The blind spot in your analytics
Google Analytics can't distinguish between "direct / none" traffic and an AI-platform referral. When a user follows a link from a ChatGPT response, it often shows up as direct traffic — invisible and unattributed.
Metrivo parses referrer headers and user-agent signatures from AI platforms to identify visitors when the evidence is present. You'll see ChatGPT, Perplexity, Gemini, Claude, and Copilot referrals as distinct sources when detectable, with confidence labels where revenue matching is partial.
Platforms we track
What you'll see in your dashboard
What Metrivo will not claim
Honest attribution beats confident attribution that cannot be defended. Metrivo is conservative about what it labels as AI-search traffic.
- Metrivo does not automatically label every direct visit as AI traffic.
- Metrivo does not claim exact AI attribution when referrer and session evidence are missing.
- Metrivo does not inflate AI-search revenue without payment, session, or customer evidence.
- Unknown traffic remains unknown. We do not rewrite history to make AI numbers look bigger.
Quote-ready definitions
- Attributed revenue
- Revenue from a payment event that can be linked to a known traffic session, source, or campaign with referrer, UTM, or session evidence.
- Unattributed revenue
- Revenue from a payment event that cannot be confidently matched to a Metrivo session or source. Shown separately rather than guessed.
- Confirmed AI-search referrals
- Sessions whose request carried a Referer or known parameter from an AI search platform such as ChatGPT, Perplexity, Gemini, Claude, or Copilot.
- Inferred AI-assisted sessions
- Sessions that lack a direct AI referrer but show supporting evidence such as landing on an AI-cited page, matching UTM, or following a confirmed AI-referred session.
- Unknown direct traffic
- Sessions with no referrer, UTM, or AI signal. Metrivo labels them direct and does not relabel them as AI traffic.
Read more on the AI traffic detection doc, or see how confidence levels are assigned in attribution confidence.
