ChatGPT Traffic Drop Linked to Tracking Bug After GPT-6 Rollout

A sharp decline in ChatGPT-attributed website traffic recorded across multiple analytics platforms beginning October 9, 2026 appears to stem from a tracking malfunction rather than a genuine fall in user interest — with a fix already rolling out as of October 15.

The discovery matters because hundreds of website owners and SEO professionals were staring at data showing traffic from ChatGPT had collapsed by as much as 57% in a single week. As it turns out, the numbers were telling a story about broken attribution tags rather than broken audience behavior. For anyone managing a content strategy or reporting AI-driven traffic to stakeholders, understanding what actually happened is critical before drawing the wrong conclusions.

---

What the Data Showed — and Why It Was Misleading

Graphite CEO Ethan Smith first flagged the anomaly in a LinkedIn post. He reported that Google Analytics 4 sessions attributed to ChatGPT fell between 22% and 57% across sites his firm tracks with meaningful AI Assistant channel traffic. The drop began on October 9 and Smith estimated it represented approximately a 45% decline from the previous baseline.

LLM Pulse independently confirmed the pattern. Sessions from ChatGPT on the sites it monitors fell approximately 46% during October 9 through 11 compared with the same days in the prior two weeks. Crucially, total traffic on those same sites remained relatively constant over the same period — an early signal that visitors were not disappearing but were simply being miscounted.

Smith identified the cause. ChatGPT had begun opening inline links in answer text without the utm_source=chatgpt.com tag or a referrer header, meaning visits from those links were being recorded as Direct traffic inside GA4 rather than being attributed to ChatGPT. This kind of attribution gap is a recurring challenge for marketers working to understand the differences between direct and organic traffic sources — and why one can suddenly spike at the expense of the other.

Cedric Huesler, a senior director of product management at Adobe, added weight to that explanation. Comparing October 9 through 13 with the prior week across approximately 500 pages where ChatGPT drove at least 5% of visits, he found:

• ChatGPT visits fell by about 31,000 per day
• Direct visits rose by about 26,000 per day
• Search traffic fell 7%, in line with broader search trends
• Direct traffic rose on two of every three pages that lost ChatGPT attribution, compared with about half the week before

He also noted that the new Direct visitors behaved like ChatGPT visitors. On mobile, the share of first-time visitors among Direct traffic climbed from 36% to 44%, moving closer to the 55% he observes among confirmed ChatGPT visitors — a behavioural fingerprint that strongly suggested misattribution rather than genuine audience loss.

---

How GPT-6's Rollout Changed Link Behaviour

Analysts tied the tracking breakdown directly to the GPT-6 rollout. GPT-6 began rolling out to paid plans on October 7 and reached Free and Go users on October 8. LLM Pulse recorded no drop on October 7 — the timing aligns precisely with the wider rollout the following day.

Huesler explained that after the GPT-6 rollout, ChatGPT began opening links in a new tab using a browser setting that withholds the referrer and stopped appending the utm_source tag that ChatGPT's citation links had carried since its search feature launched in 2024. Smith noted that:

• Inline links in answer text lost the tag
• Links in the Sources panel and shopping links retained it
• 95% of mobile ChatGPT visits to Graphite's sites relied on the tag alone, making app traffic the most exposed segment

Profound, an AI search analytics firm, tied the change to a new answer format called Intelligent UI — ChatGPT's format for charts, comparisons and small tools. The share of ChatGPT answers with at least one tagged source link fell from approximately 98% on October 8 to 66% by October 14 according to Profound's data. Over the same period, Intelligent UI went from about 2% of all answers Profound collected to approximately 35% — a rapid format shift with immediate consequences for attribution.

Tomek Rudzki at Peec AI ran the same one million prompts across two periods — October 2 through 7 and October 9 through 13. He found that ChatGPT included a direct URL in 42% of responses, compared with a previous rate of roughly 8% to 10%. He noted that 54% of those URLs pointed to home pages and that no UTM tags or referrers were attached. Those figures represent links displayed by ChatGPT and not confirmed user click-throughs — an important distinction when interpreting the data.

Why This Matters for Your Analytics Reporting

The practical implication is significant. If your analytics platform relies on referrer headers or UTM parameters to attribute ChatGPT traffic, any session arriving without those signals will be counted as Direct. For sites with substantial AI-driven audiences, this can distort your channel mix overnight — making it appear that a high-performing source has collapsed when in reality the audience is still arriving, just uncounted. Understanding the foundations of SEO reporting and web analytics is increasingly important as AI platforms introduce new referral behaviours that traditional tracking setups were not designed to handle.

---

The Fix and What Remains Unresolved

Huesler reported on October 14 that OpenAI confirmed the tracking behaviour was a bug. A fix began rolling out early on October 15 and by 9 a.m. PT, tagged ChatGPT clicks had returned to roughly 95% to 100% of the same hour the previous week across thousands of domains he tracks. He added, however, that most ChatGPT clicks still arrive without a referrer and recommended counting the UTM tag rather than the referrer as the more reliable signal going forward.

Not everyone confirmed the fix immediately. SEO consultant Glenn Gabe replied that he was not yet seeing the change on paid or free accounts. Profound stated it had not received official confirmation, and OpenAI's ChatGPT release notes for October 1 through October 15 made no mention of link tracking. The open question is how quickly tags return across every account type, since OpenAI has not publicly documented the fix.

Smith called the issue a bug and said he expects OpenAI to resolve it fully. Attribution already lost during the October 8 through 15 window cannot be recovered retroactively. Meanwhile, Google's Gemini moved in the opposite direction two weeks earlier by adding UTM parameters to its links — a contrast worth noting as AI platforms develop their own attribution standards. You can read more about how Google approaches UTM attribution in its official Analytics Help documentation.

What You Should Do Now

The episode is a reminder that in the rapidly evolving AI search landscape, a dramatic traffic number is not always what it appears. Annotating October 8 in your analytics reporting before stakeholders interpret the dip as lost audience is a practical first step.

Reviewing ChatGPT and Direct traffic side by side for each page — especially your homepage, where Rudzki found more than half of direct links pointed — will give a cleaner picture of what actually changed. If you are running ChatGPT Ads with your own UTM parameters, Smith confirmed those should be unaffected by the bug.

For site owners looking beyond this specific incident, the broader lesson is that AI-referred traffic is now a meaningful enough channel to warrant dedicated monitoring. Building a reliable strategy to increase and sustainably grow website traffic now requires accounting for AI platforms as referral sources — complete with the attribution quirks they introduce.