AI Search Is Changing Where Business Inquiries Come From: The Data for the Gulf and Egypt

Key takeaways

• Clicks from search are falling. An Ahrefs study published in February 2026 found AI Overviews correlate with a 58% reduction in click-through rate for top-ranking pages, up from 34.5% measured in April 2025.
• The visits that remain are worth more. Across several independent 2026 datasets, visitors arriving from AI assistants convert between roughly 4x and 23x better than visitors from classic organic search.
• This arrives in the Gulf earlier than in the West, not later. The UAE reported 64% of its working-age population using AI tools at the end of 2025, and Saudi Arabia reported 45.2% of internet users, more than double the previous year.
• Arabic is no longer excluded. AI Overviews reached MENA and Arabic globally, and AI Mode added Modern Standard Arabic, so the shift now applies to Arabic queries too.
• Your reporting is probably wrong before your marketing is. Most analytics setups in this region undercount both AI referrals and chat-based enquiries.

The short answer

Businesses in Egypt and the Gulf are receiving fewer visits from search and a similar or larger number of enquiries. That is not a contradiction. Search engines now answer more questions directly, so the casual browsing traffic disappears, while the people who still click through have already read a summary, formed an opinion, and decided you are worth visiting.

The practical consequence is that traffic has stopped being a useful health metric. If you are still judging your marketing by sessions, the picture will look worse than reality. If you judge it by enquiries and by where those enquiries started, the picture is usually different, and in some categories it is better than it was.

What actually changed

Three independent measurements from 2025 and 2026 point the same direction, using different methods.

Two points of caution before you use these numbers in a board deck. First, they measure correlation, not a controlled experiment. Queries that trigger an AI summary are not a random sample, and informational queries were always more likely to end without a click. Second, publisher data describes media businesses, whose traffic mix is unlike a services company or a retailer. Directionally the finding is consistent. Precisely, it is not transferable to your situation without checking your own numbers.

Fewer visits, better visits

This is the part most coverage of the topic leaves out, and it is the part that matters commercially. Several 2026 datasets found that visitors arriving from AI assistants convert far better than visitors from classic search results.

By assistant, reported conversion rates were 15.9% for ChatGPT referrals, 10.5% for Perplexity and 5.0% for Claude.

Now the honest caveat, because these numbers are quoted everywhere without it. Conversion is not defined the same way across these studies. Some count form submissions, some count purchases, some count product signups. A 23x differential measured on free product signups is not the same claim as a 4.4x differential measured on purchases. Before you import any of these figures into a forecast, check which event each one counted, and then measure your own.

What survives that scrutiny is the direction, which every dataset agrees on: a visit that begins in an AI answer is worth more than a visit that begins in a list of blue links. The person has already been told what you do and why you might fit. They arrive further along.

Why this reaches the Gulf before it reaches the West

The common assumption is that new search behavior starts in the United States and arrives here two years later. On this particular shift, the opposite is true, and the published adoption figures are not close.

ChatGPT was the most downloaded AI application in Saudi Arabia, ahead of Gemini and DeepSeek.

Read those two rows next to the click-through data and the implication is uncomfortable. A business in Dubai or Riyadh is exposed to this change sooner and more intensely than an equivalent business in London or Chicago, because a larger share of its market has already moved. Waiting to see how it plays out abroad means waiting for news about a market that is behind yours.

The demographic detail in the Saudi data is worth a second look if you sell to consumers. Adoption is highest among people aged 20 to 29 and higher among women than men. If that is your customer, the shift in how they find you has already happened, whatever your traffic report says.

Arabic removed the last excuse

Until recently there was a defensible reason for regional businesses to postpone this. The features were English-only, so an Arabic-language business could reasonably wait. That reason has expired.

• AI Overviews are now available across MENA, and in Arabic globally, on mobile and desktop.
• AI Mode added Modern Standard Arabic as part of an expansion to 36 further languages, following its English launch in MENA in August 2025.

There is a competitive consequence here that favors whoever moves first. Arabic results contain far fewer well-optimized pages than English ones for the same commercial intent. When an assistant assembles an answer to an Arabic question, it has a much thinner set of credible sources to choose from. Being one of them is materially easier in Arabic than in English right now.

What this does to your enquiry pipeline

Four changes show up in practice, in roughly this order.

1. Top-of-funnel traffic falls first. Definitional and explanatory queries stop producing visits, because the answer is given on the results page. This is the traffic that never converted anyway, which is why revenue often holds while sessions drop.
2. Enquiries arrive better informed. People who contact you have already read a summary of your category, and sometimes of you specifically. They ask sharper questions and move faster.
3. You get shortlisted before you are visited. The assistant names a handful of options. If you are not among them, you never enter the process, and nothing in your analytics records the loss.
4. Attribution degrades. A meaningful share of these arrivals show as direct traffic, so the channel that produced the enquiry becomes invisible in standard reporting.

The third point is the one that costs the most and shows up the least. Losing a ranking is visible in a report. Being omitted from a recommendation is not visible anywhere.

How to measure it properly

Before changing any tactics, fix the measurement. Almost every regional business we look at is undercounting this in at least two places.

1. Separate AI referrals from direct traffic

Create a channel group for referrals from assistant domains and treat it as its own source. Until you do, the highest-converting segment you have is buried inside direct traffic, and you cannot make a case for investing in it.

2. Count chat enquiries as conversions

This one is specific to our region and it is the larger error. WhatsApp reaches more than 83% of internet users in Saudi Arabia and above 80% of the population in the UAE. When a visitor reads your page and opens a chat to enquire, standard analytics records an exit. Tag the click as a conversion and add a source parameter so you know which page produced it. Businesses that do this routinely discover their conversion rate was understated by a wide margin.

3. Track whether assistants mention you at all

Rankings tell you where you sit in a list. They do not tell you whether an assistant recommends you when someone asks for options in your category. Those are different questions with different answers. Build a fixed set of prompts a real buyer would ask, run them on a schedule across several assistants, and record whether you appear and how you are described.

Run the same prompts in Arabic and English, and consider adding a regionally built Arabic assistant to the set. They frequently return different sets of companies, and the Arabic set is usually the shorter one, which is where the opportunity sits.

4. Stop treating “unknown source” as dirty data

Most marketing teams look at the share of enquiries with no identifiable source and treat it as a gap to be closed. It is worth reading differently. A growing unknown share is itself a finding, and it usually means the path from discovery to contact now runs through more places than your tracking covers.

The typical journey in this region now looks something like this. A buyer asks an assistant which suppliers are worth considering. They search your brand name to check you are real. They look at your site. They read something about you elsewhere. Then, days later, they message you on WhatsApp. Ask them where they found you and the honest answer is that they cannot remember, because there was no single place.

Last-click attribution cannot describe that, and no amount of tag configuration will fix it. The more useful question is not “which channel produced this enquiry” but “where did this person learn what they now believe about us”. That is a different measurement problem, and the honest first step is admitting your current setup does not answer it.

One practical substitute costs nothing. Ask every new enquiry where they first heard of you, record the answer verbatim, and review the list monthly. It is imprecise and unscientific, and it will still tell you more about this shift than your analytics will.

The case against acting on this yet

Three objections to everything above are worth taking seriously, because two of them are partly right.

“Assistants are unstable. They mention you today and forget you tomorrow.”

Correct, and anyone who tells you otherwise is selling something. Answers vary between runs on identical prompts, and models change without notice. The conclusion people draw from this is usually wrong though. Early search rankings were unstable too, and the businesses that built content and recognition during that instability were the ones holding position once it settled.

The realistic goal is not a guaranteed appearance for a specific question. It is being inside the set of companies an assistant considers credible when it is asked about your category. That accumulates slowly, which is precisely why it cannot be assembled quickly once buyers have already started asking.

“The volume is too small. Advertising delivers more.”

Also fair on volume. Referral traffic from assistants is a small fraction of total sessions for most businesses today. But volume is the wrong measure for this channel, for two reasons.

First, the conversion difference. In the datasets above, AI-referred visitors converted several times better than organic ones, so a small share of sessions can carry a disproportionate share of enquiries. Second, and more importantly, assistants influence the stage before the click. If a buyer has already asked which suppliers are credible and you were not named, your advertising now has to overcome a judgment that was formed without you. That makes the ad more expensive, and the cost never appears in the ad account.

“There is no standard method and it cannot be measured.”

Half right. It cannot be measured the way SEO is measured, because there is no position to record. It can be measured, and the fact that most people are not doing it is the opportunity rather than the objection.

What is measurable: how often you appear across a fixed question set, how you are described when you do, which sources the answer cites, whether competitors appear more often than you, and how many enquiries arrive from assistant referrals. None of that requires a vendor’s proprietary score. It requires writing down the questions and running them on a schedule.

One action, depending on your role

If you do nothing else after reading this, do the one item that matches your job.

The founder item is the one we would push hardest. Marketing teams tend to report on channels that are already measured. The gap this article describes is invisible in every dashboard you currently own, and the fastest way to believe it is to go and look.

What we cannot tell you yet

Every figure in this article comes from studies conducted on predominantly English-language, predominantly Western datasets. We have applied them to this region because they are the best evidence available, not because anyone has verified they hold here.

Nobody currently publishes measured data on which sources AI assistants cite for Arabic queries, how often those citations change, or how referral behavior from Arabic prompts differs from English ones. We are building a measurement panel to answer exactly those questions, and we will publish the method and the raw data alongside the findings. Until then, treat the regional parts of this article as reasoned inference from adoption data, clearly labelled as such.

We would rather say that plainly than present an estimate as a measurement.

Frequently asked questions

Is SEO finished?

No, and the framing is misleading. Assistants build answers largely from pages they can find, read and trust, which are the same pages that rank, a distinction we cover in AEO versus GEO. What has changed is the payoff. Ranking used to produce a click. Now it can produce a mention that leads to an enquiry without a click, which means the value is real but harder to see.

Our traffic fell but enquiries held steady. What happened?

The most likely explanation is that you lost informational visits that never converted, while your commercial visits held. Check whether the decline is concentrated in explanatory pages rather than product or service pages. If it is, the loss is largely cosmetic.

Does this apply to Arabic-language businesses?

Yes. AI Overviews are available across MENA and in Arabic globally, and AI Mode now supports Modern Standard Arabic. Arabic competition is thinner, so the opportunity to become a cited source is currently larger in Arabic than in English.

How do I know if assistants recommend my company?

Ask them, systematically. Write down the questions a buyer in your category would actually type, run them across several assistants in both languages, and record whether you appear and what is said about you. Repeat on a fixed schedule, because answers vary between runs even with identical prompts.

Where should a business in Egypt or the Gulf start?

Fix measurement first, because you cannot manage what you are not counting. If you are evaluating outside help, our buyer’s checklist for choosing a GEO agency lists the questions worth asking. Separate AI referrals from direct traffic, tag chat enquiries as conversions, and establish a baseline of whether assistants mention you. Our GEO Guide covers the full method. Those three steps take days and change what every subsequent decision is based on.

Sources

• Ahrefs, click-through rate study on AI Overviews, February 2026, and prior measurement April 2025.
• Pew Research Center, observational study of 68,000 search queries.
• Opollo AI Search Benchmark 2026, GA4 data from 312 B2B technology firms.
• Semrush 2026 cross-industry conversion data.
• WebFX session analysis, 2.3 billion sessions.
• Published AI adoption reporting for the United Arab Emirates and Saudi Arabia, 2025 to 2026.
• Google product announcements on AI Overviews availability in MENA and Arabic, and AI Mode language expansion.
• DataReportal 2026 digital reports for messaging platform reach.

Related guides

• How to Get AI Search Answers to Recommend Your Business (GEO Guide)
• How to Measure AI Visibility (Share of Model): A 2026 Playbook
• Best AI Search & GEO Agency in the Middle East, Gulf & Egypt (2026)
• Answer Engine Optimization (AEO) Services in Egypt: How to Win Zero-Click Search