ChatGPT fixed the UTM issue!
A quick recap. On October 9, ChatGPT switched to GPT-6. After that, many links to websites stopped carrying utm_source, and referrer tags were gone so Google Analytics counted those visits as Direct.
As a result, reported traffic from ChatGPT to websites dropped by 37% (based on a panel of a few hundred domains).
The traffic wasn’t gone, only misattributed.
The fix went live yesterday: late evening in Europe, late afternoon in the US.
Starting today, reported traffic from ChatGPT to websites should go back to normal.
I’m not mistaken, based on the data I see, many brands should see higher LLM referral traffic from today onward than they did in the weeks before October 7. We’ll see.
One practical note: when you analyze LLM traffic in tools like GA4 or Peec AI, treat October 9-15 as a reporting gap. Annotate it in GA4, and compare it with your Direct traffic for the same dates
Thanks to everyone who covered this: Barry Schwartz, Cedric Huesler, Lily Ray, Ethan Smith, and many more. At Peec AI we also notified OpenAI employees about this issue. I can’t say what exactly made the difference, but it shows that when our industry speaks with one voice, even companies as big as OpenAI listen.
Next on my wish list: Google adding special markers, so we can analyze LLM traffic better.
John Mueller Daniel Waisberg 🤞 🤞
A quick recap. On October 9, ChatGPT switched to GPT-6. After that, many links to websites stopped carrying utm_source, and referrer tags were gone so Google Analytics counted those visits as Direct.
As a result, reported traffic from ChatGPT to websites dropped by 37% (based on a panel of a few hundred domains).
The traffic wasn’t gone, only misattributed.
The fix went live yesterday: late evening in Europe, late afternoon in the US.
Starting today, reported traffic from ChatGPT to websites should go back to normal.
I’m not mistaken, based on the data I see, many brands should see higher LLM referral traffic from today onward than they did in the weeks before October 7. We’ll see.
One practical note: when you analyze LLM traffic in tools like GA4 or Peec AI, treat October 9-15 as a reporting gap. Annotate it in GA4, and compare it with your Direct traffic for the same dates
Thanks to everyone who covered this: Barry Schwartz, Cedric Huesler, Lily Ray, Ethan Smith, and many more. At Peec AI we also notified OpenAI employees about this issue. I can’t say what exactly made the difference, but it shows that when our industry speaks with one voice, even companies as big as OpenAI listen.
Next on my wish list: Google adding special markers, so we can analyze LLM traffic better.
John Mueller Daniel Waisberg 🤞 🤞
As posted by Tomek Rudzki on LinkedIn.