On 3 August 2026 the Interactive Advertising Bureau published Measuring Visibility in the AI Era, the first serious attempt to impose a shared vocabulary on a market that had been inventing its own. The framework is careful, well built and backed by measurement teams at Walmart, Acxiom, Microsoft, WPP Media, eMarketer and Tinuiti. It also draws a boundary that anyone buying AI placement should read closely.
Twenty vendors, twenty answers
The IAB counted more than twenty vendors selling AI visibility measurement, with little consistency in method and less in result. Two tools pointed at the same brand and the same question could return numbers that did not resemble each other. When the underlying methodology is undisclosed, a marketer cannot tell whether a rising line means the brand is doing better or the sample changed.
The stakes are not academic. McKinsey estimates that brands failing to adapt to AI discovery could see traffic fall by as much as half. Against that, only 16 percent of brands systematically track how they perform in AI search at all. Most of the market is flying on instrument readings it cannot calibrate.
Four questions, in order
The framework organises measurement into four principles, deliberately ranked. Presence asks whether the brand appears at all: mention rate, citation rate, share of voice, momentum over time. Prominence asks where it appears and in what order. Portrayal asks what the answer actually says: sentiment, framing, and the rate at which the model states something factually wrong about the brand. Persuasion asks whether any of it moved someone: recommendation strength and post citation click through.
The ordering is the useful part. A brand that is mentioned often but described badly has a Portrayal problem, not a Presence problem, and buying more coverage will not fix it. Hallucination rate sitting inside the standard as a named metric is an admission worth noticing: the industry now treats being described incorrectly as a measurable brand risk rather than an amusing anecdote.
The IAB also splits measurement into two tiers. Directional readings are early signals, fit for internal orientation only. Decision grade readings carry stated sample size, query volume and reproducibility, and only those should move budget. The guidance to vendors is blunt: differentiate on rigour rather than claims.
The line the framework does not cross
Here is the part that matters for anyone spending money inside AI answers. The framework, as published, excludes paid advertising measurement. It is a standard for organic visibility. The moment a brand buys a placement inside an AI answer, it leaves the area the standard covers.
That is not an oversight, and it is not a criticism of the IAB. Paid placement inside generative answers is roughly six months old as a real channel, and the platforms themselves publish almost nothing about how it behaves. A standards body cannot codify what its members cannot yet observe.
But it leaves a practical gap. A brand can now measure its organic AI presence against an agreed method, and has nothing equivalent for the paid side. In practice that means the two halves of an AI visibility strategy are reported in different languages, and the paid half is the one without a dictionary. Until that changes, the burden of building a defensible measurement story for GEA sits with the advertiser and whoever runs the campaigns, not with the standard.
Sources: Marketing Dive, IAB shares playbook for measuring brand visibility in AI powered platforms