ChatGPT stopped passing UTM parameters on outbound links around October 9, the week GPT-6 shipped, and Adobe's analysis of roughly 500 pages shows the traffic did not vanish — it moved into the "direct" bucket. Analytics tools now file ChatGPT referrals as direct visits. The change removes the only reliable signal publishers used to measure ChatGPT-driven traffic since June, when OpenAI first added UTM tracking to its links. OpenAI has not confirmed the removal was intentional. Takeaway: your ChatGPT traffic numbers did not collapse; your attribution did.

What exactly changed in the links?

Cedric Huesler of Adobe posted the technical detail on LinkedIn. He wrote: "GPT-6 opens links with window(.)open(url, '_blank', 'noopener,noreferrer') and drops ?utm_source=chatgpt(.)com. There's no tag and no referrer, so your analytics files the visit under direct."

Two things disappeared at once:

• The ?utm_source=chatgpt.com parameter, added in June.
• The HTTP referrer header, killed by the noopener,noreferrer window-opening call.

In plain words: there is now no tag and no referrer, so no standard analytics setup can tell a ChatGPT click from a typed URL. Takeaway: this is a code-level change, not an analytics bug on your side.

How big is the shift?

Huesler ran the numbers himself. He wrote: "To check, I looked at ~500 pages where ChatGPT drove 5%+ of visits, comparing Oct 9–13 with the week before."

His findings:

• ChatGPT-tagged traffic fell by about 31,000 visits per day across the sample.
• Direct traffic rose by about 26,000 visits per day on the same pages.
• Search traffic fell 7% — in line with the broader trend, so unrelated to the link change.
• Direct traffic rose on two out of three pages that lost ChatGPT traffic; the week before, that pattern was a coin flip.

He also checked whether the new direct visitors behaved like ChatGPT visitors. On mobile, the share seeing a consent banner — a proxy for first-time visitors — rose from 36% to 44% on those pages. For known ChatGPT visitors, the figure sits at 55%. The behavioral fingerprint matches. Takeaway: this is one analyst's sample of 500 pages, not a platform-wide study, but the numbers line up cleanly.

Who else is seeing it?

Sébastien Billard, an independent SEO observer, reported the same pattern on October 14. He wrote that since October 9 he saw "a sharp drop in Chat GPT human traffic, and also a drop of Chatgpt-user robot traffic," while OAI Search bot crawl traffic increased over the same period, starting October 7.

That second observation deserves caution. Billard stated it as an observation, not a tested conclusion, and no one has yet established a link between the crawler shift and the referrer change. Others in the SEO community, including commentators tagged in the same threads, have complained about the tracking loss as well. Takeaway: multiple independent reports agree the change landed around October 9; the crawler detail remains unverified.

What does this mean for publishers?

Any report that splits traffic by channel will now undercount ChatGPT. Publishers who saw June's UTM rollout as the arrival of measurable AI referral traffic lost that measurement layer in October, four months later.

Note what is missing from the story. OpenAI has said nothing publicly. Nobody has confirmed whether the removal is deliberate policy or a side effect of GPT-6's new link-handling code. The Adobe numbers cover 500 pages that already had meaningful ChatGPT traffic; sites with lighter AI referrals will see smaller, harder-to-detect shifts. Takeaway: treat every "ChatGPT traffic collapsed" claim from mid-October onward with suspicion — check direct-traffic trends first.

What should you monitor next?

Watch for three things: a statement from OpenAI on whether the change is intentional, any return of referrer or UTM data in a subsequent release, and whether analytics vendors build workarounds, such as first-party heuristics that infer ChatGPT visits from session patterns. Until one of those lands, direct-traffic deltas around October 9 are the closest thing publishers have to a measurement of ChatGPT referrals.