{"id":13280,"date":"2026-10-15T09:16:16","date_gmt":"2026-10-15T07:16:16","guid":{"rendered":"https:\/\/voctos.net\/chatgpt-direct-traffic-ga4-arab-websites\/"},"modified":"2026-10-18T10:33:10","modified_gmt":"2026-10-18T08:33:10","slug":"chatgpt-direct-traffic-ga4-arab-websites","status":"publish","type":"post","link":"https:\/\/voctos.net\/ar\/chatgpt-direct-traffic-ga4-arab-websites\/","title":{"rendered":"\u0632\u064a\u0627\u0631\u0627\u062a ChatGPT \u062a\u0638\u0647\u0631 \u0627\u0644\u0622\u0646 \u0643\u0632\u064a\u0627\u0631\u0627\u062a \u0645\u0628\u0627\u0634\u0631\u0629 \u0641\u064a GA4: \u0643\u064a\u0641 \u062a\u0638\u0644 \u0627\u0644\u0645\u0648\u0627\u0642\u0639 \u0627\u0644\u0639\u0631\u0628\u064a\u0629 \u0642\u0627\u062f\u0631\u0629 \u0639\u0644\u0649 \u0642\u064a\u0627\u0633\u0647\u0627"},"content":{"rendered":"

Update, 18 October 2026:<\/strong> the missing UTM tag described below appears to have been a bug, not a permanent change. Adobe\u2019s Cedric Huesler reports that OpenAI told him on 14 October it was a known bug; the tag stopped around 8 October with the GPT-6 rollout, and a staged fix began on 15 October, after which tagged ChatGPT clicks were back to roughly 95\u2013100% of the same hour a week earlier in Adobe\u2019s data. This is second-hand: OpenAI has not commented publicly and its release notes do not mention it, and we have not yet re-run our 18-link test after the fix. Sessions that GA4 recorded as Direct between 8 and 15 October are not restated, so annotate that window in GA4 (it starts on the evening of 8 October in Riyadh and Cairo) and count ChatGPT by the chatgpt.com<\/code> source value. Source: Relevant Audience<\/a>.<\/p>\n

ChatGPT links lost their tracking tag, so GA4 now files the visits as Direct. In our 15 October 2026 test, none of 18 answer links carried a UTM parameter and all 18 told the browser to send no referrer. Adobe’s Cedric Huesler reports about 31,000 fewer daily ChatGPT visits and 26,000 more direct visits across roughly 500 pages.<\/p>\n

That matters for any Arab site that reports ChatGPT as a channel. The row in your GA4 report can fall while the people keep arriving. Below is what changed, what we saw in ChatGPT’s own links, and a five-step way to estimate the visits GA4 can no longer name. We did not have a client before and after comparison, and we say so where it counts.<\/p>\n

\n
\n
0 of 18<\/div>\n
ChatGPT answer links carrying a UTM parameter<\/div>\n
VOCTOS test, 15 Oct 2026<\/div>\n<\/div>\n
\n
18 of 18<\/div>\n
links set to open with noopener and noreferrer<\/div>\n
VOCTOS test, 15 Oct 2026<\/div>\n<\/div>\n
\n
2 in 3<\/div>\n
pages that lost ChatGPT visits gained direct visits<\/div>\n
Adobe analysis, 9 to 13 Oct 2026<\/div>\n<\/div>\n<\/div>\n

What changed in ChatGPT links in the week of 7 October 2026?<\/h2>\n

OpenAI released GPT-6 as the ChatGPT default from 7 October 2026, and about a week later analysts reported that its links stopped carrying a UTM tag. Search Engine Roundtable published the report on 14 October. It says the tag appeared in June 2025 and is now gone, and that nobody yet knows whether the change was deliberate.<\/p>\n

Huesler, who works at Adobe, wrote on LinkedIn that GPT-6 opens links with window.open<\/code> using the noopener<\/code> and noreferrer<\/code> options. That drops both the referrer and the utm_source=chatgpt.com<\/code> tag, so analytics records a direct visit. His comparison of 9 to 13 October with the prior week found ChatGPT visits down about 31,000 a day and direct visits up about 26,000 a day on the same pages, while search traffic fell 7%.<\/p>\n

Direct traffic rose on two in three of the pages that lost ChatGPT visits. On mobile, the share of new direct visitors who saw a consent banner rose from 36% to 44%, close to the 55% seen for ChatGPT visitors. Huesler concluded the new direct visitors behave like ChatGPT visitors. S\u00e9bastien Billard separately reported a sharp fall in human ChatGPT visits since 9 October. Neither report names the countries in its data. Read the originals at Search Engine Roundtable<\/a>, Huesler’s LinkedIn post<\/a> and OpenAI’s ChatGPT release notes<\/a>.<\/p>\n

What did we find in ChatGPT answer links on 15 October 2026?<\/h2>\n

We asked ChatGPT three web-search questions in a signed-out browser session on 15 October 2026 and read every external link in the answers. Across 18 links to 13 hosts, none carried a UTM parameter and every one had rel=\"noopener noreferrer\"<\/code> with target=\"_blank\"<\/code>. The links therefore ask the browser to hide where the click came from.<\/p>\n\n\n\n\n\n\n\n\n
Prompt (15 Oct 2026)<\/th>\nLanguage<\/th>\nExternal links<\/th>\nWith UTM<\/th>\nnoreferrer<\/th>\n<\/tr>\n<\/thead>\n
Official Saudi sites to check a commercial registration and file a consumer complaint<\/td>\nEnglish<\/td>\n3<\/td>\n0<\/td>\n3<\/td>\n<\/tr>\n
Best places to buy electronics in Saudi Arabia, with links<\/td>\nGulf Arabic<\/td>\n8<\/td>\n0<\/td>\n8<\/td>\n<\/tr>\n
Where to buy a laptop in Egypt and at what price, with links<\/td>\nEgyptian Arabic<\/td>\n7<\/td>\n0<\/td>\n7<\/td>\n<\/tr>\n
Total<\/strong><\/td>\n<\/td>\n18 (15 unique)<\/strong><\/td>\n0<\/strong><\/td>\n18<\/strong><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

The limits are real. This was one signed-out Free session in one desktop browser on one day, with links read from the page and not followed into GA4. A pop-up blocker stopped us from opening a link to read the landing referrer, so the noreferrer<\/code> result rests on the attribute, which browsers honour by specification. We did not test the mobile app or a paid account. Our 8 October GA4 guide<\/a> had already found the same noreferrer<\/code> attribute in a desktop test, so the Direct outcome was predictable.<\/p>\n

Why does a ChatGPT click with no referrer land in GA4 as Direct?<\/h2>\n

GA4 puts a session in Direct when the source is exactly “(direct)” and the medium is “(not set)” or “(none)”. A ChatGPT click with no referrer and no UTM tag matches that rule, so GA4’s AI Assistant channel never sees it. Google’s own wording describes Direct as bookmarks and typed URLs, which is no longer what the bucket holds.<\/p>\n

The rule is in Google’s default channel group definitions<\/a>. For Arab sites the damage depends on traffic mix. Our earlier guide<\/a> measured mobile at 68.27% of web traffic in Saudi Arabia in September 2026, and mobile app and in-app browser clicks lose referrers more often than desktop ones. Sites on strict consent settings also lose cookie-based identity on first visits, which we cover in our consent mode guide<\/a>.<\/p>\n

How can an Arab site estimate the ChatGPT visits GA4 no longer names?<\/h2>\n

You cannot recover the individual visits, but you can bracket them with five checks that run on data you already own. Compare a Direct deep-page segment before and after 7 October, read AI agent hits in server logs, ask visitors directly, keep the AI Assistant row as a floor, and report the two numbers separately.<\/p>\n

    \n
  1. Annotate the dates.<\/strong> Mark 7 October 2026 (GPT-6 rollout) and 9 October (first reported drop) in GA4 so nobody reads the shift as a campaign effect.<\/li>\n
  2. Build a deep-page Direct segment.<\/strong> In Explore, filter Default channel group to Direct, landing page not the home page, and new users. Real bookmarks mostly hit the home page, so a jump on article and product URLs is the signature Huesler describes.<\/li>\n
  3. Read the server logs.<\/strong> Count fetches by ChatGPT-User and OAI-SearchBot. They show OpenAI retrieving your pages, not people clicking, but a rise in both next to a rising deep-page Direct segment is consistent. Check that your robots.txt<\/a> lets them in.<\/li>\n
  4. Ask the visitor.<\/strong> Add “ChatGPT or another AI assistant” to the “how did you hear about us” field on lead forms, and to the first WhatsApp reply script. For clinics and service businesses that sell through WhatsApp, this is the only first-party count.<\/li>\n
  5. Report two numbers.<\/strong> Keep the AI Assistant channel as the measured floor and show the estimated direct uplift as a separate, labelled range.<\/li>\n<\/ol>\n

    Check how my GA4 handles ChatGPT visits<\/a><\/p>\n\n\n\n\n\n\n\n\n\n
    Signal<\/th>\nWhere to read it<\/th>\nWhat it tells you<\/th>\nWeakness<\/th>\n<\/tr>\n<\/thead>\n
    Deep-page Direct, new users<\/td>\nGA4 Explore<\/td>\nLikely referrer-stripped clicks, ChatGPT among them<\/td>\nAlso catches other stripped sources and dark social<\/td>\n<\/tr>\n
    ChatGPT-User and OAI-SearchBot hits<\/td>\nServer or CDN logs<\/td>\nOpenAI is fetching your pages<\/td>\nFetches are not human visits<\/td>\n<\/tr>\n
    “How did you hear” answers<\/td>\nLead form, WhatsApp script<\/td>\nSelf-reported ChatGPT referrals<\/td>\nSmall samples, recall bias<\/td>\n<\/tr>\n
    AI Assistant channel<\/td>\nGA4 default report<\/td>\nVisits that kept a referrer<\/td>\nNow a floor, not the total<\/td>\n<\/tr>\n
    Search Console AI report<\/td>\nSearch Console<\/td>\nGoogle AI impressions only<\/td>\nNo clicks and no ChatGPT data<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    What does this change in our GA4 guidance for Arab sites?<\/h2>\n

    Our 8 October guide warned that ChatGPT clicks without a referrer land in Direct and that UTM-tagged links could become Unassigned. The second branch is now the minority case, because the tag is gone from the links we read. The advice to keep an AI channel group stands, but its label should read “measured AI visits”, never “all AI visits”.<\/p>\n

    We have no Arab before and after data for this change, and the published analyses do not name their countries. Behaviour on Arabic prompts matched English ones in our small test, which is a reason to expect the same pattern, not proof of the same size. We will pull a client comparison once two clean weeks exist, and say what it shows.<\/p>\n

    Our read<\/h2>\n

    Do not cut AI-search work because the ChatGPT row in GA4 fell this week. The visits moved to Direct, and the people who clicked are the same people. Equally, do not book the whole Direct jump as ChatGPT. The honest report is a floor, a range and the method behind both. A team that shows a flat line and a shrug will lose the argument to one that shows a bracket.<\/p>\n

    FAQ<\/h2>\n

    My ChatGPT traffic in GA4 dropped. Did the visits really stop?<\/h3>\n

    Not necessarily. Adobe’s analysis of about 500 pages found ChatGPT visits down 31,000 a day and direct visits up 26,000 a day over 9 to 13 October 2026. The visits likely moved to Direct. Check a deep-page Direct segment before you report a loss, and compare it against the week before 7 October.<\/p>\n

    Can I add UTM parameters to my own pages to fix this?<\/h3>\n

    No. The tag was added by ChatGPT to the outbound link, and your site only receives what the browser sends. You can tag links you control, such as your own ChatGPT Ads campaigns, but organic answer links are outside your control. Use landing-page segments, logs and a form field instead.<\/p>\n

    Is this the same problem as Direct traffic from WhatsApp?<\/h3>\n

    The effect is the same, because both arrive without a referrer, but the cause differs. In-app browsers and messaging apps strip referrers by design, while ChatGPT now sets the no-referrer option on its links. Both inflate Direct on Arab sites, so a Direct jump alone cannot name its source.<\/p>\n

    Should I stop reporting an AI Assistant channel in GA4?<\/h3>\n

    No. Keep it, and relabel it as measured AI visits. It still catches clicks that keep a referrer, including other assistants. Add the estimated direct uplift beside it as a separate range, and note the method, so the two numbers are never summed without context.<\/p>\n

    Does the change apply to Arabic ChatGPT answers?<\/h3>\n

    In our 15 October test it did. Of 18 links across English, Gulf Arabic and Egyptian Arabic prompts, none carried a UTM parameter and all 18 used the no-referrer option. That is a small signed-out sample, so we treat it as consistent with the global reports, not as a measured Arab traffic loss.<\/p>\n

    What is the better question than “where did ChatGPT traffic go”?<\/h2>\n

    The better question is how much of your Direct traffic is now AI-referred, and whether the pages it lands on are the ones you want cited. If you have lead forms or WhatsApp intake, start the “how did you hear” field today and compare it with the deep-page segment in two weeks. If you have neither, fix the intake first, because no GA4 setting can recover a referrer the browser never sent.<\/p>\n

    Help me set up AI traffic reporting for my site<\/a><\/p>\n