IAB publishes a framework for crediting AI with a sale, and the data to check it mostly doesn't exist

The Interactive Advertising Bureau has published a framework for deciding who, or what, gets credit when an AI assistant helps close a sale, sorting every AI-touched purchase into one of three buckets — human-owned, AI-assisted, or AI-executed — and spelling out the evidence a company needs before it can claim one. The document, which IAB calls the AI-Era Attribution Blueprint, follows months of work with a group that included Microsoft, Criteo, Walmart and Prudential, according to Digiday. It lands while software marketers are already being asked to justify budgets aimed at ChatGPT, Gemini and other AI assistants, with no agreed way to prove those channels actually produced a result.

What the framework actually does

The blueprint sorts a sale by who created the buying intent, who made the purchase decision, and who executed the transaction. A decision claim goes to whichever party actually influenced or made the purchase call; a contributory claim goes to a party that shaped the outcome without making the decision. IAB frames this as a supplement to the attribution systems marketers already run, not a replacement, and it says plainly that the framework alone doesn't fix the lack of observable data behind most AI-influenced purchases, according to Martech.org.

Caroline Giegerich, IAB's vice president of AI, put the underlying problem to Martech.org this way:

"The clicks and referrals that attribution depends on are thinning out or disappearing."

That's the gap the three-way split is trying to paper over: fewer clicks to measure, more purchases where an assistant quietly did the influencing.

Who this affects

This hits anyone who has to report on AI-driven revenue: marketers running ads inside ChatGPT or Gemini, paid-search teams watching branded search soak up credit that may not belong to it, and any CMO defending AI spend in a budget review. IAB's own research found 45% of ad buyers name comparing AI-driven and traditional customer journeys as their single biggest measurement problem, and 44% are using branded search and direct traffic as rough stand-ins for AI influence they can't directly track, per the organization's "2026 Outlook Study: September Update," cited by Martech.org.

Software vendors selling AI-visibility and GEO tools are squeezed from both sides of this. They need the IAB's categories to make their own reports defensible to skeptical buyers, yet most of them are themselves running on the same thin signals — branded search, direct traffic, self-reported referral strings — that the framework says can't be trusted as proxies. A vendor whose dashboard claims precise AI-attributed revenue today is almost certainly estimating it from exactly the gaps this framework was written to flag.

The size of the gap shows up in the numbers IAB compiled from outside vendors. Similarweb data in the report found a consumer who gets a brand recommendation from an AI assistant is 2.5 times more likely to visit that brand's site within a week — but only 8.8% of those visits get credited to an AI referral directly, with 55.9% attributed to search and 19.9% to direct traffic, per Digiday. Separately, measurement vendor Profound found that one in three consumers had a purchase-relevant AI conversation in the past month that went completely unattributed, a figure that excludes Google's AI Mode and AI Overviews entirely.

Why most of it can't be checked yet

Assigning a claim is the easy part; verifying one is not. IAB says much of AI's influence on a sale can't be measured today because the recommendations, decisions and transactions happening inside a chat interface are invisible to advertisers and their measurement vendors, per Digiday. Jaime Schultheis of data provider Bombora summed up the publisher side of it to Digiday:

"Publishers are feeding the answers but they don't have the receipts."

The UK's Competition and Markets Authority has ordered Google to start giving UK publishers AI search clickthrough data by December 3 — a narrow opening forced by a regulator, not offered voluntarily, and it says nothing about what advertisers themselves get to see.

• Platforms are asked to disclose whether a recommendation was paid or organic, and name the sources behind it
• Brands and agencies are asked to apply one claim-and-evidence standard across every AI surface, rather than trusting each platform's own numbers
• Publishers are asked to document where their content shaped an answer, and what evidence supports that claim

What a software marketer should do differently

In practice, that means treating any dashboard that claims to measure "AI revenue" with the same skepticism a software marketer already applies to multi-touch attribution models: ask what counts as a touch, who defined it, and whether the platform reporting the number has an incentive to report a bigger one. IAB's framework gives the industry shared vocabulary for that argument for the first time. It does not yet give anyone the receipts to settle it.