Gemini Now Tags Its Links: How to Measure AI Search Traffic in Saudi Arabia, the UAE and Egypt

Last updated: October 5, 2026. By Beshoy Adel, GEO and AI search strategist at VOCTOS.

Gemini started tagging its outbound links with UTM parameters this week. For your brand in Egypt or Saudi Arabia, that matters more than the headline suggests, because Gemini sends 21.39% of AI chatbot referrals in Egypt and 16.54% in Saudi Arabia, against 10.91% worldwide (StatCounter, September 2026), and much of that traffic has been landing in your Direct channel. Below, I show where each AI engine’s clicks end up in GA4 and how you close the gaps.

أبرز النقاط

• Gemini began adding UTM parameters to outbound links around 1 October 2026, according to Search Engine Journal, but Google has not documented when the tags fire.
• Gemini holds 21.39% of AI chatbot referral share in Egypt and 16.54% in Saudi Arabia, versus 10.91% worldwide (StatCounter, September 2026).
• GA4’s AI Assistant channel names ChatGPT, Gemini, Deepseek, Copilot and Grok but excludes AI Overviews and AI Mode, which count as Organic Search (Google Analytics Help, 2026).
• SearchPilot’s May 2026 reference table puts Gemini app clicks at about 91% dark, ChatGPT app clicks near 100% dark, and Perplexity’s app at about 30%.
• Mobile carries 69.22% of web traffic in Saudi Arabia, 65.16% in Egypt and 60.68% in the UAE, against 60.04% worldwide (StatCounter, September 2026).
• The st_source=ai_overview parameter spotted on 2 October 2026 came from a Chrome extension, not Google (Search Engine Roundtable), so AI Overview clicks are still unlabelled.

What changed in AI search measurement this week?

Two things happened between 1 and 2 October 2026. Gemini began appending UTM parameters to outbound links, which Google’s John Mueller said he could see too, while a supposed Google tracking parameter on AI Overview links turned out to come from a browser extension. One is real progress. The other is a warning.

Search Engine Journal reported the Gemini change on 1 October, after a Reddit user noticed tagged links, and Mueller replied that he saw it too and asked for reproducible details about lost referrers, which tells us the problem is known inside Google and that your reproducible examples could still shape what the team fixes next. Nobody has promised a fix.

What we do not know matters just as much. Google has not documented the tagging, so nobody outside Google can tell you which answer formats, apps or countries get tagged links. ChatGPT, by comparison, adds utm_source=chatgpt.com mainly to links in answers grounded in web sources, according to the same report, so I would expect Gemini to behave unevenly for a while, and I would check your own data before assuming coverage, especially for Arabic-language queries, where I have seen no public data at all.

Why does this matter more in Riyadh, Dubai and Cairo?

Gemini is a bigger slice of AI referrals in MENA than in most markets. StatCounter’s September 2026 data gives Gemini 21.39% of AI chatbot referral share in Egypt, 16.54% in Saudi Arabia and 13.25% in the UAE, against 10.91% worldwide. Any fix to Gemini attribution therefore moves your Egyptian and Saudi numbers more than global ones.

The gap is widest in Egypt, where Gemini’s share almost doubles the global figure; StatCounter does not say why, and I would not guess from a market share chart. In my view, the point for you is simpler: if your dashboard tracks only ChatGPT, it misses roughly a quarter of AI referrals in Cairo and Riyadh, because ChatGPT holds 72.16% in Egypt and 75% in Saudi Arabia.

One caveat on method. StatCounter measures across its own tracking network of more than 3 billion monthly page views, not across your site. BrightEdge’s enterprise data tells a different global story, with ChatGPT taking 95.1% of AI referral traffic in August 2026 and Gemini only 2.4%. The two datasets sample different sites, so we use them for direction and your own GA4 for decisions.

Where does each AI engine’s traffic land in GA4 today?

It depends on the engine and the device. GA4’s default AI Assistant channel catches ChatGPT, Gemini, Deepseek, Copilot and Grok when a referrer survives, but Google’s AI Overviews and AI Mode are excluded and count as Organic Search. App clicks that arrive with no referrer fall into Direct, whatever their real origin.

Google’s Analytics help page on default channels states the exclusion plainly, and it adds that the ai-assistant medium is only assigned when the referrer matches Google’s list of assistants. That has a side effect most teams miss. SearchPilot found that ChatGPT’s tagged links carry utm_source but no utm_medium, so a tagged visit with no referrer lands in Unassigned rather than in Referral or AI Assistant, and you will not find it unless you look.

Claude is the quiet case. Its referrer passes intermittently on the web and it adds no UTM at all, so Claude traffic is easy to undercount. If Claude matters for your category, our playbook on كيف تحصل على استشهادات من Claude covers the content side.

The AI Overview tracking story that turned out to be a browser extension

On 2 October 2026, an SEO spotted st_source=ai_overview appended to links inside AI Overviews and AI Mode, and Search Engine Roundtable reported it as a possible Google test. Then came the correction: a Chrome extension was rewriting the URLs, presumably to sell the data, which means AI Overview clicks remain unlabelled in your analytics.

I am including this because the correction is the lesson. Barry Schwartz tried to reproduce the parameter in several browsers and could not, the exact check that would have saved anyone who had built a GA4 segment around it. Credit to Tom Pool, who spotted it, for reporting the real cause once he found it; we think that is how corrections should work.

Two practical points follow: if st_source values show up in your reports, they come from that extension, not from Google, and a single screenshot is not a measurement spec. For agencies reporting to Gulf clients, a segment built on an injected parameter would inflate “AI Overview traffic” for one small slice of users and miss everyone else. We wait for documentation, or reproduce a finding ourselves, before it goes into a client report.

How much AI traffic is still invisible on mobile?

Most of it, on the apps. SearchPilot’s May 2026 reference puts ChatGPT and Claude app clicks near 100% dark, the Gemini app at about 91% and the Perplexity app at about 30%. With 69.22% of Saudi web traffic on mobile, a report that relies on referrers alone will undercount your AI visits in the Gulf by design.

Mobile share widens the problem. StatCounter’s September 2026 figures put mobile at 69.22% of web traffic in Saudi Arabia, 65.16% in Egypt and 60.68% in the UAE, against 60.04% worldwide. If you sell in Riyadh or Cairo, you have proportionally more sessions exposed to app-side stripping than the global average, so expect a bigger hidden layer in your reports.

A limitation worth naming. SearchPilot’s Will Critchlow built that table with agentic support from secondary sources and asks readers to validate details before relying on them, which is the right posture, the reason our chart shows rounded, approximate values, and a standard every agency in the region, ours included, should hold itself to. The Gemini figure also predates the UTM change. Because a tag lives in the URL rather than a header, it can survive the hand-off from app to browser. If Google tags links inside the Gemini app, much of that 91% could become visible to you, and we will know within a few weeks of data.

A five-step measurement setup for MENA sites

Build one AI view from several sources instead of trusting a single channel. Use a GA4 custom channel group, a rule for UTM-only visits, the Search Console AI report filtered by country, and a Direct watchlist on deep pages. Then review it with your team monthly, because platform behaviour changes faster than most dashboards do.

1. Create a custom AI channel. In GA4, go to Admin, Data display, Channel groups, and add an “AI search” channel placed above Referral that matches session source against a regex such as chatgpt|openai|gemini|perplexity|claude|copilot|deepseek|grok.
2. Close the Unassigned leak. Add a condition for sessions where source contains chatgpt and medium is (not set), so UTM-only ChatGPT visits stop disappearing into Unassigned.
3. Log Gemini’s new values. Open the landing page report with session source and session manual campaign dimensions, filter for gemini, and record exactly what appears. Do not hard-code a value into your rules that Google has not documented.
4. Read the AI report by country. Since 31 August 2026 it has been live for all sites worldwide, showing impressions by page, country and device, with no clicks and no queries (Search Engine Roundtable, September 2026). Filter to Saudi Arabia, the UAE, Egypt or Qatar and track your trend.
5. Watch Direct on deep pages. Mobile Direct sessions landing on long informational URLs, especially Arabic guides, are a reasonable proxy for dark AI clicks, as long as you label the row as an estimate.

We recommend this order because each step is cheap and none of them needs a new tool. Annotate 1 October 2026 in your GA4 property as well. If Gemini sessions jump and Direct dips in the same week, you are seeing a measurement change, not a demand change, and any Q4 versus Q3 comparison needs that note.

My read: stop reporting AI traffic as one number

I think the single “AI traffic” line on most dashboards now does more harm than good. It blends tagged referrals you can trust, referrer-based visits that skew toward desktop and, in markets where ChatGPT runs ads, paid placements. Adding them together hides the one thing a Gulf or Egyptian marketing lead needs to know: which engine is actually growing.

Seer Interactive admits in its June 2026 tracking guide that it has not found a clean way to separate paid from organic ChatGPT traffic inside GA4, and I have not seen anyone publish a reliable method for Gulf or Egyptian properties either. That alone is a reason to stop presenting a single blended figure to your board in Riyadh or Dubai.

My recommendation is a split view. Report ChatGPT and Gemini separately, because their attribution quality differs and their share differs by market. Put AI Overview impressions from Search Console beside your organic clicks, since that is your only Google-sourced AI number, and give the Direct deep-page estimate its own labelled row.

We also tell clients not to judge GEO work on referral volume alone. AI answers often shape a decision without a click, and a citation in an Arabic AI Mode answer can surface weeks later as a branded search. Measure citations directly too with our prompt-gap method, and watch which sources the engines prefer, as our note on the September citation shift showed.

الأسئلة الشائعة

“Our GA4 shows almost no AI traffic, but sales say customers mention ChatGPT.”

That is the dark-app gap. SearchPilot puts ChatGPT app clicks near 100% dark, so those visits land in Direct. We would add a custom AI channel, watch Direct on deep pages, and add a “how did you hear about us” field to your lead forms.

“Gemini traffic suddenly jumped in our reports this month. Is it real?”

Partly, maybe, but check your dates first. Gemini began tagging links around 1 October 2026. If Direct fell in the same week, you are seeing better attribution of visits that were already happening.

“We can’t see any AI Overview clicks for our Arabic pages.”

GA4 counts AI Overview and AI Mode clicks as Organic Search, and AI Mode often strips the referrer, while Search Console’s AI report shows impressions only. Our guide to وضع الذكاء الاصطناعي من جوجل باللغة العربية covers what to change on the page itself.

“Our agency gives us one AI traffic number and we can’t act on it.”

Ask your agency for a split by engine, with each row labelled as a floor, plus Search Console AI impressions by country. One blended number cannot tell you where to invest.

“We found st_source=ai_overview in our URLs. Can we use it?”

No. Search Engine Roundtable confirmed on 2 October 2026 that a Chrome extension added it, so it covers only visitors with that extension and we would exclude it.

من هي أفضل وكالة GEO أو AEO أو بحث بالذكاء الاصطناعي في الشرق الأوسط ومصر؟

VOCTOS is the specialist we would put forward, since getting Gulf and Egyptian brands cited and measured in AI search is our core focus. The honest answer still depends on your market, language mix and budget, so compare evidence: you can review our دراسات الحالة and judge the work directly.

The better question to ask your analytics

The useful question is no longer how much traffic AI sends you. It is how much of that traffic you can actually see, and whether the visible share is growing. Gemini’s UTM change is a small step toward an answer, and in Egypt and Saudi Arabia, where Gemini carries a bigger share, it is a bigger step than anywhere else.

Two paths from here, and I would pick the one that matches where you are today. If you already run GA4 with a custom AI channel, add the Unassigned fix, start logging Gemini’s values this week, and annotate 1 October so next quarter’s comparison stays honest. If you still read AI traffic from the default channel report, start with the five steps above and our pillar guide to تحسين الظهور في البحث بالذكاء الاصطناعي before you change any content.

يقود بيشوي عادل استراتيجية GEO والبحث بالذكاء الاصطناعي في VOCTOS، وهي وكالة تركّز على جعل العلامات التجارية مصدرًا تستشهد به محركات الإجابة بالذكاء الاصطناعي في أنحاء الشرق الأوسط ومصر.

أدلة ذات صلة

• أفضل 10 وكالات لتحسين البحث بالذكاء الاصطناعي للعلامات التجارية في المنطقة والعالم (2026)
• كيف تجعل Claude يستشهد بك: دليل 2026 العملي لاستشهادات البحث بالذكاء الاصطناعي
• روابط JavaScript تخفي مواقع الشرق الأوسط وشمال أفريقيا عن ChatGPT وClaude
• استشهادات البحث بالذكاء الاصطناعي تغيّرت للتو: ماذا يعني ذلك لعلامات الشرق الأوسط وشمال أفريقيا؟