ChatGPT Tracking Bug Explains Recent Referral Traffic Drop

OpenAI confirmed on October 14, 2026, that a technical bug caused a sharp drop in ChatGPT-attributed traffic across various analytics datasets. This issue, which saw reported drops of 46% in some datasets, was caused by links opening without necessary tracking tags or referrers. If you manage AI search attribution and observed a sudden decline in ChatGPT traffic in your GA4 property between October 8 and October 15, this analysis provides the technical confirmation needed to explain the shift to stakeholders.

The ChatGPT traffic drop was a tracking bug, not a loss of interest

The ChatGPT traffic drop was primarily a failure in technical attribution rather than a decline in user behavior. Graphite CEO Ethan Smith reported that ChatGPT began opening links in the answer text without the utm_source=chatgpt.com tag or a referrer header. (referrer header: a piece of data sent by a browser to tell a website the address of the previous page.) (utm_source=chatgpt.com: a specific tracking tag used in website links to identify the source of incoming web traffic.) When both the utm_source tag and referrer header are removed, Google Analytics 4 (GA4) defaults the session to the Direct channel. Consequently, visits that should have been attributed to the AI Assistant channel were instead captured as Direct traffic.

Why mobile traffic was disproportionately exposed

Mobile traffic suffered a higher rate of misattribution because 95% of mobile ChatGPT visits to Graphite’s sites relied on the tag alone. Because the utm_source tag was stripped from the answer text, the lack of a fallback referrer header meant these sessions became invisible to the AI Assistant channel. This mechanism explains why the drop was so severe for mobile users compared to desktop users who might still have partial attribution via other headers.

Data highlights the correlation between Intelligent UI and attribution loss

The shift toward ChatGPT’s Intelligent UI—the platform’s format for charts, comparisons, and small tools—directly correlates with the decrease in captured attribution. Profound recorded the share of ChatGPT answers with at least one tagged source link falling from 98% on October 8 to 66% by October 14. Simultaneously, the use of Intelligent UI rose from 2% to 35% of all answers collected. This suggests that as ChatGPT moves toward more interactive, widget-based outputs, the traditional method of passing utm_source tags through text links is being bypassed.

Analysis of the 31,000 daily visit drop

Adobe’s Cedric Huesler measured a specific shift where ChatGPT visits to certain pages fell by about 31,000 a day while Direct visits rose by about 26,000 a day between October 9 and October 13. Huesler noted that these new Direct visitors behaved similarly to ChatGPT visitors, and the rise in Direct traffic occurred on two of every three pages that lost ChatGPT traffic. Further analysis by Peec AI showed that ChatGPT began placing direct URLs into 42% of its responses, up from a previous rate of 8-10%, with 54% of those URLs pointing to home pages.

OpenAI confirmed the bug and a fix on October 15

OpenAI confirmed the tracking failure to Cedric Huesler on October 14, 2026, and began rolling out a fix early on October 15. By 9 a.m. PT on October 15, tagged ChatGPT clicks returned to roughly 95% to 100% of the levels seen a week earlier across thousands of domains. While the fix restored the tags, Huesler noted that most ChatGPT clicks still arrive without a referrer, making the utm_source tag the primary reliable metric for attribution.

Audit steps for the October 8–15 tracking window

SEO practitioners should annotate the October 8–15 window in all internal reports to prevent stakeholders from interpreting the drop as a loss of user interest. Analytics teams should cross-reference the increase in Direct traffic with a corresponding rise in home page hits to verify the missing ChatGPT volume. To ensure future reliability, technical teams should prioritize monitoring utm_source tags over referrer headers, as the latter remains inconsistent in the ChatGPT interface.

Who ChatGPT matters to

GA4 analysts who monitor the AI Assistant channel should compare the 22% drop in ChatGPT-attributed sessions against the Direct channel volume before and after October 8 to identify the impact of the tracking bug. SEO growth engineers managing technical site health should configure their server-side tracking scripts to capture any available referrer headers while monitoring OpenAI’s release notes for further link-handling changes. Digital marketing leaders who report monthly traffic growth should verify that the 31,000 daily fall in ChatGPT visits has been corrected in their latest reports to explain the variance to stakeholders.

Who this is not for: Beginner SEO students seeking a general introduction to search engine

Beginner SEO students seeking a general introduction to search engine optimization should skip this technical analysis and instead read a foundational guide on how GA4 and UTM tags function. Social media marketing managers who focus on engagement metrics for platforms like Instagram or TikTok should read a guide on social media marketing instead.

Iker runs a large e-commerce site and notices a 40% drop in ChatGPT-attributed traffic

Iker runs a large e-commerce site and notices a 40% drop in ChatGPT-attributed traffic in his weekly GA4 report. Iker checks the dates and sees the drop began around October 8, then looks at his Direct traffic and sees a corresponding rise in home page visits. Iker identifies the 54% of direct links pointing to the homepage as the source of the missing data and correctly attributes the shift to a tracking bug rather than a loss of interest. If the total traffic had also dropped by 40% instead of remaining constant, Iker would have correctly identified a real loss of interest and pivoted his content strategy accordingly.

What this replaces: Panic-driven reporting on 'lost' AI traffic

Panic-driven reporting on ‘lost’ AI traffic is replaced by a technical explanation of a tracking bug and a confirmed fix date, allowing for more accurate reporting to stakeholders.

Reported by Search Engine Journal on 15 October 2026: ChatGPT Traffic Drop Appears To Be A Tracking Problem via @sejournal, @MattGSouthern.