Attribution was built on a trail of clicks and referrals. When a shopper asks an AI assistant which stand mixer to buy and then types the brand name into a search box, the trail points at search. The Interactive Advertising Bureau released a framework on Nov. 12 meant to sort those sales by how much of the decision an AI made, and four sets of figures published in the past five weeks show how much of the evidence it asks for is still missing.
What the IAB published
The AI-Era Attribution Blueprint is a classification scheme that sits next to existing attribution systems. The IAB says it was written by a working group of brand, agency, publisher, AI platform and measurement leaders, and that it extends the trade body’s earlier Measuring Visibility in the AI Era playbook from how brands show up in AI platforms to how that presence connects to sales.
The framework asks three questions of every conversion: who or what created the intent, who or what made the decision, and who or what completed the purchase. The answers put each sale into one of three pathways:
- Human-Owned: a person created the intent, decided and bought.
- AI-Assisted: an AI shaped the journey along the way, and a person still made the purchase.
- AI-Executed: an AI completed the conversion.
The pathway decides which parties may file a decision claim or a contributory claim, if any, and what evidence each claim needs. The IAB draws a line between a claim and the credit itself: “A claim is not credit; it is the right to be evaluated for credit.” Brands, agencies, AI platforms and publishers each have a role, and a publisher, for example, can identify the contributory claims its content creates. Every claim then has to meet the evidence its pathway requires.
“As AI takes on more of the purchase journey, from sparking interest to completing the sale, the clicks and referrals that attribution depends on are thinning out or disappearing,” said Caroline Giegerich, VP, AI, at the IAB. “AI-influenced sales are sometimes reported as search or direct traffic, which can send the next marketing investment to the wrong channel.”
Why the IAB says it needed one
The trade body tied the release to its own survey. In the 2026 Outlook Study: September Update, based on more than 200 brand and agency decision-makers, ad buyers named comparing AI-driven and traditional customer journeys as their top measurement challenge, at 45%. Another 44% said they rely on branded search and direct traffic as proxies for AI influence they cannot track.
That second number describes the problem Giegerich named. If a person reads an AI recommendation and then searches for the brand, the search click gets the sale in a last-touch model and a share of it in most multi-touch models. The AI answer that produced the brand search gets nothing, because it left no click to count. Budget follows the credit, so the search line grows and the AI line stays invisible.
David Cohen, CEO of the IAB, framed the blueprint as a starting point. “The AI-Era Attribution Blueprint is honest about what the industry can’t yet substantiate, and we’re inviting everyone in the ecosystem to apply it and help make AI’s role in commerce more observable, verifiable, and measurable,” he said.
How big each pathway looks today
The IAB release gives no shares for the three pathways, and it asks brands to estimate their own. Three recent data sets give a rough sense of the split, each with its own limits.
Shoppers mostly use AI to decide, then buy themselves
Salesforce published a Consumer Shopping survey on Nov. 12 of more than 2,500 shoppers in the US, the UK, Canada and Australia. Almost 20% said AI is now their first stop on the way to the store, and 74% said they trust the recommendations it gives them. Thirty percent of AI shoppers said these tools frequently introduce them to brands they did not know, and 11% said they would skip a purchase if an AI assistant advised against it.
Fewer are ready to let the agent pay. Thirty-one percent of consumers have tried a personal AI agent, and only 19% of those users have deployed one at checkout. About 73% of shoppers are not comfortable giving a personal agent their payment information, and 49% are not interested in agents browsing or buying on their behalf at all. Of the agent users who do let agents buy, 36% have set up a separate credit card for the agent.
Mapped onto the IAB’s scheme, that is a large AI-Assisted pathway and a thin AI-Executed one. Salesforce also projects that AI will drive $75.8 billion in Cyber Week spending, up 13% from 2025, and that AI-driven recommendations will influence more than one in five holiday orders. The company does not publish the method behind those projections, so they are best read as Salesforce’s estimate of the AI-Assisted share across its own commerce data.
“Everyday shoppers are asking AI what to buy, where to buy it, and whether it’s worth the money before they ever land on a retailer’s site,” said Caila Schwartz, head of agentic commerce shopper insights at Salesforce.
Agents reach checkout pages, with a person often still present
Microsoft Clarity counted more than 5.2 million sessions involving AI agents across 230,511 sites between Aug. 13 and Sept. 9, according to its agentic analytics study. Checkout appeared in 25.21% of those sessions, against 14.65% of human-only sessions, a gap we examined when Clarity’s 5.2 million agent sessions ranked checkout as the top action.
Clarity’s own caveat matters for the blueprint. An agentic session may include a human handoff, and Clarity says it cannot currently separate the agent and human portions into different sessions, so a recorded checkout does not prove an agent completed the purchase. That is exactly the distinction between AI-Assisted and AI-Executed, and the analytics layer cannot yet draw it.
Brand agents talk to buyers, and the outcome rarely reaches the ad platform
The third data set covers the brand’s own agents. Invoca’s Broken Funnel Report, released on Nov. 10, surveyed 451 US B2C marketing leaders between Sept. 14 and Oct. 5. Eighty-six percent said they can identify leads who started their research in an AI answer engine. But 74% of those running AI agents cannot fully track an agent-handled lead from first click to conversion, only 26% can tie agent conversations to the campaign or ad that drove them, and 19% reliably send conversation outcomes back to their ad platforms. We covered the survey in Invoca Survey Finds 74% of Marketers Cannot Track AI Agent Leads to a Sale.
“Too many consumer brands measure their AI agents like a help desk and expect them to perform like a sales team,” said Gregg Johnson, CEO of Invoca, which sells software to connect those conversations to ad data.
Where the evidence sits
The blueprint is direct about the gap. Much of the evidence needed to support claims is either not captured or sits inside AI platforms, the IAB says, and it gives each part of the market a job:
- Brands can estimate the share of their conversions in each pathway.
- AI platforms can disclose whether recommendations are paid or organic and which sources informed them, and can label AI-originated traffic consistently.
- Agencies can apply one claim-and-evidence standard across platforms, separating platform-reported claims from independently verified ones.
- Publishers can identify the contributory claims their content creates and the evidence needed to support them.
Read against the three data sets, each request lines up with a missing field. Consistent labelling of AI-originated traffic is what Clarity needs to tell an agent session from a person who switched on an assistant halfway through. Disclosure of the sources behind an answer is what a publisher needs before it can file a contributory claim. Sending outcomes back to ad platforms is a step only 19% of Invoca’s agent-running respondents reliably take.
In our reading, the AI platform requests carry the most weight, and the IAB has the least power to make them happen. The framework asks platforms to disclose and label, and it has no mechanism to make them. The pieces brands and agencies control, their own pathway estimates and a single evidence standard, can start now. The same dependency showed up in ChatGPT ads measurement, where one sale can count twice across two reporting routes.
What it means for the marketing leader
The blueprint gives marketing teams vocabulary before it gives them data. That is still useful, because most of the immediate risk sits in how budgets are argued. A search team reporting strong branded search growth this quarter may be reporting AI-Assisted sales. A display or content team arguing for AI visibility work has no claim to file under last-touch rules. The three-pathway split lets both sides describe the same sale without either one losing it.
It also changes what to ask vendors. A platform that reports AI-driven conversions should be able to say which pathway those conversions fall into and what evidence supports the claim. Under the blueprint’s own terms, a platform-reported claim and an independently verified one are different things, and agencies are asked to keep them apart.
What to do before the 2027 plan is final
- Estimate your pathway split. Add a post-purchase survey question asking whether the buyer used an AI assistant to choose the product, run it on a sample of orders, and split the answers into the IAB’s three pathways. Salesforce’s 19% checkout figure suggests most AI-touched orders will land in AI-Assisted.
- Watch branded search and direct as one signal. If branded search rises while non-brand search and paid social stay flat, test whether AI answers are the cause before giving search the budget.
- Send agent outcomes back. If your brand runs a sales or service agent, create a conversion action for agent-closed sales in each ad platform and carry the campaign ID into the conversation.
- Write the evidence rule down. Decide which claims count as verified, which stay platform-reported, and record it once, the discipline argued in Keep One Written Copy of Your Conversion Definition.
- Ask AI platforms the IAB’s questions. In every 2027 media negotiation with an AI platform, ask whether recommendations are labelled paid or organic, which sources informed them, and how AI-originated traffic is tagged when it reaches your site.
The IAB says it will bring brands, agencies, AI platforms, retailers and publishers together to share what they learn as more data becomes available. Until platforms label their traffic, a brand’s own survey and conversion data will carry most of the weight in sorting its sales into the three pathways.
Source: IAB
