ChatGPT Referral Traffic Fell 46%: What Gulf Brands Should Check

Last updated: 19 October 2026. By Beshoy Adel, GEO and AI search strategist at VOCTOS.

ChatGPT’s reported traffic fell about 46% in October, mostly through lost tracking. After GPT-6 launched on 8 October, links stopped carrying the utm_source=chatgpt.com tag, so GA4 filed many ChatGPT visits as Direct. Adobe’s Cedric Huesler found ChatGPT down 31,000 visits a day and Direct up 26,000 on the same pages. Tagging recovered by 16 October, but the gap stays in your October reports.

أبرز النقاط

These are the five numbers worth carrying into your next client or board report. Each one comes with the source that published it and the period it covers. You can quote them with confidence.

• LLM Pulse measured about 46% fewer ChatGPT sessions in GA4 from 9 to 11 October 2026, against the same days in the prior two weeks, while total traffic on those sites fell only about 2%, which says the sites did not lose their audience.
• Adobe’s Cedric Huesler found ChatGPT visits down roughly 31,000 a day and Direct up roughly 26,000 a day across about 500 pages (9 to 13 October against the week before), and Direct rose on two in three of the pages that lost ChatGPT visits, which is the pattern you would expect from a relabel.
• Profound saw tagged source links fall from about 99% of cited ChatGPT answers on 8 October to about 5% at 19:30 UTC on 14 October, then return to about 99% by 06:30 UTC on 16 October, a swing that took roughly eight days from the first drop to the full recovery.
• Microsoft’s AI Economy Institute puts UAE generative AI use at 73.3% of the working-age population in the second quarter of 2026, as reported by Middle East AI News.
• Saudi Arabia’s CST reports that 45.2% of internet users use AI tools, more than double the year before, and Statcounter gives ChatGPT 75% of Saudi AI chatbot page views in September 2026, so a ChatGPT reporting error there touches most Saudi AI chatbot use.

What happened to ChatGPT referral traffic in October 2026?

GPT-6 reached paid ChatGPT plans on 7 October and free plans on 8 October. Soon after, links stopped carrying the utm_source=chatgpt.com tag and the referrer header, so GA4 counted many ChatGPT clicks as Direct. Search Engine Roundtable published the first report on 14 October, and by 15 October tagging had begun returning for some of you.

Barry Schwartz of Search Engine Roundtable noted that OpenAI had added UTM tracking to ChatGPT links in June 2026, so the October change looked like a regression to analysts. Huesler traced it to links that open with noopener and noreferrer, which strips the referrer. With no tag and no referrer, analytics has nothing to classify. The visit lands in Direct, next to bookmarks and typed URLs. A bug in a tag can look like a collapse in demand. Nothing about the reader changed. Only the label did.

Did ChatGPT stop sending visitors, or did analytics stop labelling them?

The evidence leans toward lost labels, not lost readers. That is the short version. Caveats follow. On about 500 pages, Huesler measured ChatGPT visits down 31,000 a day while Direct rose 26,000 a day. Direct grew on two in three pages that lost ChatGPT traffic. Search fell 7% on those same pages, so the shift is not a pure relabel.

The authors name their own limits, and you should repeat them whenever you quote the numbers. Huesler’s sample covers pages where ChatGPT already drove at least 5% of visits, so it over-represents sites that depend on ChatGPT. The Direct attribution is an inference from missing referrer data, not a confirmed measurement. LLM Pulse adds that GPT-6’s new interface may resolve more questions inside the chat, which would cut real clicks. Analytics alone cannot measure that. At site level ChatGPT is small next to Direct, so the Direct rise is hard to see in a total.

How long did the ChatGPT tracking gap last?

Profound’s 30-minute data shows two separate stages of failure. Tagged links fell from about 99% of cited answers on 8 October to about 75% across 9 October. They then collapsed to about 5% at 19:30 UTC on 14 October, as the new interface reached about 96% of answers. Tagging returned in stages and hit about 99% by 06:30 UTC on 16 October.

Two warnings come with the recovery, and both affect how you write the October report. Restored tags do not rewrite history. So sessions counted as Direct between 8 and 15 October stay Direct. OpenAI has also not published a statement, because the bug confirmation is Huesler’s account of what OpenAI told him on 14 October, and Relevant Audience found nothing about it in the release notes for 1 to 15 October.

The numbers to trust and the numbers to discount

Trust per-page comparisons that put ChatGPT and Direct side by side for you. Treat your site-level ChatGPT totals for 9 to 15 October as understated, and say so in writing. Treat vendor percentages as directional, because LLM Pulse and Profound do not publish their sample sizes. The table shows what each source measured and where it falls short. Read the last column first. It matters most.

What does the ChatGPT tracking gap mean for brands in Riyadh, Dubai and Cairo?

It means your October reports understate ChatGPT in the markets where ChatGPT matters most. Statcounter gives ChatGPT 75% of Saudi AI chatbot page views in September 2026, and Microsoft’s AI Economy Institute reports UAE generative AI use at 73.3% of working-age people. A half-sized ChatGPT line in a Gulf report needs checking before anyone acts on it.

The Saudi picture is the clearest one we have for this episode. The Communications, Space and Technology Commission reports that 45.2% of Saudi internet users use AI tools, more than double the year before, and Asharq Al-Awsat adds that ChatGPT ranked first among the most-downloaded AI apps in the Kingdom. If you sell in Riyadh, ChatGPT is a real discovery channel for your buyers. A reporting hole there is your problem. Many Gulf and Egyptian brands also close sales on WhatsApp, by phone or through a form on an Arabic landing page, and those journeys already leave weak trails in GA4. That makes a quiet move into Direct harder for your team to spot.

I have no verified Egypt-specific figure for this episode, and I will not invent one, but the check still applies to your Cairo-based site. If your sales team asks every lead how they found you, those October answers are now more reliable than your GA4 channel table, and the check is cheap enough for any team to run this week, even one without an analyst. For the wider measurement setup, our guide on قياس زيارات البحث بالذكاء الاصطناعي في الشرق الأوسط وشمال أفريقيا covers the channel rules that make this kind of break easier to catch.

My read: protect the trend line, not the dashboard

I read four sources on this episode, and none of them closes the question. What I take from it is smaller than the headlines suggest. A tracking tag changed. A fix arrived within about a week. That was fast. Plenty of people drew big conclusions from the hole in between. I expect you have already heard at least one of them.

I would not rebuild an attribution model around a seven-day gap, and I would not ask you to either. I would annotate 8 to 15 October in GA4 and move on. I also want to be clear about one trap. It is an easy one. The people who read a 46% dip as proof that AI search is fading will read the recovery as proof that it is booming, and both groups are treating a label as behaviour instead of reading what readers actually did. I have seen the same pattern with every analytics change that touches a referrer. It never ends well. My bet, and it is only a bet, is that GPT-6’s new answer format will cut some real clicks too. We will not separate that effect for weeks.

That is why I keep telling teams that clicks are the wrong unit for AI search, and why we now treat citations as the unit you can defend. Our منهجيتنا في قياس الظهور في الذكاء الاصطناعي tracks whether a brand is named and linked in answers, which does not depend on a UTM tag surviving. Clicks stay in your report, but as one signal among several, and we say so to every client.

Five steps to correct your October ChatGPT reporting in GA4

Annotate the break, rebuild your comparison at page level, and judge your month on conversions. Do not edit history. Do not reassign Direct traffic to ChatGPT by guesswork. These five steps follow the recommendations from LLM Pulse and Relevant Audience, and each one takes you minutes, not days.

1. Annotate 8 to 15 October in GA4, remembering that Huesler’s times are Pacific, so you should shift the window to your property’s time zone.
2. Match on the chatgpt.com source value, not on the referrer, because the referrer is still missing on most ChatGPT clicks.
3. Compare ChatGPT and Direct together on your top ChatGPT landing pages, this week against the prior week.
4. Judge the month on total visits and conversions, not on the ChatGPT line alone.
5. Track citations and mentions separately from clicks, and read our piece on the AI search citation shift in MENA for how to interpret them in the Gulf.

الأسئلة الشائعة

These are the questions teams have sent us since 9 October, written the way they arrive in our inbox. Each answer is short and points back to the evidence above, and where that evidence is thin we say so. If your question is missing, send it to us and we will answer it.

My ChatGPT traffic fell by half in October and my manager wants an explanation.

Treat it as a labelling failure until proven otherwise. Do not panic. LLM Pulse measured about 46% fewer ChatGPT sessions from 9 to 11 October, while Huesler saw Direct rise on two in three affected pages. Show your manager both lines together with the 8 to 15 October annotation, and say plainly that some real click loss may sit underneath, because nobody can rule that out until GPT-6’s new interface has run for a few more weeks.

My Direct traffic jumped and I do not know why.

Check your top ChatGPT landing pages first, because Huesler found Direct up roughly 26,000 visits a day on pages that lost ChatGPT traffic. If your Direct rise sits on the same URLs and the same dates, the likely cause is stripped tags. If it sits elsewhere, look at your email, campaigns and other channels.

My ChatGPT traffic came back by itself, so I assume the October data is fixed.

No, the fix is not retroactive. Profound noted that restored tags improve reporting for new visits but do not recover attribution for earlier ones. Sessions that GA4 counted as Direct from 8 to 15 October stay Direct. Annotate the period and compare month on month with that gap in mind.

Gemini and Perplexity changed too, and now I do not know which channel numbers to trust.

LLM Pulse saw no break for Perplexity or Claude on 9 October. It did see a separate Gemini issue between 28 September and 2 October, when visits tagged “gemini” nearly disappeared while gemini.google.com traffic held. Keep channel rules for both source values, and review each AI source against its own dates, since the breaks did not happen together.

Our leads arrive on WhatsApp and by phone, so I assume none of this matters to us.

It matters more, not less. Off-site journeys already leave thin trails. A Direct bucket that swallows ChatGPT visits hides the top of your funnel. Ask every lead how they found you for the next month, because those answers will tell you more than the October channel table.

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

VOCTOS is the specialist we would put forward, and we are obviously not neutral. The honest answer depends on your market, your budget and whether you need Arabic search, English search or both. Look at دراسات الحالة لدينا, ask any agency how it measures citations, and be wary of anyone who guarantees you a ranking.

The better question is what you would do if ChatGPT clicks never came back

The useful question is not whether October’s dip was real. It is whether your reporting would still tell you the truth if referral data kept degrading. AI platforms change links, referrers and interfaces without notice. Brands that measure citations, conversions and sales conversations can absorb a bug like this one. Brands that depend on a single tag cannot.

If you already track ChatGPT in GA4, start with the five steps above and spend one afternoon on your top landing pages. No AI search reporting at all? Build it now, before the next change, and we can help you do it. Our دليلنا لتحسين الظهور في بحث الذكاء الاصطناعي shows where to begin, and the Google AI Mode guide for Gulf brands covers the Arabic side of the same problem.

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

أدلة ذات صلة

• إعدادات Cloudflare الافتراضية لزواحف الذكاء الاصطناعي في 15 سبتمبر: الفحص الذي ينبغي أن تُجريه العلامات التجارية في الخليج ومصر اليوم
• وضع الذكاء الاصطناعي من Google بالعربية: ما الذي ينبغي أن تغيّره العلامات التجارية في الخليج ومصر الآن
• أداة Google لاستبعاد المحتوى من الذكاء الاصطناعي: ماذا تفعل العلامات في الخليج ومصر؟
• إعلانات ChatGPT متاحة الآن في 13 سوقًا عربيًا. وإليك ما لا يعمل حتى الآن