Your buyers are asking AI which companies to consider. In Arabic, yours often is not one of them.

A procurement manager in Riyadh needs exactly what your company sells. He opens ChatGPT, asks in Arabic which companies do it, and gets three or four names back.

If your company is not among them, you will never learn the opportunity existed, because he did not visit your website, did not fill in a form, and did not call.

A group of colleagues in a Riyadh office at dusk, the city skyline behind them

Here is that same question, asked twice, for a distributor we will call Mada Technology.

A buyer in Riyadh, asking in Arabic
أفضل شركات توزيع حلول تكنولوجيا المعلومات في السعودية
من أبرز الشركات العاملة في توزيع حلول تكنولوجيا المعلومات في السوق السعودي:
  1. شركة الأفق للأنظمة
  2. مجموعة التقنية المتحدة
  3. دار النظم للحلول المتكاملة
وتُعد هذه الشركات من الأكثر حضوراً لدى الموزّعين وشركاء القطاع.
Mada Technology was not named
The same buyer, asking in English
best IT distributors in Saudi Arabia
The leading enterprise IT distributors operating in Saudi Arabia include:
  1. Al Ufuq Systems
  2. United Technology Group
  3. Dar Al Nuzum Solutions
  4. Mada Technology
Each maintains vendor authorisations and regional distribution coverage.
Named, fourth of four

An illustration of the pattern these audits keep finding. Mada Technology and the companies beside it are invented. The shortlist a buyer receives depends on which language the question was asked in, and nobody inside the company ever sees either answer.

A study of close to four thousand B2B buyers last year found that in 95 percent of cases, the company that won the deal was already on the list the buyer had written before contacting anyone.1 Buyers get about two thirds of the way through a decision before they speak to a single vendor, and 94 percent of them use AI somewhere along the way.

The shortlist is written before anyone calls you.

Arabic is spoken by more than four hundred million people and makes up under one percent of the content on the web, while the UAE and Saudi Arabia have the highest rates of workplace AI use in the world. So the region asking the most questions is asking them in the language these systems know least, and the answers get assembled from whatever thin and outdated material exists.

Tawaator counts how often AI assistants name your company when the buying question is asked in Arabic, compares that against English, shows which competitors appeared in your place, and then corrects the sources those answers are built from.

We work with B2B companies selling into Saudi Arabia, the UAE and Egypt.

What you actually get

Three numbers that did not exist before, each recorded before any work starts and measured again every month afterwards.

Corrections take weeks rather than days to show, because a search engine has to find and index the new pages before any assistant can read them. Nothing here is promised by a date. Everything is counted.

A dense field of faint tally marks with four marks standing sharp and upright near the centre

Three or four names make the list. Everything else in the field is invisible to the buyer.

The problem, in three parts

None of this is about a badly built website. All three failures are invisible from the inside, because every page involved renders perfectly in a browser.

A procurement manager working alone at a desk in an empty office, lit only by a screen
i.

The decision happens without you

By the time a buyer makes contact, the comparison is already finished and the list is already ranked.

A buying decision used to start with a phone call. It now starts with a question typed into a chat window by someone you will never meet, who describes the problem, asks which companies solve it, and gets three or four names back.

Nothing about that process is visible to a vendor. No form is filled in, no page is visited, no call is made, and no analytics system records that the opportunity existed.

A single aged red index card pulled forward from a drawer of hundreds of identical cards
ii.

Somebody else writes your Arabic description

In Arabic, the account a buyer receives usually comes from a directory entry nobody inside your company has read.

The assistant does not answer from memory. It runs a live search and writes its answer out of the pages that come back, and to compare several companies it needs sources that mention several companies together. That means directories and trade press rather than anyone's own website.

Arabic is spoken by over four hundred million people and makes up under one percent of the content on the web, so that third party material is thin and mostly years out of date.

Eight scattered faint variations of the same mark resolving into one sharp red mark
iii.

Your name competes against itself

Split across six names, none of them accumulates enough authority to win a comparison against a competitor using one.

Most companies here circulate under several names at once. An English trading name, a legal entity name, an abbreviation, then two or three Arabic spellings each chosen by a different person in a different decade.

Every mention of your company attaches to whichever name was used that day, and an assistant reading all six has to decide whether they describe one company or four. It decides differently on different attempts.

A worked example. One fictional company, six names in circulation.

In English

  • Al Noor Industrial
  • AlNoor Industries LLC
  • Al-Noor Industrial Company

In Arabic

  • النور الصناعية
  • شركة النور للصناعات
  • النور إندستريال
The trade licence says one of these. The website header says another. LinkedIn says a third. The Arabic press release from 2019 says a fourth. To an AI assistant reading all six, this may be one company or it may be four unrelated ones.
A stone archway with three misaligned red blocks near its crown

What we do about it

Nobody can edit an AI assistant. There is no console, no submission form and no appeal. What can be changed is the material it reads before it answers, and since it builds every answer out of the pages a search returns, changing those pages changes the answer.

That single fact is the whole business. Everything below follows from it.

  1. Build the question set

    Twenty five to forty real buying questions for your sector, written in Modern Standard Arabic, in Gulf phrasing, in Egyptian phrasing and in English. This is research into how your market actually buys rather than a translation exercise, and it is where most of the value sits.

  2. Count what comes back

    Every question is asked twenty times in each language, across a mix of assistants, from clean sessions with no account signed in. Ask twice and you get two different answers, so a single screenshot proves nothing. The result is a count out of twenty rather than a yes or a no.

  3. Find the cause

    Which sources the answers were built from, how many names you circulate under across both scripts, which of your own facts contradict each other, and whether your pages can be read at all.

  4. Correct the sources

    Arabic pages on your own domain, one canonical name in each script applied everywhere, structured data stating your facts, corrections filed with the directories and data brokers, and the technical fixes that let your pages be read.

  5. Measure again against a control

    The same questions, the same number of runs, repeated at six weeks and three months, alongside two or three named competitors who had no work done. If you rise and they hold steady, the movement came from the work.

Engagements

Every engagement starts with a diagnostic, because the findings decide what is worth fixing. Prices are listed because you should be able to decide without booking a call.

An audit tells you what is happening. The build is where it gets corrected.

These are priced separately on purpose. An audit establishes what AI assistants currently say about your company, which competitors appear instead, and which sources are producing those answers. It does not include the corrective work, and most companies want to see the findings before committing to that. Monthly tracking comes after a build, never before, because there is nothing meaningful to track until something has changed.

Source audit

$750
Three working days
  • Every name you circulate under, in both scripts
  • The sources answers about you are built from
  • Contradictions across your own properties
  • Technical legibility of your site
  • Fix list ordered by effect
Diagnosis and a plan. No measurement of assistants.

Full audit

$2,500
Five working days
  • Everything in the source audit
  • 25 to 40 questions in four language registers
  • Counted results across five assistants
  • Your appearance rate against named competitors
  • Targets for three and six months
  • Two hour working session on your questions
The report arrives three days ahead, so the session goes on your questions rather than a walkthrough.

The build

$4,000
to $6,000, four to eight weeks
  • Arabic pages written and published on your domain
  • One canonical name applied across every surface
  • Structured data across the site in both languages
  • Directory and data broker corrections filed
  • Technical fixes deployed and verified
  • Re-measurement at six weeks and three months
Tracking is available afterwards and is described below.

Monthly tracking

$999
per month, sold only after a build

Appearance rates move slowly and they move in both directions. A competitor publishing an Arabic site, a directory updating its records, or an assistant changing how it retrieves sources will all shift your position without anyone telling you. Tracking is what turns a one time correction into something you can defend.

What arrives each month

  • Fifteen questions from your frozen set, re-run twenty times in each language across five assistants. Same wording every month, because changing a word invalidates the comparison
  • Your appearance rate in Arabic and English, against the baseline recorded before any work started
  • The same two or three named competitors measured alongside you, so a rise can be told apart from a market wide shift
  • The sources feeding the answers, ranked, with any new source flagged and any source that changed its description of you
  • One prioritised recommendation, which is usually a page to write or a listing to correct
  • A written report of three to four pages. No dashboard and no login
No call is required in any month. If you want to discuss a report, ask and we will find a time.
Colleagues in discussion around a meeting room table, seen through a glass wall

How the working session runs

The report reaches you three days before we speak. You read it, mark up whatever you want to go into, and the two hours go entirely on your questions and on deciding what to fix first.

There is no deck and no walkthrough, because reading a report aloud to the people who already have it wastes everybody's afternoon.

If you would rather not meet at all, send your questions in writing and you will have a written answer inside two working days. A good share of clients prefer this and it changes nothing about the outcome.

Start with a free scan

Three Arabic questions for your category, run against the major assistants. You get back whether your company was named, which companies were named instead, and the source the assistant leaned on hardest.

Every scan is run by hand rather than by a tool, because the programming interfaces these companies offer do not return what your buyers actually see. That takes a day and it costs nothing.

A work email on your company's own domain, please. It is how we confirm the request is yours to make.

Nothing is automated and nothing is generated. The three questions are written for your sector, the runs are counted, and the reply names the competitors that appeared in your place.

No call is required at any point. If you would rather ask questions in writing, that is the default.