Was the ChatGPT traffic drop in October 2026 a UTM bug, now fixed?

Yes. The apparent ChatGPT traffic collapse in October 2026 was caused by a bug in GPT-6's link handling, not a real drop in visitors, and OpenAI began rolling out a fix on 15 October. Tagged links have since recovered to near-normal levels, though referrer data remains incomplete and site owners should still rely on UTM parameters rather than referrer strings to track ChatGPT traffic.

What happened to ChatGPT traffic in October 2026

In mid-October, site owners and analysts started noticing that ChatGPT-attributed traffic was falling off a cliff in their analytics dashboards. Adobe's Cedric Huesler investigated by examining roughly 500 pages where ChatGPT had previously accounted for at least 5% of visits, comparing the week of 9 to 13 October against the preceding week. Taken at face value, the drop looked like a sudden collapse in ChatGPT as a referral source, which matters for anyone tracking AI visibility across platforms.

What caused it

The cause was technical, not a sign that users had stopped clicking through from ChatGPT. Huesler found that GPT-6 was opening links using `window.open(url, "_blank", "noopener,noreferrer")`, which strips both the UTM parameter and the referrer header. Visits that should have shown up as ChatGPT referrals were instead landing in analytics as Direct traffic, because there was no utm_source=chatgpt.com tag and no referrer to fall back on. OpenAI had only added UTM tracking to ChatGPT links back in June, so the GPT-6 rollout undid months of improved attribution overnight. The change coincided with GPT-6's new Intelligent UI, which began rolling out on 7 October for paid plans and 8 October for Free and Go. For the broader shift in citation behaviour, see how ChatGPT's Intelligent UI shakes up what counts as a citation.

How big was the misattribution

The scale is well documented. Profound's tracking data, cited by The Prompt Insider, showed the share of ChatGPT answers with at least one tagged source link fell from about 98% on 8 October to 66% by 14 October, while Intelligent UI's own tagging rose from 2% to about 35% of the answers it collected. Adobe's investigation put real numbers on it: across the roughly 500 pages studied, Huesler identified around 31,000 fewer daily visits attributed to ChatGPT, alongside an increase of approximately 26,000 Direct visits. That near one-to-one swap is strong evidence that the decline was caused by incorrect attribution rather than visitors actually disappearing. Metrics Rule's analysis of the same issue found the bug caused ChatGPT-attributed traffic to drop by up to 46% in GA4 because of the missing UTM tags. Webepex's analysis flagged a mobile weak point too: 95% of mobile ChatGPT visits in one tracked dataset relied on the UTM tag alone, with no referrer as backup, which made mobile traffic especially exposed once the tag vanished.

Timeline: discovery, confirmation and fix

The timeline is tight. Barry Schwartz at Search Engine Roundtable reported on 14 October 2026 that UTM parameters had gone missing, with no OpenAI response at that point. Huesler contacted OpenAI on 14 October, and the company confirmed it was a known bug with a fix expected that week. Huesler updated his post on 15 October to confirm it was a bug, and the fix began rolling out from that same date. French outlet Siecle Digital independently corroborated that the UTM parameter returned on 15 October, and Huesler then reported that ChatGPT was tagging its links with `utm_source=chatgpt.com` again.

Has it actually been fixed

Largely, yes, with caveats. French outlet sitebrunch reported in a 24 October update that UTM tagging works again, and since 16 October almost all ChatGPT links carry the `utm_source=chatgpt.com` parameter. It also noted that OpenAI has since confirmed and fixed the bug privately, but has not made a public announcement about it. Margo Digitech's telemetry, covering thousands of domains, found tagged ChatGPT clicks recovered to roughly 95 to 100 percent of the levels seen in the same hour the previous week. DEV Community's write-up adds a nuance worth flagging: the fix makes measurement more reliable going forward, but it doesn't retroactively repair the reporting gap from the affected period, so anyone pulling historical reports needs to account for that hole.

What to do about it in your own analytics

Search Engine Land's recommended response is straightforward: annotate 8 to 15 October in Google Analytics 4 or Adobe Analytics, avoid changing your attribution models since the apparent loss was misclassification rather than a genuine decline, and keep using UTM tracking, since referral information is still missing from most ChatGPT clicks. Gen3 Marketing's coverage of the same issue advises that you filter on the `utm_source=chatgpt.com` tag rather than the referrer, which is still absent on most ChatGPT clicks even after the fix. One piece of good news for advertisers: Search Engine Journal reported that clicks on ChatGPT Ads with your own UTM parameters were unaffected by the bug, so paid traffic measurement wasn't part of this. For the practical fallout and how to rebuild clean reporting, see ChatGPT dropping its tracking tags and why your AI traffic numbers are probably wrong. If you're also tracking visibility across Google's AI surfaces, pair this with the Generative AI performance report in Search Console rather than relying on ChatGPT referral data alone. It's also worth reviewing a related drop in Reddit's ChatGPT citation volume over a short window, since citation volatility and referral tracking issues both complicate any single-platform view of AI visibility.

Our view: this episode is a useful reminder that AI referral tracking is fragile infrastructure built on assumptions a model vendor can break with a single rendering change. If your entire AI visibility picture depends on one tag surviving every future ChatGPT update, you're one release note away from another false alarm.

Caveats: this isn't necessarily over

Not everyone treats this as a one-off bug. Comergent's analysis of Shopify merchant data took a more cautious line, arguing site owners should treat this as the current state of ChatGPT's outbound links rather than a temporary bug, and plan attribution around the possibility that any platform could make a similar change without notice. That's a fair limitation: the notes show recovery and a confirmed fix, but there's no public commitment from OpenAI guaranteeing this won't happen again with a future model release. If your analytics strategy leans on a single UTM parameter as the only signal of AI-driven traffic, build in redundancy rather than assuming the current fix is permanent.

Frequently asked questions

Did ChatGPT traffic actually fall in October 2026?

No, not by anywhere near the margin dashboards showed. Adobe's analysis found the apparent decline was largely caused by incorrect attribution, with visits showing up as Direct traffic instead of ChatGPT referrals.

What exactly broke the tracking?

GPT-6 started opening links with a `noopener,noreferrer` method that strips both the UTM tag and the referrer header, so analytics tools couldn't identify the traffic source.

When was it fixed?

OpenAI confirmed the bug on 14 October and the fix began rolling out from 15 October, with tagged clicks recovering to 95 to 100 percent of prior-week levels shortly after.

Should I still use UTM parameters to track ChatGPT traffic?

Yes. Referrer data remains largely missing even after the fix, so `utm_source=chatgpt.com` is still your most reliable signal for identifying ChatGPT-driven visits.

Could this happen again?

Possibly. One analysis argued this should be treated as the current reality of ChatGPT's link handling rather than a one-time glitch, since nothing prevents a future model update from changing link behaviour again without notice.