A sudden drop in website traffic attributed to ChatGPT has website owners checking their analytics dashboards twice. The likely explanation isn't that fewer people are clicking through from ChatGPT. It's that those clicks have stopped showing up correctly in the data.
What happened
Several analytics platforms and marketing teams noticed a steep decline this month in traffic labeled as coming from ChatGPT. On the surface, that looks like a signal that OpenAI's chatbot is sending fewer people to outside websites, which would matter a lot to businesses that have started treating ChatGPT citations as a new source of visibility, alongside Google search results.
But analysts digging into the issue found a more mundane cause. Links shared through ChatGPT are increasingly opening without the tracking tags that let analytics tools identify where a visitor came from. Without those tags, often small pieces of code appended to a URL, a visit from ChatGPT can look identical to someone typing a web address directly into their browser. The result is that real traffic isn't vanishing. It's being miscounted as "direct" traffic instead of "referral" traffic.
This isn't the first time referral data has gone murky. Analysts have long referred to this broader phenomenon as "dark traffic," visits that happened but can't be cleanly attributed to their true source because of missing or stripped tracking information. It became common years ago when browsers started hiding referrer data for privacy reasons, and it has resurfaced repeatedly as new platforms, apps, and now AI tools send traffic in ways that don't always preserve a clean trail back to the origin.
Why it matters
This fits a pattern that has followed every major shift in how people find websites. When social platforms, mobile apps, and voice assistants became significant traffic sources, each introduced its own attribution blind spots, and marketers spent months recalibrating before they trusted the numbers again. AI chat tools are now going through the same growing pains, and the stakes are a little higher because businesses are actively trying to figure out whether being cited by ChatGPT, Perplexity, or Google's AI Overviews actually drives visitors.
What's genuinely new here isn't the concept of lost attribution. It's that ChatGPT has become significant enough as a referral source that its tracking quirks are now visible in aggregate data across many websites at once, not just a handful of marketing blogs.
What this means for small businesses
If your analytics show ChatGPT referrals falling off a cliff this month, resist the urge to conclude your content has lost favor with the chatbot or that AI search visibility is declining. Check your "direct traffic" numbers for a corresponding spike around the same time. If direct traffic jumped as ChatGPT referrals dropped, that's a strong sign the visits didn't disappear, they just got reclassified.
This is also a reminder not to build marketing decisions on a single metric from a single source. If you're using AI-referred traffic to justify time spent optimizing content for AI chatbots, cross-check with server logs or a second analytics tool if you have access to one. Attribution gaps like this tend to understate AI's actual role in bringing customers to your site, not overstate it.
What to watch
Keep an eye on whether OpenAI makes any changes to how shared links are formatted, and whether Google Analytics or other major platforms roll out better detection for AI-driven visits. Also watch whether your own direct traffic numbers stay elevated or settle back down, since that pattern will tell you how much of the "lost" ChatGPT traffic was a measurement issue versus a real shift in behavior.
The bottom line
Treat this month's ChatGPT traffic numbers with skepticism rather than alarm, and verify any apparent drop against your direct traffic data before changing your content or SEO strategy.