What Happened
Over the past month, several analytics datasets recorded a sharp decline in traffic attributed to ChatGPT. Analysts who investigated the anomaly lean toward the view that the cause is not waning user interest in the AI assistant, but rather the behavior of the links themselves.
The leading hypothesis: links from ChatGPT are increasingly opening without tracking parameters (UTM tags), which means analytics platforms cannot correctly associate the visit with its source. As a result, sessions either land in the direct traffic bucket or are marked as referrals without an identifier, muddying the overall attribution picture.
Why It Matters For Marketers
The situation exposes a fundamental vulnerability in modern tracking: we remain heavily dependent on how a platform passes link parameters. If ChatGPT, an in-app browser, or an intermediate redirect strips those tags, the entire channel becomes invisible in reports.
- Attribution suffers: real visits exist, but get credited to another source.
- Decisions are made on distorted data: budgets shift toward channels that may actually perform worse than they appear.
- Server-side tracking gains importance, along with independent measurement methods.
Market Context
Over the past year, ChatGPT and other AI assistants have become a meaningful source of referral traffic — especially for content projects, services, and information resources. Many teams now dedicate a separate row in their dashboards to AI-driven traffic. That is precisely why the current tracking glitch raises alarm: it hits one of the fastest-growing content distribution channels.
For affiliates and performance marketers, this is one more signal: you cannot build optimization solely on client-side analytics data. The more intermediaries between a user and a site, the higher the risk of losing the source.
Editorial Take
The story of the ChatGPT traffic drop is less about ChatGPT itself and more about the maturity of the measurement market. Attribution has always been a compromise between technical feasibility and accuracy, and with the arrival of AI platforms and closed ecosystems, this problem is only intensifying.
Our conclusion is simple: teams should stop treating a single counter as the ultimate truth. A combination of server-side tracking, logs, internal surveys, and cohort analysis delivers a far more resilient picture. Otherwise, any technical glitch at a major platform will look like a channel collapse — even when the channel is alive and well, just temporarily invisible.