Recent data shows a sharp drop in ChatGPT-attributed traffic. Analysts suggest a tracking failure linked to GPT-6 rather than a loss of actual visitors.

For digital marketers and business owners, the analytics dashboard is the primary source of truth. When a primary traffic channel suddenly plunges, the immediate reaction is panic. This is exactly what happened across thousands of websites starting October 9, as ChatGPT-attributed traffic appeared to vanish from Google Analytics 4 (GA4) reports.

The numbers are stark. Ethan Smith, CEO of Graphite, reported that ChatGPT-attributed sessions fell by 22% to 57% for sites with significant AI Assistant channel traffic. Smith estimates this represents a roughly 45% drop from the previous baseline. Similarly, LLM Pulse observed a 46% decrease in sessions from ChatGPT between October 9 and 11 compared to the preceding two weeks. On the surface, it looks like a catastrophic loss of visibility within the world's most popular AI interface.

The phantom decline in AI referrals

Despite the alarming drop in attributed sessions, a critical detail emerges: total traffic on these sites remained relatively constant. If users had actually stopped clicking links within ChatGPT, the total session count would have dipped in tandem with the AI referral metric. Instead, the traffic did not disappear; it simply changed its identity in the eyes of tracking software.

The prevailing theory among analysts is that this is a tracking failure rather than a behavioral shift. When a user clicks a link, the browser typically sends a referrer header or a UTM tag (like utm_source=chatgpt.com) that tells the destination website where the visitor came from. Without these markers, GA4 defaults the visit to Direct traffic, the digital equivalent of a visitor typing the URL manually into their browser.

How the GPT-6 rollout altered the data

The timing of this anomaly is not accidental. The shift began immediately following the rollout of GPT-6. According to Cedric Huesler, a senior director of product management at Adobe, the new version of the AI began opening links in new tabs using browser settings that withhold the referrer. Furthermore, the specific UTM tags that accompanied citations since the launch of ChatGPT search in 2024 have started to vanish from inline links.

The impact is not uniform across all parts of the interface. Data indicates a divergence in how different link types are handled:

Inline links within the answer text have largely lost their tracking tags, whereas links located in the Sources panel and specific shopping links continue to carry them.

This creates a fragmented reporting environment. For businesses relying on Search Engine Journal insights to monitor their AI visibility, the data is now skewed. The most affected users are those on mobile; Smith noted that 95% of mobile ChatGPT visits to Graphite sites relied solely on the tag, making app-based traffic the most invisible segment of the current crisis.

Tracing the migration to Direct traffic

To prove that users are still visiting, Huesler analyzed approximately 500 pages where ChatGPT previously drove at least 5% of visits. His findings provide a clear map of the traffic migration. Between October 9 and 13, ChatGPT visits to these pages fell by roughly 31,000 per day, while Direct visits rose by approximately 26,000 per day. This near-perfect inverse correlation suggests that the users are still there; they are just arriving unannounced.

Huesler further validated this by analyzing user behavior. On mobile devices, the share of Direct visitors who encountered a consent banner—a strong indicator of a first-time visit—rose from 36% to 44%. This movement brings Direct visitor behavior closer to the 55% rate typically seen among ChatGPT visitors, confirming that these new Direct visitors are behaving exactly like AI-referred users.

The rise of Intelligent UI and new formats

The technical cause may be linked to a fundamental change in how OpenAI presents information. Profound, an analytics entity, observed a shift in the answer formats provided by the AI. The share of ChatGPT answers containing at least one tagged source link plummeted from 98% on October 8 to just 66% by October 14.

Simultaneously, the adoption of Intelligent UI—a format used for charts, comparisons, and small interactive tools—surged from 2% to 35% of all collected answers. It appears that as OpenAI moves toward more complex, integrated UI elements to deliver data, the traditional method of appending tracking tags to simple hyperlinks is being discarded or overlooked in the new architecture.

Measuring success in a post-referrer era

This situation highlights a growing vulnerability for businesses: the reliance on third-party platforms to provide transparent attribution. As AI assistants become the primary gateway to the web, the difficulty in measuring referral traffic becomes a strategic risk. If a company cannot prove that an AI is driving leads, it becomes harder to justify content strategies optimized for LLMs.

The current crisis suggests that the Direct traffic category is becoming a dumping ground for various AI-driven visits. For entrepreneurs, this means that the traditional GA4 'Acquisition' report is no longer a reliable mirror of reality. The focus must shift toward holistic metrics—such as total conversion rates and overall traffic volume—rather than obsessing over the specific source of every click.

Strategic implications for US and UK enterprises

For businesses operating in the USA and UK, this tracking anomaly is a warning about the volatility of the AI ecosystem. Unlike the structured environment of Google Search, where indexing and referral rules are relatively stable, the AI landscape is subject to overnight changes during model updates. In the US market, where data-driven attribution is central to marketing budgets, this lack of transparency can lead to incorrect pivots in strategy.

From a regulatory perspective, while the EU AI Act focuses heavily on safety and transparency of the model's output, the US and UK markets are more concerned with the economic impact of AI on the open web. The erosion of referral data complicates the ability of publishers to negotiate fair compensation or visibility agreements with AI providers. When the 'receipt' of a visit (the referrer tag) is removed, the evidence of value creation is erased.

Enterprises should now prepare for a world where Dark Social and Dark AI traffic dominate. The strategy should move away from relying on platform-provided tags and toward first-party data collection, such as 'How did you find us?' surveys and advanced server-side tracking, to recapture the visibility lost in the transition to GPT-6.

FAQ

Did ChatGPT stop sending traffic to websites?

No. Evidence shows that total traffic remained constant, but the visits are now categorized as Direct instead of AI referrals due to missing tracking tags.

Why did this happen after the GPT-6 rollout?

GPT-6 introduced changes in how links are opened (new tabs with withheld referrers) and a new 'Intelligent UI' format that often lacks the traditional UTM tags.

How can I tell if my traffic drop is a tracking issue or a real loss?

Check your Direct traffic. If Direct visits increased by a similar amount that your ChatGPT referrals decreased, it is likely a tracking problem.

Which devices are most affected by this change?

Mobile users are the most affected, as a vast majority of mobile visits relied exclusively on UTM tags for attribution.

---

Sources: Searchenginejournal, Semasocial, Msn ·