AI Traffic Attribution Convention (ATAC), Version 0.1 — Draft for Public Comment

An open convention for classifying, declaring, and reporting AI-mediated traffic and AI-influenced revenue. Current web analytics conventions were designed for a click-based web and misclassify AI-mediated activity in three ways: traffic referred by AI assistants is scattered across “referral” and “direct”; agentic sessions are invisible or miscounted as human; and recommendation-influenced revenue arrives disguised as branded search or direct traffic, systematically credited to the wrong cause. This document defines: (1) a three-class taxonomy of AI-mediated traffic (AI-Referred, Agentic, AI-Influenced) with explicit confidence semantics; (2) ai_ref, a lightweight query parameter by which AI platforms and tools can declare referral provenance; (3) an agent-access extension to the /.well-known/brand.context resource by which site operators declare canonical machine-readable endpoints for agent consumption; and (4) minimum reporting requirements for any party publishing AI-influenced revenue estimates, including the prohibition of point estimates for inferred quantities. The convention is deliberately minimal: it standardises vocabulary, declaration, and honesty requirements, and leaves implementation methods to implementers. It is designed to interoperate with adjacent work rather than duplicate it — consuming cryptographic agent authentication (HTTP Message Signatures, RFC 9421, and the IETF Web Bot Auth effort) as its highest confidence tier, drawing methodological lineage from the Media Rating Council's Invalid Traffic framework, and addressing the organic discovery layer that sits alongside the IAB Tech Lab's agentic advertising transaction standards. Status: Draft for public comment. Adoption is voluntary and free of charge, and requires no licence from, relationship with, or notification to AIVO Standard. Comments are invited from all parties at [email protected] or [email protected]; a versioned change log is maintained at aivostandard.org.