites, conference programmes, anything that is not your own domain. This is the slow half, and it is the half that decides whether you are ever quoted.
Watch the citation set, not your position
There is no ranking to track. What you can track is which sources an assistant names for your questions, logged monthly in a spreadsheet. The useful signal is movement. A set that never changes means the topic has settled and you are late. A set that churns means it is still winnable. The audit itself is a spreadsheet and an afternoon.
What we would not bother with
Publishing high volumes of thin articles to feed the models does not work, and it costs you the only asset you had, which is a reason for anyone to reference you. Nor does bolting schema onto pages with nothing in them. The common failure is a brand producing more content while remaining uncorroborated, which is the expensive way to stay invisible.
None of this is exotic. It is the same discipline that won featured snippets a decade ago, pointed at a new set of readers. In Arabic, where the field is still close to empty, it is startlingly cheap.