The Interactive Advertising Bureau is developing a framework to measure and credit ads served to AI agents, a market shift that threatens to sever the link between publishers and the humans who used to click through to their sites. The framework, due out Nov. 12, addresses a core problem: when an AI agent reads a product page, weighs it against competitors, and buys on a user's behalf, no one can agree on who gets paid for that exposure. The publisher whose content shaped the answer? The platform that ran the agent? Nobody, because the evidence trail doesn't exist yet.
"What does it mean to advertise to an agent? One side might think that 'this is an interesting test,' and the other side is like, 'that's deception,'" said Caroline Giegerich, vp of AI at the IAB, who is drafting the framework based on conversations with a working group of tech companies, publishers, agencies, measurement vendors, and brands. She declined to name the companies involved.
The core issue is that traditional attribution signals - UTM parameters, referral data - don't reliably survive the customer journey inside AI platforms. So the framework will try to establish what evidence should count as proof that an ad encountered by an AI agent contributed to a conversion.
Two kinds of AI influence
Giegerich said the framework will likely separate AI impact into two categories: when AI serves something to a user (an awareness or intent layer) and when AI helps to split the decision by that user. The IAB has already released frameworks this month on measuring AI visibility and disclosing AI usage in content.
The biggest hurdle has been assigning credit when AI platforms and tech companies are not known for sharing this kind of data. "Some of that evidence doesn't exist and the IAB, we want to influence those conversations to happen. So, if the evidence doesn't exist-if there's a certain party who could provide it-then this is a good open discussion to have," Giegerich said.
When asked what was the hardest thing for all parties in the working group to agree on, she said "all of them." Publishers argue their content informed the response the AI system serves to the user, and they don't want to be left out of the attribution conversation.
Reciprocity and black boxes
The IAB's work could create a foundation for a better dynamic between publishers and tech platforms, said Tracy Schultz, head of global data partnerships at Captora, who leads a network of B2B publishers, brand sites, and data providers. "Many big tech partners have grown big audiences off of publishers, and there has famously been very little reciprocation. If there's a framework that can attribute credit, then it directly speaks to reciprocity," she said.
But even an industry standard won't necessarily open the AI platforms. Mike Bishop, co-founder of AI native advertising platform OpenAds, pointed out that AI platforms may remain black boxes for external measurement firms. If companies like OpenAI control the signals for how AI influenced marketing outcomes, outside firms may have to rely on integrations or data supplied by those same platforms.
"If you look at how those measurement vendors worked with Facebook, for example, they were actually running their own JavaScript tags. Facebook controlled the integration and effectively had the measurement vendors testing to verify that integration-so the black box was maintained, and the vendors were basically acting as a auditor trust layer," Bishop said. "And it looks like that pattern is repeating with AI platforms."
The framework will need to answer some hard questions, Bishop said: what is being measured, who is doing the measurement, how it's being measured, and at what layer that measurement happens. For developers and IT professionals building on these platforms, the answers will determine what data is available to them through APIs and integrations. The IAB's work on AI Agents & Automation and related measurement standards will shape what signals become available to technical teams trying to build attribution into their own systems.
Why this matters for IT and development
For engineers and technical teams, this framework will determine what telemetry you can actually get from AI platforms. If the IAB succeeds in pushing for standardized measurement signals, developers building on OpenAI, Anthropic, and other platforms may gain access to data that currently doesn't exist in their analytics pipelines. If it fails, the black box pattern Bishop described will persist, and your attribution systems will keep relying on whatever limited integration points the platforms choose to expose. The Nov. 12 release is worth watching for anyone whose code depends on understanding where traffic and conversions actually come from.
Your membership also unlocks:
