As agentic auctions and cross-channel attribution remain undefined gray areas for audio, the Media Rating Council is moving to define standards for AI’s fastest-moving corners as part of what it calls step one of a two-part effort.
The guidance, published July 8, was developed through MRC’s engagement with measurement services and the audit firms conducting its accreditation reviews, alongside supporting organizations including the ANA, the 4As and IAB Tech Lab.
This first release pulls together standards MRC says already apply to AI in measurement. The group says that guidance existed before but was never organized or labeled specifically for AI. Step two will bring new standards to cover the gaps MRC has identified. That work started in the first quarter of 2026 and isn’t expected to wrap until early 2027.
Nine principles, among them fairness, transparency and accountability, get tied to rules already on the books, spanning documents like MRC’s Minimum Standards, Invalid Traffic Detection Addendum and Auction Transparency Standards. An appendix lines up dozens of existing rules against those principles, the point being there’s no AI loophole where current rules don’t already reach.
The gaps are where things get more consequential. Six priority areas make up the next phase: general AI governance, invalid traffic, brand safety, base digital measurement, identity and big data, and auctions.
Some of those areas have no defined metrics yet, and others need new minimum requirements MRC hasn’t written. Audio’s two growth areas, podcast measurement and cross-channel outcome effectiveness measurement, covering brand lift, attribution and creative return on investment, both fall into that unwritten category.
Auditors face a separate list of risks they can’t fully evaluate yet: bias in AI training data, telling organic content apart from generative content during model training, and what agentic AI, and eventually artificial general intelligence, means for invalid traffic detection.
A handful of gray areas round that out: human-involved agentic activity, zero-click measurement and monetization, and how answer engine optimization and AI scraping affect invalid traffic classifications at the property level. None of it has standards attached, a gap with direct bearing on podcast download counting and any audio property competing for discovery outside traditional platforms.
Until that work is finished, part of the burden shifts to measurement users. The guidance closes with a list of questions MRC recommends buyers, sellers, and intermediaries ask their providers: where AI shows up in a given workflow, what training data the models rely on, how methodology changes get disclosed, who can access input and output data, and whether the provider’s process has MRC accreditation.
