ChatGPT Drops UTM Tracking, Blurring AI Referral Traffic Data

ChatGPT appears to have stopped passing UTM parameters in its referral links, and that single technical change makes AI-driven traffic significantly harder to measure. For publishers and marketers who spent two years building dashboards around those tags, the timing is unwelcome.

UTM parameters are the short strings appended to URLs — the utm_source, utm_medium, utm_campaign tags — that tell analytics platforms where a visitor came from. Strip them out, and a click from ChatGPT looks like a bare referral, or worse, like direct traffic. Attribution gets murky fast.

The change surfaced in reports from ALM Corp, which observed that ChatGPT referral traffic no longer carries the tracking markers it previously did. The operative word is "appears": OpenAI has not, so far, published documentation confirming the shift or explaining it.

What does this change for publishers?

Plenty, in practice. Most analytics stacks — Google Analytics included — depend on UTM tags to sort inbound visitors into clean channels. Without them:

• ChatGPT referrals may collapse into "direct" or unlabeled referral buckets
• Campaign-level reporting on AI traffic becomes unreliable
• Historical comparisons between periods break, since old data carries tags and new data does not
• Editors lose a concrete number for how many readers chatbots actually send

That last point stings most. The whole industry argument about AI search — whether it cannibalizes traffic or complements it — rests on measurement. Remove the measurement, and the argument turns theological.

Why might the tags have disappeared?

Nobody outside OpenAI knows, and OpenAI is not saying. Link sanitization is a plausible motive; platforms sometimes strip tracking parameters for privacy or cleanliness reasons. But without a statement, attribution teams are left reconstructing intent from behavior.

The industry has been here before. Email clients, privacy browsers, and social platforms have each taken turns stripping parameters over the years, and each time analysts adapted with workarounds: server-side tagging, referral-path rules, user-agent segmentation. Those tools work, after a fashion. None of them matches the simplicity of a well-formed UTM string.

Who feels this first?

Publishers and B2B content teams, mostly. Anyone who reports monthly on channel performance now has to explain why the AI line item went ambiguous overnight. Marketing budgets follow attribution, and attribution just got vaguer.

Skeptics will note the broader pattern: as AI platforms become traffic sources, they also control how visible that traffic is to the sites receiving it. A chatbot that decides which parameters survive the click holds a quiet kind of power over the publishers' own analytics.

What can teams do about it now?

Until OpenAI clarifies, analysts have limited options. They can segment by referral domain where it still resolves, flag unexplained direct-traffic shifts, and treat any pre-change AI traffic figures as a baseline rather than a trend. None of this is elegant. All of it is temporary, presumably.

The takeaway: ChatGPT referrals no longer announce themselves, and every AI traffic number you report this quarter deserves an asterisk until further notice.