Tru Commerce

← Glossary · Measurement

Server-side AI attribution

Also known as: AI attribution layer

Server-side infrastructure that identifies AI-driven traffic without relying on the HTTP Referer header. Three-signal architecture: fingerprinting, query-param negotiation, tracking pixel.

Server-side AI attribution is the engineering fix for DACT. Because AI surfaces strip the HTTP Referer header on outbound clicks, client-side analytics (GA4, Adobe, Mixpanel) can't identify the traffic source. A server-side attribution layer combines three parallel signals:

Signal A — Surface fingerprinting: JA3 TLS hash, User-Agent pattern, IP range cluster, session timing profile. Combined into a per-request confidence score for surface attribution.

Signal B — Query-parameter negotiation: ACP and UCP both expose mechanisms for merchants to negotiate outbound URL decoration. Where available, this is the strongest signal (confidence 1.0) and serves as labeled ground truth for training Signal A.

Signal C — Tracking pixel: lightweight JavaScript pixel on product pages capturing document.referrer, navigator.userAgent, JS-API access patterns, and timing. Cross-referenced with Signal A to close gaps.

A confidence model combines the signals via product-of-posteriors, capping combined confidence at 0.97 when Signal B is absent. Sessions below 0.6 combined confidence are bucketed as "AI-likely-unattributed."

Build cost: 4-8 weeks for a two-person team (backend + data engineer). Or buy — we ship this as the DACT panel inside Citation Rank.

See also

Related on Tru Commerce

See where Server-side shows up in your Citation Rank.

Get your free scan or book a demo