ChatGPT Referral Traffic Dip? Fix Your GA4 Reports
If your ChatGPT referral traffic dropped in mid-October while direct traffic climbed, your audience probably didn’t vanish. On October 20, Search Engine Land reported that ChatGPT links had stopped carrying the utm_source=chatgpt.com tag, and that OpenAI confirmed it was a known bug. A fix started rolling out on October 15. The visits were still arriving. Analytics tools just couldn’t tell where they came from. Here’s what happened, why it matters for anyone reporting on AI search, and a six-step GA4 audit to clean up your October numbers before they reach a client deck.
What happened to ChatGPT referral tracking
ChatGPT has been adding a UTM source tag to outbound links, so a click from a ChatGPT answer shows up in analytics as coming from chatgpt.com. After the GPT-6 release, those tags went missing. Adobe’s Cedric Huesler also reported that links were being opened in a way that suppresses referrer information. With neither a UTM tag nor a referrer, analytics platforms have nothing to go on, so they file the visit under direct traffic.
Huesler looked at roughly 500 pages where ChatGPT had previously driven at least 5% of visits and compared October 9 to 13 with the week before. On those pages, ChatGPT traffic fell by about 31,000 visits a day while direct traffic rose by about 26,000 a day. Two-thirds of the pages that lost ChatGPT referrals saw direct traffic go up, and the new “direct” visitors behaved like the ChatGPT visitors who came before them. Search traffic on the same pages was down 7%, which he described as in line with wider search trends.
Huesler contacted OpenAI on October 14. OpenAI said it was a known bug, and the fix began rolling out early on October 15. By 9 a.m. Pacific time that day, tagged ChatGPT clicks were back to about 95% to 100% of the level seen at the same hour a week earlier, based on telemetry across thousands of domains.
Why this matters beyond one bad week
The bug is mostly fixed, but it exposed something agencies should take seriously. Most ChatGPT clicks still don’t pass referrer data, according to Huesler, so the UTM tag is doing nearly all the attribution work. When a platform changes how it opens links, AI traffic can quietly move into another bucket overnight, and nobody gets an alert.
That has real consequences. A client who sees “ChatGPT traffic down sharply” might cut budget on the content, PR and brand work that earns AI mentions. A team might switch attribution models to “fix” a dip that was really a labeling problem. And month-over-month AI visibility reports for October will be wrong unless someone explains the gap. If you already track AI visibility, as we covered in our guide to measuring AI visibility, referral data is the part that ties those mentions to visits and leads.
A six-step GA4 audit for ChatGPT referral traffic
1. Compare ChatGPT and direct traffic on the same pages
Open the Traffic acquisition report and look at session source for chatgpt.com from October 1 to now. Then check direct traffic on your top AI-referred landing pages over the same dates. If ChatGPT dips between roughly October 8 and 15 while direct rises on those same pages, you’re looking at the tracking bug, not a change in demand. Pages with few bookmarks or branded visits are the clearest test, because a sudden jump in direct traffic there is hard to explain any other way.
2. Annotate the affected dates
Huesler recommends marking October 8 to 15 in GA4 or Adobe Analytics. In GA4, you can right-click a point on any line graph in Reports and choose to add an annotation with a title, description and date range. Google’s help page says you need Analyst access or higher on the property to create one, and the note then shows on every line graph in that property. Write something plain, such as “ChatGPT UTM bug: AI visits counted as direct,” so anyone reading the report later understands the dip.
3. Don’t change your attribution model
This is the easiest mistake to make. The visits didn’t disappear; they were mislabeled for about a week. Changing attribution settings in response would distort every other channel’s numbers going forward. Leave the model alone and explain the gap in reporting instead.
4. Give AI assistants their own channel
By default, GA4 usually lumps AI assistant visits in with Referral traffic, which makes shifts like this hard to spot. Google’s own help page on custom channel groups includes an example for an “AI assistants” channel. You create a channel group, add a channel with a “matches regex” condition on source for assistants such as ChatGPT, Gemini, Copilot, Claude and Perplexity, and move it above Referral so it gets matched first. Once it’s in place, a sudden drop in AI traffic stands out on its own line instead of hiding inside referrals.
5. Keep UTMs on the links you control
Huesler’s advice is to keep relying on utm_source=chatgpt.com, because referrer data is still missing from most ChatGPT clicks. For links you place yourself, such as in press releases, guest posts, GPT listings or partner content, tag them consistently so you aren’t depending on any one platform to label traffic for you. A shared UTM naming sheet across your SEO, PR and paid teams keeps the data clean.
6. Tell clients before they ask
If you send monthly reports, add a short note to October’s: what happened, which dates were affected, and that the traffic was reclassified rather than lost. Clients who hear it from you first will trust the next AI traffic number you show them. Clients who spot the dip themselves will start asking whether the whole AI strategy is working.
What to watch next
AI assistants are becoming a real discovery channel, and each one handles outbound links differently. Search Engine Land’s takeaway was that changes in how AI platforms open links can make measurement harder even when the number of actual visitors stays about the same. Expect more of these breaks as products update. The safest approach is a dedicated AI channel, annotated anomalies, consistent UTMs, and a habit of checking direct traffic whenever an AI source suddenly drops.
It also helps to measure the inputs, not just the clicks. Coverage, citations and brand mentions on trusted sites are what get you named in AI answers in the first place. That’s where Digital PR and guest posting come in, and why tracking set up properly through our Growth Setup service matters before you judge any channel. For the content side, see our post on generative engine optimization.
FAQ
Why did my ChatGPT referral traffic drop in October 2026?
ChatGPT links stopped carrying the utm_source=chatgpt.com tag after the GPT-6 release, and referrer data was also missing. Analytics tools counted those visits as direct traffic. OpenAI confirmed it was a bug, and a fix began rolling out on October 15.
Which dates were affected?
Adobe’s Cedric Huesler recommends annotating October 8 to 15, 2026 in GA4 or Adobe Analytics as the affected period.
Is ChatGPT traffic tracking fixed now?
Mostly. By 9 a.m. Pacific time on October 15, tagged ChatGPT clicks were back to about 95% to 100% of the previous week’s level, based on telemetry across thousands of domains. Referrer data is still missing from most ChatGPT clicks, so the UTM tag remains the main signal.
Should I change my attribution model because of this?
No. The visits were mislabeled, not lost. Changing your attribution model would distort other channels. Annotate the dates and explain the gap in reporting instead.
How do I track ChatGPT traffic separately in GA4?
Create a custom channel group with an “AI assistants” channel that matches AI sources such as chatgpt.com with a regex condition, then place it above Referral. Google’s help page on custom channel groups includes this exact example.
Need cleaner AI traffic reporting or more brand mentions in AI answers? Read more on the Tech Hustler blog, see our press release distribution service, explore all services or book a free strategy call.
Sources: Search Engine Land (Oct 20), Google Analytics Help: About annotations, Google Analytics Help: Custom channel groups
