Short answer: Because GPT-6 now opens outbound links without UTM tags and without a referrer, GA4 can only file those visits as direct traffic. An Adobe analyst tracking about 500 pages saw ChatGPT visits fall by roughly 31,000 a day between October 9 and 13 while direct visits rose by about 26,000. The visitors did not disappear; the attribution did. Checking landing pages, setting up an AI channel and asking on forms recovers most of the picture.
On October 14, Search Engine Roundtable reported that utm_source=chatgpt.com had vanished from ChatGPT's outbound links. Adobe's Cedric Huesler wrote on LinkedIn that GPT-6 opens links with noopener,noreferrer, so there is no tag and no referrer. Comparing about 500 pages where ChatGPT drove 5% or more of visits, he found ChatGPT visits down about 31,000 a day for October 9–13 versus the prior week, and direct visits up about 26,000 on the same pages. In one line: ChatGPT did not stop sending visitors; it stopped telling you it sent them.
Why now? OpenAI launched GPT-6 on October 7, saying ChatGPT has more than 1.2 billion weekly users and answers search-backed questions 44% sooner than before. Over the past 18 months AI assistants went from occasional referrers to a steady source for many B2B and local-service sites. Marketers had only just split AI traffic into its own GA4 channel, and Gemini began adding UTM tags to outbound links in early October. Suddenly the largest source on those new dashboards went dark. Browsers and in-app browsers have trimmed referrers for years; this time it hit the fastest-growing channel.
Platforms are diverging. Google's Gemini adds tags; Microsoft's Clarity separates AI platforms into their own channel and offers citation reports; Bing Webmaster Tools added a Topics Portfolio view to its AI performance report on October 14; paid tools such as Profound and Ahrefs measure whether AI answers mention you. The category is shifting from counting clicks to counting citations, because clicks are getting harder to measure.
For small businesses: if ChatGPT traffic fell while direct traffic rose, do not cut content or budgets yet. Below are the numbers, a tool comparison, and a no-subscription way to rebuild the data.
The details: where the missing traffic went
Bottom line: the data shows most of the lost ChatGPT traffic moved into direct, with matching behaviour.
- Timing: from about October 9, after GPT-6 rolled out; OpenAI has not said whether it was deliberate.
- Scale: across about 500 pages, ChatGPT visits fell about 31,000 a day and direct rose about 26,000; search fell only 7%, in line with other sources.
- Corroboration: direct traffic rose on two in three pages that lost ChatGPT visits, versus roughly a coin flip the week before.
- Behaviour: the share of direct visitors seeing a mobile consent banner, a proxy for new visitors, rose from 36% to 44%; for ChatGPT visitors it is 55%.
- Another signal: SEO consultant Sébastien Billard also saw human ChatGPT traffic drop from about October 9 while OAI-SearchBot crawling increased.
Caveat: this is one analyst's sample of about 500 pages, not an OpenAI statement. The direction is credible; the size will vary by site.
What three kinds of readers should do now
Bottom line: owners should avoid misreading the drop, marketers should change metrics, and developers should store attribution where they control it.
- Brand owners: in monthly reports, look at ChatGPT plus direct combined before and after October, rather than cutting content because one row fell.
- Marketers and SEOs: use three layers of AI metrics: citation counts (Bing, Clarity), direct traffic to likely AI landing pages, and self-reported source on forms.
- Developers and agencies: save first landing page and self-reported source on every form and CRM record; neither depends on platform changes.
AI traffic tool comparison
Bottom line: free tools are enough for small-business trend tracking; paid tools answer whether you are mentioned, not whether you were clicked.
| Tool | Measures | Price | Impact of this change |
|---|---|---|---|
| GA4 custom channel group | Clicks from AI sources | Free | High: no referrer, no classification |
| Microsoft Clarity AI channel | AI clicks plus citation reports | Free | Medium: clicks affected, citations not |
| Bing Webmaster Tools AI performance | Copilot/Bing citations and topics | Free | Low: citations, not clicks |
| Profound, Ahrefs and similar AI visibility tools | Brand mentions in AI answers | Paid subscription, see vendor plans | Low: prompt monitoring, not traffic |
What the vendors will not tell you
Bottom line: AI traffic was undercounted from the start; this change just makes it obvious.
- Apps rarely passed a referrer anyway: links opened inside mobile apps often lacked a referrer, so past ChatGPT numbers were already low estimates.
- Direct traffic gets misread as brand strength: a rise in direct visits is easy to celebrate as brand awareness when it may just be relabelled AI visitors.
- Paid monitoring has blind spots: prompt tracking measures fixed questions, which may not match how real customers ask, so good scores do not guarantee more enquiries.
A no-subscription alternative for small businesses
Bottom line: landing-page analysis, an AI channel rule and a form question cost close to nothing.
- In GA4, filter direct traffic to deep landing pages such as pricing, FAQ and articles; growth after October 9 is likely AI traffic.
- Follow the GA4 custom channel group guide to build an AI channel that still catches Gemini, Perplexity and Copilot.
- Add a single-choice how-did-you-hear question to enquiry forms and compare it with GA4 monthly.
FAQ
Has ChatGPT traffic really dropped?
Not necessarily. Most of it is being filed as direct traffic; the visitors still arrive, GA4 just cannot see the source.
Can GA4 still identify ChatGPT traffic?
For links opened from GPT-6, mostly no, because there is no UTM and no referrer; some older paths may still pass one.
Should we buy an AI visibility tool right away?
No rush. GA4, Clarity and a how-did-you-hear field on your forms recover most of the picture for free.
Is Bing's AI performance report useful?
Yes, but it counts citations, not clicks, so treat it as a trend indicator for AI visibility.
What should a small business do first?
Add a how-did-you-hear-about-us question to enquiry forms with ChatGPT, Google and AI summaries as options. It costs almost nothing.
My take
The mainstream reaction is to hunt for a new way to recapture ChatGPT clicks. My contrarian view: small businesses should stop betting AI performance on click attribution at all. The platform controls how links open; today it drops UTM, tomorrow it may switch to in-app previews, and you will always be chasing. What you can control is whether AI answers cite you and what customers say on your forms. The first lives in Bing and Clarity citation reports; the second is your own data.
That is the direction ScriptWalker (a Taiwan-based Laravel/Flutter custom development studio) has been taking with clients: website forms and CRM back offices store first landing page, UTM and self-reported source by default, and reports show which sources produce enquiries and deals, not just visits. For agencies, this is a sturdier service than reselling AI monitoring tools: build measurement into the client's own system instead of renting it on someone else's platform.
Sources
- Search Engine Roundtable: OpenAI Drops UTM Parameters From ChatGPT Results (GPT-6) (2026-10-14)
- Cedric Huesler (Adobe) LinkedIn analysis
- OpenAI: GPT-6 for everyone (2026-10-07)
- Microsoft Learn: Clarity AIPlatform and PaidAIPlatform channels
- Google Analytics Help: Custom channel groups
- Search Engine Land: Bing Webmaster Tools AI performance report adds Topics Portfolio