Why it happens
- Apps: links opened from mobile and desktop apps may not pass a referrer.
- Copy and paste: buyers copy a URL from an answer.
- Privacy settings strip referrers.
Four ways to find it
- Use the tag. ChatGPT adds utm_source=chatgpt.com to many links; filter on it, not only on referrer.
- Watch deep-page direct traffic. Nobody types a long comparison URL; a rise there is a clue.
- Ask. Add "ChatGPT or another AI assistant" to "How did you hear about us?"
- Identify companies. Company identification names the visitor regardless of source.
Report it honestly
Show tagged AI visits, an estimate for untagged ones and the method you used. A range is better than a false precise number.
An estimate you can defend
- Take direct sessions to deep pages (not the homepage) this month.
- Compare with the same pages before you started AI work.
- Treat the growth as an upper bound for hidden AI traffic.
- Report: "Tagged AI visits: X. Estimated hidden AI visits: up to Y."
Questions
Could other sources cause the rise?
Yes: email, chat apps and bookmarks also arrive as direct. That's why the estimate is a range.
Do AI apps always strip referrers?
Not always. Behaviour varies by app, browser and platform.
Will this get better?
Tagging like utm_source=chatgpt.com helps. Expect some loss to remain.
Why this matters for decisions
If AI visits hide in direct traffic, AI's contribution looks smaller than it is, and budgets follow the visible numbers. A team might cut AI visibility work because "AI brings 1% of traffic" while that work drives a share of the direct visits that convert best.
Three numbers to report together
- Tagged AI visits: the conservative floor.
- Estimated hidden AI visits: the range from deep-page direct traffic.
- Self-reported AI: leads who say they found you through an AI assistant.
Shown together, they give leadership an honest picture: a floor, a range and a signal from buyers themselves.
Start here: Company identification from AI. More articles in Company signals.