If you looked at your website stats this month and saw traffic from ChatGPT fall off a cliff while your direct traffic jumped, here is what actually happened: you did not lose customers. You lost the label.
What changed
Since June 2026, links from ChatGPT have carried a small tracking tag, a piece of text called a UTM parameter that tells your analytics "this visitor came from ChatGPT." On October 8, shortly after the GPT-6 release, that tag quietly disappeared.
Adobe analyst Cedric Huesler spotted it and ran the numbers. He looked at about 500 pages where ChatGPT was driving at least 5 percent of visits and compared October 9 to 13 against the week before. Traffic labeled as coming from ChatGPT fell by roughly 31,000 visits per day. On those same pages, direct traffic rose by roughly 26,000 visits per day. Two thirds of the pages that lost the ChatGPT label gained it back as direct traffic.
Read that again. The visitors kept coming. They were just showing up under a different name.
Huesler asked OpenAI about it on October 14 and was told it was a known bug. A fix started rolling out on October 15, and by that morning tagged ChatGPT clicks were back to roughly normal across the domains Adobe tracks. OpenAI has not published a statement of its own, so treat the week of October 8 to 15 as a measurement gap, not a traffic problem.
Why this matters to your business
If you only glance at one line in your reports, that week lies to you. It looks like AI discovery stopped working for your business. It did not. The assistants kept sending people to websites. The tracking broke for a week.
This is the part most business owners miss: when a number moves in your analytics, the first question is never "what do I do." It is "is this number still measuring what I think it measures." For ChatGPT referrals in early October, the answer was no.
What to do about it
- Mark the gap. If you use Google Analytics, add an annotation for October 8 to 15, 2026, so nobody reads that dip as a real drop six months from now.
- Look at three numbers together, not one. Referral traffic, where ChatGPT visits normally show up. Direct traffic, where they hid that week. And landing page performance, which tells you whether real people are still arriving and doing something.
- Count the tag, not the referrer. Even after the fix, most ChatGPT clicks still arrive without a referrer. The UTM tag is the reliable signal, so make sure your reports are built on it.
If direct went up that week and your landing pages stayed steady, your AI visibility is fine. The label moved, not the customers.
The bigger point
AI assistants are becoming a real source of website visits for local businesses. People ask ChatGPT or Claude for a plumber, a detailer, an insulation contractor, and they click through. That trend did not reverse because a tracking tag went missing for a week.
But it does mean guessing is over. You cannot judge whether AI search is working for you by glancing at one chart. You have to measure it on purpose, the same way you would measure anything you spend money on.
That is exactly what my free SEO and AI Search Checker does. It scores how your business shows up in AI search, in under a minute, with real Google mobile speed data. Run your site through it and see where you actually stand instead of guessing.
Sources
- Search Engine Roundtable: OpenAI drops UTM parameters from ChatGPT results
- Cedric Huesler's follow-up: the UTM tag loss was a bug and a fix is rolling out
Traffic numbers are Huesler's measurements across roughly 500 pages, October 9 to 13 versus the prior week.
Where to go next
More posts on the blog. If you want to know whether Google and AI assistants can currently find your business, there is a free ten minute check. Every price is on the pricing page.
Send me your website
I will look at it and tell you what I find, free, whether or not you hire me.
Mándame tu sitio y te digo qué encontré. Hablamos español.
Get a free look