If your analytics showed ChatGPT traffic falling off a cliff in the second week of October 2026, look at your Direct channel before you conclude anything. Several analysts reported the same pattern at the same time, and it looks like a tracking change rather than lost visitors.
What was reported
Search Engine Roundtable
(Barry Schwartz, 14 October, updated 15 October) and
Search Engine Journal
(Matt G. Southern, 15 October) describe the same change. After the GPT-6 rollout, links in
ChatGPT answers stopped carrying the utm_source=chatgpt.com tag and stopped passing a
referrer. A visit without a tag or a referrer is recorded as Direct, so ChatGPT traffic
disappears from the channel where you used to look for it.
The numbers behind the story come from several parties, each looking at different data:
| Source | What it saw |
|---|---|
| Graphite (Ethan Smith) | ChatGPT sessions in GA4 down 22% to 57% on sites with meaningful AI assistant traffic, from 9 October |
| LLM Pulse | ChatGPT sessions about 46% lower on 9 to 11 October than on the same days in the prior two weeks, with total site traffic roughly constant |
| Adobe (Cedric Huesler) | On about 500 pages where ChatGPT supplied at least 5% of visits, ChatGPT fell by about 31,000 visits a day while Direct rose by about 26,000 a day, and search fell 7% |
Huesler also found that Direct rose on two out of three pages that lost ChatGPT traffic.
How far to trust it
Treat this as strong circumstantial evidence, not proof. Aggregated data cannot separate traffic that was relabelled as Direct from traffic that was genuinely lost. LLM Pulse saw no clear rise in Direct, although Direct is so much larger than ChatGPT that a small shift would be hard to see. Huesler’s figures rest on four days of sampled data, and several of the sources sell SEO or AI search services, while Huesler works at Adobe, as Search Engine Journal points out.
What changed afterwards
According to Huesler’s later update, OpenAI confirmed a bug on 14 October and began rolling out a fix on 15 October. By that morning tagged clicks were back to roughly 95 to 100% of the same hour the week before, across the domains he tracks. Search Engine Journal notes that OpenAI’s release notes through 15 October did not mention it.
One practical point stays after the fix: most ChatGPT clicks still arrive without a referrer. The UTM tag is what lets you count them, so the tag is the number to watch, not the referrer.
What to check in your own analytics
- Annotate 8 to 15 October in your reports before you read the drop as lost demand.
- Compare ChatGPT and Direct side by side, per page ChatGPT used to send visitors to, and for the homepage. A fall in one with a matching rise in the other on the same pages points to relabelling.
- Do not compare October with September without that annotation. A month-on-month comparison that includes the gap will understate your AI traffic.
- Leave links with your own UTM parameters alone. Graphite’s Ethan Smith says links from ChatGPT Ads that carry your own parameters should be unaffected.
What this means for how you measure AI visibility
This is the weakness of measuring AI search through sessions alone: the number depends on a tag that one company can change overnight. That is why we measure visibility from the answers themselves, which prompts name you, at what position and which sources the answer draws on, against a baseline frozen before the work starts. Sessions remain useful, as the second half of the picture. When one of the two moves for a reason that is not about your site, the other tells you so, and the report says it.
Frequently asked questions
Did ChatGPT traffic really drop in October 2026? Many sites saw ChatGPT-attributed sessions fall sharply from around 8 or 9 October, but the evidence points to a tracking change. The links lost their UTM tag and referrer, so the visits were filed as Direct.
Is the problem fixed? Huesler reports that OpenAI confirmed a bug on 14 October and began rolling out a fix on 15 October. Most ChatGPT clicks still arrive without a referrer, so count the UTM tag.
How do I check whether my traffic was relabelled? Annotate 8 to 15 October, then compare ChatGPT and Direct sessions side by side per page. A drop in one with a matching rise in the other on the same pages points to relabelling.
