ChatGPT appears to have stopped passing UTM tracking parameters on the links it sends to websites, according to a report from Search Engine Land. For publishers and marketers, that change makes AI referral traffic harder to measure.

UTM parameters are the short tags — such as utm_source or utm_medium — that analytics tools attach to URLs. When a reader clicks a tagged link, the destination site can see exactly where the visit came from. Strip those tags, and the visit still arrives, but it arrives without a return address.

The shift matters because AI chatbots are becoming a meaningful source of traffic. If ChatGPT referrals no longer carry UTMs, they may instead land in analytics reports as "direct" traffic, mixed in with people typing a URL by hand. That hides a growing channel inside an uninformative bucket.

How did the change surface?

The observation came from Search Engine Land's monitoring of referral data. The publication reports that ChatGPT "appears to drop UTM tracking," which means the behavior looks deliberate rather than accidental, but OpenAI has not confirmed the mechanism or the reason.

There is no official statement on timing either. The report frames the finding as an observed pattern in traffic data, not a documented product announcement.

What does this change for publishers and marketers?

The practical question is simple: how do you count visitors that ChatGPT sends you if the link no longer says so? A few consequences follow.

• AI referral counts may understate reality, because stripped-UTM visits no longer register under an AI source.
• "Direct" traffic numbers may inflate, absorbing visits that actually came from chatbot answers.
• Campaign attribution weakens — marketers lose the ability to tie a visit to a specific AI entry point.
• Trend analysis gets noisier, since a channel that was measurable becomes blended with another.

For teams that report monthly traffic by channel, this is not a crisis. It is a measurement problem: the visitors still exist, but the labels they carried are gone.

What can sites do about it?

The report does not prescribe fixes, but the situation points to standard workarounds. Analysts can watch server logs for ChatGPT-related referrer strings, where available. They can also look at landing-page patterns, since chatbot referrals often concentrate on specific answer-linked pages.

None of these methods replaces clean UTM data. They are approximations, and they require more effort than reading a dashboard line that no longer tells the truth.

Why strip the tags at all?

Search Engine Land does not state OpenAI's motive. One plausible reading is privacy: UTM tags are a form of tracking, and removing them fits a broader industry pattern of limiting cross-site data collection. Another reading is simply a technical change in how ChatGPT constructs outbound links.

Without confirmation from OpenAI, both remain speculation. What is documented is the effect: UTM tags from ChatGPT referrals are no longer showing up as before.

The bigger picture

The measurement gap arrives as the industry argues over how AI platforms should credit publishers. If chatbot traffic becomes invisible in standard analytics, publishers lose one of the few levers they have for quantifying what AI services send them — and what that traffic is worth.

For now, the advice is practical: treat your "direct" traffic numbers with suspicion, check server logs, and expect AI referral figures to look smaller than the reality behind them.