We reviewed the SEO of a language institute whose traffic figures looked impressive. Much of that traffic came from the United States and India. The institute did not operate in those markets or offer its service online.

The numbers described website activity, but they did not demonstrate that the business was reaching suitable customers.

An AI visibility report can create the same problem. A brand may appear frequently in answers collected from another country, in the wrong language or for questions unrelated to the services it wants to sell. Before celebrating the percentage, check what it represents.

“

Before celebrating the percentage, check what it represents.

Barish Elrekabi, DGTL Force

A country setting can change the competitive field

OtterlyAI’s October 2026 regional study compared more than 4.25 million citations across four markets and seven AI engines. UK answers sent 30.7% of citations to UK sites, while the corresponding local-source figures were approximately 76% for Germany and France. The share also varied substantially by engine.

This is vendor research covering Western markets, with its own prompt set and source-classification method. It contains no UAE or Arabic benchmark. We would not use its percentages to forecast Dubai performance.

It does, however, give us a reason to test regional differences rather than assume that one global collection describes every customer’s experience. The UAE implications below are a proposed measurement approach, not findings established by that study.

Define the audience before collecting answers

“Dubai customers” can include residents comparing nearby services, procurement teams evaluating UAE suppliers, parents choosing an academy or overseas investors researching a local provider. Those audiences ask different questions.

Record the target service, geography, customer type and stage of decision. A business serving one neighbourhood needs a different question set from a supplier selling across the GCC.

English and Arabic questions deserve separate attention where both are commercially relevant. Work from how each audience actually asks, with fluent review of the phrasing. A literal translation may preserve the words while missing the local terminology or intent.

Build a question set that reflects buying decisions

Separate informational questions from commercial ones. “What is a volleyball academy?” and “Which volleyball academy offers beginner classes near me?” serve different needs. Both may matter, but their visibility should not be presented as equivalent evidence of purchase intent.

For each service, include a manageable mix of category discovery, provider comparison, location-specific needs and important constraints. Age group, delivery area, certifications, availability or budget may change which recommendation is suitable.

Keep branded questions separate from unbranded discovery questions. It is useful to know whether an assistant describes the business accurately when asked by name, but that does not show whether it introduces the business to someone choosing a provider.

Ask your agency to disclose the measurement conditions

These conditions do not make a test representative of every customer. They make its scope understandable and its changes easier to interpret. An API result and a consumer-interface result should not silently be treated as interchangeable.

Score suitability as well as appearance

A nursery recommended in the wrong city is a poor match. So is an academy recommended for an age group it does not accept. Record whether the answer correctly represents location, service scope and suitability, as well as how prominently it names the business.

Track the supporting sources separately. An assistant may recommend the client using third-party evidence without citing the client’s website. Equally, it may cite the website for a fact without recommending the business.

For a repeatable pilot, select 20 commercial questions per relevant language and run each three times under documented settings. Keep the set unchanged for four weekly checks. This is a practical starting design, not a statistically representative estimate of market share.

Use Google’s generative AI performance reporting as a complementary view of actual impressions by page and country where available. Keep that first-party exposure data distinct from a synthetic prompt benchmark.

Connect regional visibility to the business

Check whether identifiable referrals become suitable enquiries, appointments and customers. Ask where those customers are based and which service they need. Monitor branded search and direct traffic as supporting context, while recognising that other marketing and market conditions affect both.

If visibility rises without enquiries, investigate the questions, geography, recommendations, landing pages and buying cycle. Our AI SEO measurement framework explains that sequence. Our guide to website foundations before more blogs covers the changes that investigation may identify.

We believe clients deserve to understand whether a report reflects the people they can actually serve. That principle shapes our approach to AI SEO in Dubai. Speak to DGTL Force about a baseline grounded in your market, services and customers.

Publishing more articles will not resolve those gaps by itself.

Our starting point is research. Understand the industry, the customer, the competition and the opportunities the business can realistically pursue. Then make sure the core website can support that strategy. Supporting content should have a clear job within it.

Research the opportunity before setting a publishing quota

A keyword with substantial search volume can still be a poor business target. It might attract people in the wrong market, answer a question unrelated to the offer or be too competitive to justify the immediate investment.

We want to understand which services matter commercially, which questions signal buying intent and where competitors leave useful gaps. Search volume, ranking difficulty and the site’s existing strength help prioritise conventional search opportunities. Observed AI answers add another perspective: which brands appear, which sources support them and which decision criteria dominate?

Neither a keyword score nor a single AI answer should decide the strategy on its own.

The result should be a prioritised page and topic map, with a reason for each investment. A monthly article quota is a delivery schedule; it does not substitute for that map.

“

A monthly article quota is a delivery schedule; it does not substitute for that map.

Barish Elrekabi, DGTL Force

Build the pages a buyer actually needs

Review whether the website properly represents the business:

• Does each important service have a useful page explaining scope, suitability and the next step?
• Can customers identify the locations or areas the business genuinely serves?
• Are product specifications, pricing considerations, credentials and constraints easy to find?
• Do case studies explain what was done and what the evidence supports?
• Can visitors move from a useful answer to an appropriate service and enquiry route?

A location-based business may need branch or service-area pages. Those pages should contain meaningful local information, such as facilities, availability, directions or service differences. Creating near-identical pages with different city names does little to help someone make a decision.

This is where the information architecture matters: clear labels, logical navigation and internal links that connect related needs. Good structure helps people find the answer and gives search systems a coherent route through the site.

Make the evidence accessible

Check indexing eligibility, crawler access, canonical URLs, broken links and the availability of important text. Test whether key information is accessible in the environments relevant to the client, including content dependent on JavaScript or protected by security systems.

Google’s documentation for AI features says existing SEO foundations remain relevant. Supporting pages must be indexed and eligible for a search snippet; Google does not require special AI markup or additional machine-readable files for inclusion.

Structured data can describe information consistently when it matches the visible page. Useful FAQs can answer genuine objections. Neither should be presented as a guaranteed route to an AI citation.

Relevance matters more than a formatting checklist

In What Drives AI Citations?, Mahe Chen and Nick Koudas studied 158,099 answers. Pages whose titles and passages matched questions were cited more often. Associations for FAQ blocks, structured data and other page properties disappeared when comparing pages within the same website.

The work is a preprint based on particular commercial markets, so it cannot establish a universal optimisation formula. Its practical implication is to examine whether the page supplies relevant evidence before prescribing superficial additions.

A product page with current technical specifications and clear applications may be a more valuable next investment than another broad explainer. An academy’s branch page may need class ages, schedules and booking information. The right decision follows the buyer’s information gap.

Give every blog a specific role

Once the foundations are understood, blogs can make an important contribution. A post might explain a difficult buying decision, share first-hand experience, analyse original research or resolve a question that does not belong on a service page.

Ask three questions before commissioning it:

1. Which customer need does this answer?
2. What can we add from our experience, evidence or judgment?
3. Which relevant page should the reader visit next?

An article about choosing a fire-rated door core should help someone understand the specification decision and lead naturally to relevant product evidence. An article about academy training should serve the parent or athlete and connect to the appropriate programme or location.

External authority matters too. Identify the credible publications, professional bodies, directories and other sources that actually appear in the client’s relevant answers. Pursue accurate information and useful contributions where there is a legitimate audience, rather than collecting placements without a purpose.

Prioritise by the obstacle to the next result

If the right page does not exist, build it. If it cannot be accessed, investigate the technical barrier. If it lacks evidence, strengthen it. If it attracts the wrong audience, revisit the targeting. If suitable enquiries fail to progress, examine the follow-up process.

Our article on what an AI SEO agency should measure explains how to identify those breaks. For regional businesses, measuring what customers in Dubai actually see is part of the same discipline.

Our SEO services and AI SEO work begin with that investigation. The aim is to put effort where it can make a useful difference, then keep adjusting as the evidence develops.

Imagine a client’s AI mentions increase for three months, but enquiries remain flat. We would call that promising but unproven. The next job is to understand where the connection breaks.

Is the business appearing in buying conversations? Is it being recommended, or simply cited as a source? Are potential customers reaching the website and finding a reason to enquire? Does the buying cycle mean a result will take longer to emerge?

Those questions belong in the report, alongside the visibility chart.

Start with the business decision

Before choosing metrics, establish what success means for the client. A nursery needs suitable families to book visits and enrol. An industrial supplier may need fewer enquiries, provided they come from qualified buyers with substantial projects. A local service business may care most about booked appointments and repeat customers.

The baseline must reflect that difference. The competition matters too. A specialist academy with a handful of relevant competitors faces a different challenge from a coffee shop competing with hundreds of nearby alternatives.

We would set targets after researching the market, current visibility, website and competitive gaps. A universal promise made before that research is difficult to defend.

Measure the journey in layers

These layers should remain distinguishable. A citation to a technical explanation is different from a recommendation to hire the company. A recommendation is different from an enquiry, and an enquiry is different from a customer.

Define the calculations too. For example, recommendation frequency can mean the proportion of eligible test answers that explicitly recommend the client. “Share of voice” needs a stated competitor set and denominator; otherwise two apparently similar percentages may measure different things.

A single screenshot is weak evidence

TJ Digital’s October 2026 citation study tracked 193,507 answers across ChatGPT, Google AI Mode and AI Overviews. Only 24–28% of pages cited for a question were cited again the following week.

It is an agency-client sample, not a universal benchmark, but it makes a useful point: one successful answer does not establish stable visibility. Repeated checks using a consistent question set are more informative. Track the model, interface, market and collection date so a change in measurement does not masquerade as a change in performance.

Our companion article explains why Dubai AI visibility needs its own measurement settings.

Follow the lead beyond the form submission

With a nursery client in Saudi Arabia, marketing was already generating enquiries and enrolments. We still believed conversion from enquiry to visit, and from visit to enrolment, could improve.

After reviewing the marketing, we examined WhatsApp and direct-message conversations. Replies were generic, and follow-up was inconsistent. Parents making a considered decision needed personal conversations and continued contact, rather than being treated as a stream of new enquiries.

We recommended a more personal approach and helped introduce follow-up processes. Both booked visits and enrolments subsequently improved. This was a broader marketing engagement, not an AI SEO experiment, and we cannot isolate the contribution of every change. It illustrates how we believe an agency should work: investigate the business result even when the next improvement involves the sales process.

“

A recommendation is different from an enquiry, and an enquiry is different from a customer.

Barish Elrekabi, DGTL Force

Treat attribution honestly

Someone may discover a business in an AI answer and later search its name, call or visit directly. Referral analytics will miss some of that influence.

Use identifiable AI referrals alongside CRM records and a simple “How did you first hear about us?” question. Monitor branded searches and direct traffic as supporting context. None of these, alone, proves that AI SEO caused a sale.

Google’s generative AI performance report documentation describes impressions across AI Overviews and AI Mode, with page, country and device views. That helps measure exposure; it is not a record of qualified leads or revenue.

For longer buying cycles, compare enquiry cohorts over a suitable period. Report revenue and lifetime value only where the underlying data supports them, and label forecasts separately from realised results.

Expect accountability and a clear next step

An agency can commit to research, implementation, testing and an honest response to the evidence. Rankings, citations, recommendation share and referrals are outcomes it can influence, rather than fully control.

The report should explain what improved, what remains uncertain, where the business is losing opportunities and what will change next. Sometimes that means fixing the core website before publishing more content. Sometimes it means improving follow-up or revising the offer.

That is the thinking behind our AI SEO services. If your current reporting leaves those questions unanswered, talk to DGTL Force about connecting visibility to the results your business needs.