Agentic search — journeys where AI agents research, compare and sometimes complete purchases on a user’s behalf — is early, and measurement hasn’t kept up. Similarweb says more than 90% of AI-driven sales are currently being miscredited as branded search, direct traffic or retail media.
The IAB has put forward a starting point: the AI-Era Attribution Blueprint. It creates three categories for AI-attributable conversions, each defined by what the AI actually did in the buying journey.
Human-owned, AI-assisted or AI-executed
The blueprint asks marketers to classify a conversion by the level of AI involvement:
- Human-owned: AI may have been involved in decision-making, but it did not substantively create purchase intent, influence the product decision or carry out the purchase.
- AI-assisted: AI narrows down and suggests options. The IAB breaks this into “AI-influenced,” where AI shapes intent but a human makes the final choice, and “AI-recommended,” where AI shapes the choice itself.
- AI-executed: A human gives an agent permission to purchase. This covers both a specific “buy this” instruction and an agent-selected purchase where the AI makes the selection.
The definitions are deliberately descriptive rather than measurement-ready. Caroline Giegerich, IAB’s VP of AI, said on a webinar this week: “We define this based on what’s actually happening, not what we can measure, per se, today.”
Why it matters
If AI-influenced and AI-executed sales continue to be counted as branded search or retail media, budgets will keep flowing to channels that may not be doing the influencing. That becomes more costly as agentic search and commerce take a larger share of the purchase journey.
The blueprint is not granular enough for programmatic bidding yet. It doesn’t dictate tracking parameters or a conversion API. Instead, it offers a vocabulary that can travel across DSPs, retail media networks and agent platforms as they build measurement.
For brands and agencies, that makes it an operating framework rather than a report. Teams can use it to pressure-test their attribution models: if a sale currently shows up as branded search but a consumer used an AI assistant to compare options, which label fits? The answer changes budget allocation.
What marketers can do now
Treat the blueprint as an internal classification layer rather than a replacement for your multitouch model. Start by mapping where AI touchpoints appear in your conversion paths, then tag a sample of sales using the three categories. The goal is to build a common language now, so the eventual measurement changes are less disruptive.




