# AI referral attribution methodology > How an identifiable AI referrer or campaign signal can connect a known visit to a matched payment, and what missing signals cannot establish. Canonical page: [https://attribiq.com/guides/ai-referral-attribution-methodology](https://attribiq.com/guides/ai-referral-attribution-methodology) Category: Revenue Attribution Last updated: 2026-09-02 Topics: AI referral attribution, AI referral traffic, referrer evidence, attribution limits ## Summary AttribIQ classifies reviewed AI-provider referrers and, when enabled, trusted campaign sources. Revenue is reported only when the resulting visit connects to a payment through a known visitor or session ID. Missing traffic, prompt mentions, citations and visibility estimates do not receive revenue credit. ## What this guide covers - Understand what counts as an observed AI referral - Distinguish AI-sourced revenue from an earlier AI touch - Understand why AI visibility is not revenue evidence - Recognize the limits of missing and dark referral traffic AI referral attribution connects a visit with an identifiable referrer or campaign signal to a later matched payment. It answers a narrow question: did a visible referral reach the site and remain connected through checkout? The method starts with evidence from the visit. It does not start with prompt monitoring, a visibility score, or a guess about which answer influenced a purchase. ## What counts as an observed AI referral An observed AI referral requires a signal attached to a real visit. AttribIQ currently classifies: - a browser referrer hostname that matches the versioned registry of reviewed AI-provider domains; or - an explicit campaign source that matches a reviewed provider rule when trusted campaign classification is enabled. AttribIQ's web tracker records referrer, source, UTM, landing-page, visitor, and session context. The classifier maps reviewed domains to providers and records the matched rule and classifier version. Referral hosts outside the registry remain available in the retained traffic data but are not automatically counted as AI traffic. Provider referrer behavior can change, and some assistants do not send a stable referrer. A platform name alone is not sufficient evidence. ## How visits connect to payment revenue The browser journey and payment system are separate until an identifier connects them. 1. The tracker records the observed visit and creates visitor and session identifiers. 2. The site passes those identifiers into checkout or its billing record. 3. Stripe webhooks or the generic Payment API send the payment event to AttribIQ. 4. AttribIQ joins the payment to the known journey and applies the selected attribution model. 5. Refund and churn events adjust the revenue reports according to the existing payment-event rules. Read the [browser attribution API](https://attribiq.com/guides/browser-attribution-api) for the client-side context and the [Payment API reference](https://attribiq.com/guides/payment-api-reference) for server-side revenue events. ## AI-sourced revenue and earlier AI touches AI-sourced revenue means the selected attribution touch for a payment carries verified AI-source evidence under the active attribution model. A first-touch model and a last-touch model can select different sources for the same journey, so the model must remain visible. An earlier AI touch means the same known journey includes an observed AI visit, but another source is selected for the payment. AttribIQ reports those mutually exclusive events as assisted AI revenue, separately from direct AI revenue, so one payment is not counted in both totals. This distinction prevents double counting. One payment should not silently appear as both fully AI-sourced and fully sourced to another channel. ## Observed traffic versus modeled visibility AI visibility products can monitor prompts, brand mentions, citations, or estimated share of voice. Those records describe model outputs, not site visitors. AttribIQ does not give revenue credit to a mention, citation, or visibility estimate. Those records can be useful research, but only a separate observed visit matched through supported journey identifiers becomes revenue evidence here. This is also why AttribIQ is not a general-purpose product analytics suite or a prompt-monitoring index. Its documented role is connecting acquisition context and customer journeys to payment outcomes. ## Unknown and dark AI traffic Not every AI-influenced visit is observable. Referral information may disappear when someone: - opens a link in a private or embedded browser; - copies a URL from an answer into another browser; - moves from one device to another; - uses an assistant that does not send a referrer; - follows a redirect that strips campaign context; or - declines or blocks the local tracking needed for the journey. This traffic can appear direct or unknown. AttribIQ does not classify it as AI traffic without another explicit, supported signal. “Dark AI” is therefore a limitation to document, not a number to manufacture. ## Important reporting limitations - Attribution is a consistent reporting model, not proof that a source caused a purchase. - Ordinary `google.com` referrals remain search traffic unless a distinct supported signal identifies an AI feature. - Referrer hosts and campaign markers should be reviewed when providers change their link behavior. - Cross-device and privacy-restricted journeys may remain unmatched. - Unlike currencies should not be summed unless the report has a supported conversion method. - EU hosting and privacy-minimizing product choices are not legal guarantees. For the product scope, see [AI referral revenue attribution](https://attribiq.com/ai-revenue-attribution). For the broader model, read [What is revenue attribution?](https://attribiq.com/guides/what-is-revenue-attribution). ## Frequently asked questions ### Can AttribIQ measure revenue from AI referral traffic? It can connect revenue when a visit matches a reviewed provider referrer or trusted campaign rule and the later payment carries the AttribIQ visitor or session identifier. Missing referral evidence cannot be reconstructed reliably. ### Does an AI mention or citation receive revenue credit? No. A mention, citation, or modeled visibility record is not a visitor journey. It must remain separate unless a real observed visit can be matched through supported evidence. ### Is ordinary Google traffic classified as AI traffic? No. An ordinary google.com referral is search traffic unless a separate, supported signal identifies an AI search experience. AttribIQ does not infer that distinction from a normal Google referral. ### Can AttribIQ identify all dark AI traffic? No. Some assistants, apps, privacy tools, copied links, and cross-device journeys omit referral evidence. That traffic may appear direct or unknown and should not be relabeled without a supported signal.