The IAB plans to publish a framework on November 12 for measuring and crediting conversions influenced by AI. Its stated purpose is a shared model for cases where AI sits between an advertiser or publisher and a customer, while traditional signals such as referral data and UTM parameters may be reduced or absent. digiday.com
That is directly relevant to GEO work. Teams can already observe parts of AI-mediated discovery: whether a brand or publisher appears in AI answers, whether AI sources send visits, and whether those visits convert. The hard problem is proving how those observations relate to one another when an AI system researches, compares or helps transact without preserving a conventional click path. The proposed IAB work is important because it names that gap. It should not be mistaken for a solved attribution system before the framework is published. marketingtechnews.net
The actual proposal is narrower than “measuring AI”
The framework described by IAB’s vice president of AI, Caroline Giegerich, is about classifying and crediting conversions influenced by AI, whether or not an advertisement was served. It is not simply a standard for counting AI referrals, measuring brand mentions in answers, or proving that an AI platform caused a sale. digiday.com
The IAB is drafting it with a working group of technology companies, publishers, agencies, measurement vendors and brands. The participants have not been named. That matters: the eventual usefulness of a common framework will depend not only on its definitions but also on whether the systems holding relevant data participate in the resulting measurement process. digiday.com
The reported model will likely divide AI impact into two categories:
- AI surfaces something to a user, at an awareness or intent layer.
- AI helps the user make a decision.
That distinction is sensible for GEO reporting because being included in an answer and being part of a decision are different events. A brand can be visible in an AI response without being selected; a publisher’s material can inform a response without generating a referral; and an AI agent can contribute to a purchase without encountering the ad or content exposure that influenced the eventual buyer. marketingtechnews.net
Why conventional attribution loses the trail
The IAB’s core use case is not difficult to understand. An agent might read a product page, weigh alternatives and help a user buy without creating the trail that follows a direct click. In that situation, traditional signals do not necessarily survive the journey well enough to establish who influenced the conversion. mediacopilot.ai
The more consequential implication for GEO is that an observed downstream visit is only one type of evidence. Adobe data cited in the reporting measured visits arriving at US retail websites from AI sources; it found AI-sourced traffic rose year over year in the first three months of 2026 and converted at a higher rate than non-AI traffic in March 2026. But that is measurement of identifiable AI-referred visits, not proof of every AI interaction that may have influenced a purchase. marketingtechnews.net
Those are separate questions, and conflating them is an easy reporting error. A referral can be counted when it exists. Influence may happen without a referral. A framework for the latter therefore needs to say what evidence qualifies for credit, how much credit it earns, and which party can verify it.
The extracts do not provide the IAB’s future rules for any of those decisions. They establish the release date, the likely two-part model and the unresolved evidence problem. They do not establish a conversion-credit formula, a required data schema, a platform participation requirement, or an independent validation method. Until the framework appears, claims beyond that are speculation.
Visibility is an input, not a conversion claim
This attribution project follows separate IAB guidance on measuring visibility in AI systems. That earlier work addresses whether brands, products, publishers or content appear in AI-powered discovery, and it sets out a metrics hierarchy plus a distinction between decision-grade and directional measurement. The reporting says more than 20 companies offer AI visibility tools and that different methodologies can produce different results for the same organisation. marketingtechnews.net
For GEO practitioners, that is a useful boundary. Visibility measurement answers questions such as whether an entity appears in an AI discovery environment. The forthcoming attribution work is intended to address how AI involvement connects to a conversion and what evidence supports assigning credit. Neither category substitutes for the other. marketingtechnews.net
I would therefore keep reporting in distinct layers rather than roll them into a single “AI performance” figure:
- Visibility: appearance-related measures such as AI inclusion rate, citation share and depth, corpus presence, and rank stability, which IAB summit participants identified as emerging measures.
- Observable traffic and outcomes: referrals and conversions that reach a site with an identifiable source.
- Attributed influence: conversion credit assigned to AI involvement under a disclosed methodology and evidence standard.
The first layer can show whether content or a brand is being surfaced. The second can show the measurable result of visits that arrive. The third is the unresolved layer the IAB is attempting to structure. This separation prevents a visibility uplift from being presented as incremental revenue evidence before the measurement connection has been established. marketingtechnews.net
The implementation constraint is platform-controlled data
The biggest limitation is not taxonomy; it is access to evidence. Giegerich said that some evidence needed for attribution does not exist, and that the IAB wants discussion with parties that could potentially provide missing information. digiday.com
Michael Bishop of OpenAds made the related point that even an industry standard may leave AI platforms as black boxes. If platforms control the signals showing how AI influenced outcomes, external measurement firms may have to rely on integrations or data supplied by those platforms. He compared that possibility to a model in which the platform retains control over underlying measurement systems while outside vendors serve more as an auditor layer. digiday.com
This is why a shared framework could improve comparability without automatically delivering independence. It can define what evidence should count and reveal which evidence is absent. It cannot force a platform to expose data merely by naming it as desirable. The sources do not show that AI platforms have committed to supplying the signals the framework may require.
Publisher credit makes the limitation particularly visible. Publishers argue that their material may have informed an AI response and that they should not be excluded when credit is assigned. That is a legitimate attribution question, but the existence of an informed response does not itself settle the amount of credit due or prove a causal contribution to a specific sale. mediacopilot.ai
What changes for GEO measurement now
The near-term change is methodological discipline, not a new dashboard metric. GEO teams should document which layer each result belongs to, retain the methodology behind visibility measures, and avoid treating AI referral performance as a proxy for the unobserved journeys the IAB is trying to address.
There is also a useful standard to apply to vendors and platforms: ask what event was observed, who observed it, what data connects it to an outcome, and whether the methodology is directional or decision-grade. Those questions follow directly from the measurement issues identified in the IAB visibility work and the attribution debate. marketingtechnews.net
The IAB effort could provide a common vocabulary for conversations that currently mix discovery, traffic, content contribution and conversion credit. That would be meaningful progress. But its published framework should be evaluated by its treatment of missing evidence, data provenance and platform opacity—not by the fact that it offers an AI attribution label. The central uncertainty is already explicit: when AI intermediates a journey, the evidence trail may not exist. digiday.com