
Key takeaways
- From 9 October, many sites saw ChatGPT referrals fall by about half while Direct traffic rose on the same pages.
- The cause was ChatGPT links opening with no utm_source tag and no referrer after GPT-6 arrived. Adobe’s Cedric Huesler later called it a bug.
- Tagging was back by 15 October, but visits recorded as Direct during the gap won’t be relabelled.
- Annotate the gap in your reports, and track whether AI assistants cite you rather than relying on one referral number.
What happened, day by day
ChatGPT had been adding a utm_source=chatgpt.com tag to links in its answers. That small tag is how Google Analytics 4 and most other tools know a visitor came from ChatGPT. In October it went missing for about a week.
| Date (2026) | What happened | Source |
|---|---|---|
| 7-8 October | GPT-6 reached paid ChatGPT plans on 7 October and free plans on 8 October, with a more visual answer format OpenAI calls “Intelligent UI”. | LLM Pulse |
| 9-11 October | ChatGPT sessions across sites LLM Pulse monitors fell about 46% against the same days in the previous two weeks. | LLM Pulse |
| 14 October | Search Engine Roundtable reported that UTM parameters had disappeared from ChatGPT links. | Search Engine Roundtable |
| 15 October | Cedric Huesler of Adobe updated his analysis: “It was a bug. ChatGPT is tagging its links with utm_source=chatgpt(.)com again.” | Search Engine Roundtable |
As far as we can find, OpenAI hasn’t said anything publicly about the cause or the fix. The “bug” explanation comes from Huesler’s update, and Search Engine Roundtable’s Barry Schwartz also said it was likely a bug.
Why it looked like traffic had collapsed
Huesler said GPT-6 opened links in a way that stripped the tag and blocked the referrer. In his words, reported by Search Engine Roundtable: “There’s no tag and no referrer, so your analytics files the visit under direct.”
His numbers back that up. He looked at about 500 pages where ChatGPT drove at least 5% of visits and compared 9-13 October with the week before. ChatGPT visits fell by about 31,000 a day. Direct visits to the same pages rose by about 26,000 a day. Huesler called this an early signal, and the two figures don’t fully match, so some real loss can’t be ruled out.
LLM Pulse saw the same pattern from a different angle. Total traffic on the sites it monitors fell only about 2%, and Perplexity and Claude referrals showed no break on 9 October. Its analysis doesn’t say how many sites are in the sample, so treat the 46% as directional.
It’s fixed, so why does it still matter?
Because the damage to your history is permanent. LLM Pulse points out that even with tags back, the missing history won’t be recovered. Any report that compares October with September will show a ChatGPT dip that never really happened, and a Direct bump nobody can explain.
It also wasn’t a one-off. LLM Pulse found that Gemini referrals tagged “gemini” nearly disappeared between 28 September and 2 October, while visits from gemini.google.com held steady or grew. Its reading is that Gemini changed how it tags links. Two assistants changed how they label links in two weeks. If your AI traffic reporting depends on one parameter that the platform controls, it can break again without warning.
There’s one caveat worth keeping. LLM Pulse notes that richer answer formats may settle more questions inside the chat, which would mean fewer clicks overall. Analytics can’t measure that, so don’t assume every lost click came back.
What to check in your analytics this week
- Add an annotation in GA4 and any dashboards covering roughly 7-15 October, so nobody compares those days with normal weeks.
- Open your top ChatGPT landing pages, usually product, comparison and guide pages, and look for a Direct spike from 9 October. That’s the mislabelled traffic.
- Look at conversions and total visits before you react. If they held steady, don’t cut content budgets because one channel line dropped.
- Confirm the tag is back on your own site. LLM Pulse’s tracker showed a ChatGPT tagging rate of 97.53% on its page updated 25 October.
- Write down how you classify AI referrals (source, referrer, landing page rules) so the next change is quicker to spot.
If your site’s tracking setup is patchy, this is a good moment to tidy it. Clean event tracking and consistent landing pages are part of how we approach website development, and the same reporting can feed an automated weekly check through AI automation.
Our view: measure visibility, not just clicks
We think referral clicks are the wrong single number for AI search. They depend on how each assistant formats links, and that changes with every model release. A sturdier approach is to track a short list of real customer questions and check, every month, whether ChatGPT, Perplexity and Gemini mention or cite your business. LLM Pulse makes the same recommendation: measure visibility, citations and share of voice inside answers, which don’t rely on tags.
Then work on the content those assistants actually pull from. Clear answers, current facts and pages that are easy to quote matter more than chasing a referral line. Our AI content automation work builds that kind of publishing routine with human review before anything goes live, and our guide on getting cited in Google AI Overviews and AI Mode covers the Google side. If you’d like a second pair of eyes on your AI search reporting, book a strategy call.
Frequently asked questions
Did ChatGPT really send less traffic in October 2026?
Is ChatGPT tagging links again?
Can I fix the mislabelled data in Google Analytics?
What should a small business track for AI search instead?
Sources
- LLM Pulse: ChatGPT Referral Traffic Fell by Almost Half on October 9, 2026. Here’s What We Know
- Search Engine Roundtable: OpenAI Drops UTM Parameters From ChatGPT Results (GPT-6)
- Relevant Audience: ChatGPT appears to drop UTM tags and referrer, sending clicks to Direct
- LLM Pulse: ChatGPT referral tracking (tagging rate tracker)