Why attribution modeling needs to be rebuilt for an AI-driven funnel

Attribution modeling in marketing was designed around a linear funnel: a user becomes aware of a brand through advertising, considers it through content or comparison, and converts through a transaction. The touchpoints in this funnel were predominantly owned and trackable — paid ads, owned content, email, direct website visits. Current attribution models are optimised for this architecture, which increasingly does not match how consumers actually find and evaluate brands.

AI search tools have introduced a touchpoint that most attribution systems cannot see. A user who discovers a brand through a ChatGPT recommendation, visits the website directly, and converts registers in most attribution systems as a direct visit with no prior touchpoint. The AI recommendation that initiated the journey is invisible. Brands that rely on these attribution systems to understand their acquisition economics are systematically missing one of the most important awareness channels operating on their behalf.

What rebuilding attribution for an AI funnel requires

The short-term fix is supplementary research: regular surveys of new customers asking how they first discovered the brand, tracked over time to detect whether AI recommendation or AI search appears in the discovery path. This is imperfect and resource-intensive, but it surfaces information that click-based attribution cannot.

The medium-term requirement is building content and a brand presence specifically designed to appear in AI-generated recommendations — which means investing in GEO (generative engine optimisation), structured data, authoritative content, and brand authority signals that AI models weight. A brand that invests in being recommended by AI tools will see the attribution gap close over time, because more of its traffic will arrive through trackable channels as the AI recommendation builds brand familiarity rather than serving as the sole touchpoint.