If your ChatGPT referral traffic suddenly fell in Google Analytics this month, you are not alone, and you may not have lost a single visitor. From 9 October 2026, many businesses saw visits attributed to ChatGPT drop sharply while “Direct” traffic quietly rose. For a business that has started reporting AI traffic to the board, or deciding where to invest based on it, a chart like that can trigger the wrong decisions.
This guide explains what has been reported, how to check whether your drop is real or just a labelling problem, and how to set up your reporting so the next change does not catch you out.
- From around 9 October 2026, many sites saw ChatGPT-attributed sessions fall in GA4 while Direct traffic rose, which points to a tracking change rather than lost visitors.
- Reports say links inside ChatGPT answers stopped carrying the utm_source=chatgpt.com tag, so visits arrived unlabelled.
- Adobe’s Cedric Huesler reported that OpenAI confirmed a bug on 14 October and that a fix began rolling out on 15 October. Check your own data rather than assuming it is fixed.
- Annotate the affected dates, compare ChatGPT and Direct traffic page by page, and judge AI traffic on enquiries and sales, not sessions alone.
Why your ChatGPT referral traffic dropped: what has been reported
The numbers
According to Search Engine Journal, analytics firm Graphite saw ChatGPT-attributed sessions in Google Analytics 4 (GA4) fall by between 22% and 57% across the sites it tracks, starting on 9 October. Another monitoring tool, LLM Pulse, saw ChatGPT sessions around 46% lower over 9 to 11 October, while total traffic stayed roughly flat.
The most telling figure came from Adobe. Across around 500 pages that receive a lot of ChatGPT visits, Adobe’s data showed ChatGPT visits falling by about 31,000 a day while Direct visits rose by about 26,000 a day. That pattern, one channel down and another up by a similar amount, is the classic sign of visitors being relabelled rather than lost.
The likely cause
ChatGPT normally adds a tracking tag, utm_source=chatgpt.com, to the links it shows people. Analytics tools use that tag to credit the visit to ChatGPT. The reports suggest that, around the time OpenAI rolled out a new model and answer format in early October, links placed inside the answer text stopped carrying that tag. Links opened without both a tag and a referrer simply land in “Direct”.
Accounts of the exact mechanism differ slightly between researchers, and OpenAI’s public release notes did not mention it at the time of writing. Adobe’s Cedric Huesler said OpenAI confirmed to him on 14 October that it was a bug, and that tagged clicks were back to roughly normal levels on 15 October across the domains he tracks. Others had not yet seen the fix. Several of the companies quoted sell AI search tools, so treat individual percentages as indicative rather than exact.
How to check whether your drop is real
Work through these steps before you change any budget or strategy.
1. Mark the affected dates
Add an annotation in your analytics and reporting tools covering roughly 8 to 15 October 2026. Anyone reading your reports later will then know the dip was a known tracking issue, not a performance problem.
2. Compare ChatGPT and Direct traffic side by side
In GA4, open Reports > Acquisition > Traffic acquisition and compare the period from 9 October with the same number of days before it. Look at sessions from chatgpt.com and sessions in the Direct channel. If ChatGPT fell and Direct rose by a similar amount, your visitors are still arriving; they are just unlabelled.
3. Check at page level
Find the pages ChatGPT usually sends people to (often your homepage, service pages and in-depth guides). Use a landing page report filtered to those pages. If total visits to them held steady while the ChatGPT share fell, that confirms a labelling problem.
4. Look at what really matters: enquiries and sales
Check form submissions, calls, bookings and sales over the same period. If they are stable, the business impact is close to zero, whatever the channel report says.
5. Re-check after the fix
Compare the days after 15 October with the week before 9 October. If ChatGPT sessions have recovered, you can close the issue. If they have not, keep the annotation in place and keep watching, as rollouts can reach different users at different times.
Steps and menu names in GA4 and other tools change from time to time, so confirm the current steps in the platform’s official help centre, such as the Google Analytics Help Centre.
How to report AI traffic more reliably
This episode is a reminder that AI referral data is fragile. Most AI assistants, in every market from the UK and the UAE to the US, Europe and Asia, send some visits without any referrer at all, for example from mobile apps or copied links. A few changes make your reporting far more resilient.
Use an AI channel in your reports
GA4 has introduced an AI assistant channel that groups recognised AI referrers separately from ordinary referral traffic. Coverage of individual tools is not fully documented, so many businesses also create a custom channel group that matches the hostnames of the AI tools their customers use (for example chatgpt.com, gemini.google.com, claude.ai, perplexity.ai and copilot.microsoft.com). Place this rule above the Referral rule, as GA4 applies rules in order.
Count the tag as well as the referrer
Because many AI clicks arrive without a referrer, a report filtered on utm_source values such as chatgpt.com can catch visits that referrer-based reports miss. Google’s Gemini has also reportedly begun adding UTM tags to some links, though Google has not documented this.
Treat AI sessions as a minimum, not the full picture
Whatever your setup, some AI-driven visits will always show as Direct. Report AI traffic as “at least” a figure, and pair it with what people tell you. Adding “How did you hear about us?” to forms, with an AI assistant option, is a simple way to capture what analytics cannot.
Judge success on outcomes
Sessions are a means to an end. Tie your AI visibility work to enquiries, leads and revenue so that a tracking glitch never looks like a business crisis.
How to prevent the next false alarm
- Set alerts with context: when a channel drops sharply, check whether another channel rose by a similar amount before escalating.
- Keep a change log: note platform updates, site releases and tracking changes alongside your reports.
- Review tracking quarterly: test a real click from each major AI assistant and confirm how it appears in your analytics.
- Follow reliable sources: industry news and official release notes often explain a sudden drop before your own data does.
Frequently asked questions
Did my business lose visitors from ChatGPT in October 2026?
Probably not many. Reports show ChatGPT-attributed visits falling while Direct visits rose by a similar amount, which suggests visitors were relabelled rather than lost. Check your own Direct traffic and conversions to confirm.
What is utm_source=chatgpt.com?
It is a tracking tag ChatGPT adds to many of the links it shows. Analytics tools such as GA4 read it to credit the visit to ChatGPT. When the tag is missing and there is no referrer, the visit appears as Direct.
Has the ChatGPT tracking issue been fixed?
Adobe’s Cedric Huesler reported that OpenAI confirmed a bug on 14 October 2026 and that a fix began rolling out on 15 October. Not everyone had seen it at that point, so compare your own data for the days after 15 October with the week before 9 October.
Should I change my AI search strategy because of this drop?
No. A tracking problem is not a visibility problem. Confirm the drop is real, then judge your strategy on enquiries and sales over a longer period.
Not sure what your numbers are telling you?
If your reports are giving mixed signals, our Digital Growth Audit reviews your tracking, search visibility and conversion data so you can make decisions with confidence. Or get in touch and we will help you make sense of it.