The IAB Says Publishers Can Claim Credit for AI-Shaped Sales. The Proof Sits Inside the AI Platforms.
For two years the attribution problem in AI search has been described from one side. Brands lose sight of the moment an assistant recommends them. The sale turns up days later as branded search or direct traffic, and the dashboard thanks the wrong channel.
On 12 November the Interactive Advertising Bureau published a framework that finally names the other side of that transaction. The AI-Era Attribution Blueprint is written for brands, agencies, AI platforms and measurement providers, and it lists publishers among them. That is the part worth slowing down on. A standards body has now said, in writing, that the content an AI read to build its answer is a participant in the sale.
What it has not done is give anyone the evidence to prove it.
What did the IAB release on 12 November?
The AI-Era Attribution Blueprint is a shared framework for classifying the role AI plays in a conversion and for evaluating claims for credit (reported, IAB). According to the IAB's own summary, it contains three things: definitions for three AI influence pathways to purchase, a framework for deciding who can claim credit and what evidence each claim needs, and guidance on the evidence that exists today alongside the measurement gaps that remain (reported, IAB).
It is the second half of a pair. In August the IAB published Measuring Visibility in the AI Era, which covered organic visibility inside AI answers and explicitly left commerce attribution out of scope, flagging it as future work (reported, IAB). The new blueprint is that future work. The IAB's accompanying webinar describes it as building on the visibility framework and extending it into commerce (reported, IAB).
The full report sits behind a free IAB account. The detail below on claim types comes from the IAB's published summary and from excerpts quoted in trade coverage, and is flagged accordingly.
What are Human-Owned, AI-Assisted and AI-Executed conversions?
The blueprint's core move is to stop treating "AI influence" as one thing. It defines three pathways to purchase (reported, IAB):
- Human-Owned: the person researches, chooses and buys. AI may be nearby but is not the decisive influence.
- AI-Assisted: an AI system shapes the decision, for example by recommending a product, but a human completes the purchase.
- AI-Executed: an agent completes the transaction on the person's behalf.
The definitions matter because they separate service from influence. As excerpted in trade coverage, the blueprint's position is that executing a purchase does not by itself earn a claim: an agent reordering a product the shopper already picked is performing a service, while an agent choosing which product to buy may qualify for a decision claim (reported, blueprint excerpt via trade coverage).
Why does attribution break when AI recommends a product?
Because the signals attribution was built on do not survive the journey. A shopper asks an assistant, reads the answer, closes the tab and later searches the brand by name. No click, no referral ID, no UTM parameter links the recommendation to the order.
The IAB's framing is that AI-influenced sales do not vanish from reporting; they get credited to whatever channel the dashboard can observe, and the next budget follows that channel (reported, blueprint excerpt via trade coverage). Search captures demand that an AI answer created. Advertisers then pay search to capture it again.
Where do publishers fit in the blueprint?
This is the new ground. As excerpted in trade coverage, publishers whose reviews or reporting informed an AI's choice may qualify for contributory claims, separate from the decision claim an agent might hold (reported, blueprint excerpt via trade coverage).
That did not happen by accident. When Digiday reported on the framework in development, the IAB's VP of AI, Caroline Giegerich, said publishers had pressed to be included, on the argument that their content fed the answers AI systems serve (reported, Digiday).
Two limits come with it. Qualifying for a claim is not the same as being paid for one, and the blueprint does not set how much credit any participant receives or establish incremental value (reported, blueprint excerpt via trade coverage). More importantly, much of the evidence needed to substantiate a claim remains inside the AI platforms (reported, blueprint excerpt via trade coverage).
What does this mean for publishers?
It gives publishers a vocabulary for something they have been arguing without one. "Our content shaped that answer" is now a recognised category of claim, not a grievance.
But a contributory claim is only as strong as the record behind it. A publisher cannot point to a click that never happened, and most publishers cannot see the read at all. Live Search Agents, dispatched when a user asks ChatGPT, Perplexity or Claude a question, do not execute JavaScript, so no analytics fire and no ad tags load on the visit that actually informed the answer. The AI bot share of publisher traffic is already material: roughly 1 in 50 publisher visitors is an AI bot (named-party figure, OPA Europe / Rossel, Jan 2026).
The practical consequence: publishers who log and verify agent reads at the edge will have an evidence trail to bring to any future claims process. Publishers who rely on page-level analytics will have nothing.
What does this mean for advertisers?
For brands, the blueprint is a warning about their own reporting. If AI-Assisted conversions are being booked as search or direct, search looks more efficient than it is and AI visibility looks worthless when it may be doing the heavy lifting.
It also sharpens the GEO question. If the AI answer is where intent is created, brand visibility in LLM answers is upstream of the channels currently getting the credit. Display, paid search and social do not appear inside the answer. Being present in the content the assistant retrieves, and being able to evidence that presence, is the part of the funnel the blueprint says current measurement misses.
What is still missing?
The evidence layer. The blueprint is a set of rules for evaluating claims; it is not a data feed. The IAB itself lists current measurement gaps as a section of the report (reported, IAB), and the platforms that hold retrieval and grounding logs have not committed to share them. Until they do, every claim, from brand, agent operator or publisher, rests on inference.
the blankspace read The IAB has just conceded the point publishers have been making for two years: the read creates value even when no click follows. That is the right principle and the wrong place to stop. A contributory claim that depends on the AI platform's own logs is a claim the platform can decline to evidence. Publishers should not wait for that cooperation. The read happens on their infrastructure, at the edge, before any answer is generated. Measure it there, monetise it there, and the attribution argument stops being about who gets credit later and becomes about who got paid at the moment of the read.
Key figures
| Figure | Value | Provenance |
|---|---|---|
| AI-Era Attribution Blueprint publication date | 12 November 2026 | reported, IAB |
| AI influence pathways defined | 3 (Human-Owned, AI-Assisted, AI-Executed) | reported, IAB |
| Intended users of the framework | brands, agencies, publishers, AI platforms, measurement providers | reported, IAB |
| Predecessor framework, Measuring Visibility in the AI Era | published 3 August 2026; commerce attribution out of scope | reported, IAB |
| Publisher claim type | contributory claim, where content informed the AI's choice | reported, blueprint excerpt via trade coverage (full report gated) |
| Execution alone | earns no claim under the blueprint | reported, blueprint excerpt via trade coverage (full report gated) |
| Share of publisher visitors that are AI bots | about 1 in 50 | named-party figure, OPA Europe / Rossel, Jan 2026 |
Primary source, opened and verified: IAB, The AI-Era Attribution Blueprint (12 November 2026)