How to Track ChatGPT and AI Referrals That GA4 Cannot Tie to Revenue

When potential customers ask ChatGPT, Claude, Perplexity, or Gemini for product recommendations, the visit that follows is easy to label and hard to value.

Labelling it got easier in May 2026, when Google Analytics 4 added an AI assistant channel (medium=ai-assistant). It recognises ChatGPT, Gemini, DeepSeek, Copilot and Grok from the referrer. If you use GA4 and you have wondered whether you already have this covered: for those five, with a referrer present, you largely do.

Three gaps remain, and they are the ones that decide budget.

Where the Channel Label Runs Out

It does not cover every engine. GA4’s AI assistant channel excludes Perplexity, and it does not separate Google’s own AI Overviews or AI Mode — those arrive with a google.com referrer and read as organic search. (For transparency: SourceTrack does not separate AI Overviews or AI Mode either. Nobody can, from the referrer alone.)

It is forward-only. GA4 does not reclassify history. Sessions that landed before the change stay in whatever bucket they were given, permanently.

It is a label on a session, not a line on a journey. This is the one that matters. GA4 is session-scoped: a visitor who finds you through ChatGPT in week one and buys in week four arrives as a new session with no memory of the first. The channel gets a label; the order does not get a source. There is no revenue-to-journey join in the product at all.

Traditional analytics also relies on the HTTP Referer header, and modern LLM interfaces and mobile chat apps often strip it. That part is genuinely unrecoverable — by anyone. When a referrer is absent and no UTM is present, the visit is indistinguishable from someone typing your address in, and it belongs in Direct. No tool can recover it without inventing a source.

The SourceTrack Approach to AI Attribution

SourceTrack matches the referrer host and any campaign parameters against a named list of assistants, with no third-party cookies and no device fingerprinting.

1. Named referrer matching: The referrer host and UTM source are matched against 16 named assistants and answer engines on arrival — a fixed list, not a guess.
2. First-Touch Stitched Timeline: Preserves the originating AI discovery channel across subsequent visits before purchase.
3. Revenue Attribution: Connects Stripe purchase webhooks and Shopify orders directly to the initial AI referral touchpoint.

By filling the dark traffic gap, growth teams get clear visibility into which specific AI search platforms drive real enterprise pipeline and recurring revenue.