Measuring Brands in AI Just Got More Interesting. The IAB’s “Measuring Visibility in the AI Era”.

On August 3rd something foundational happened in the world of brands in AI. For the past couple of years, CMOs have known AI would eventually become an intermediary between customers and their brands — and a grim reaper, omitting all but a few in its category responses. A remote threat at first, then suddenly a thing to get ahold of — a thing that will impact sales, budgets and bonuses. All the while, agencies and other problem-solvers did their level best to solve for their clients. I’ve lost count of the times I’ve heard “Brands in AI? That’s like SEO or GEO, right?” Or advice to publish something new every week or so, to be seen by AI.

Then, on August 3rd, the IAB dropped “Measuring Visibility in the AI Era”. It is a 37-page explanation for a non-technical audience — specifically for brand leaders and their agencies. The goal: a common vocabulary and framework for measuring AI visibility, and a way for service providers to show how their work fits. The document was produced by a cross-industry working group — brands, agencies, publishers, and measurement providers — led by Caroline Giegerich, VP, AI at IAB.

In full disclosure, I have a system for measuring brand visibility in AI. It is one reason why this topic is so close to me. I’ve been working through it and its related concepts for over 2 years. For me, the IAB’s document was a cool wind in my sails. I began building my measurement tools and method because I felt Marketing was too focused on AI for productivity and efficiency — and not at all on how brands would fare when customers used AI to learn about categories, compare brands and make decisions. The IAB document is an invaluable compendium for CMOs and agencies — and it gives them something they haven’t had before: a way to understand the process of visibility and to ask whether the measurement they’re buying is good enough for the decisions it will drive.

Until now -- and I am imagining here, not being a CMO -- there was no widely agreed framework for judging the quality of those measurements. Methods, success metrics, and definitions of rigor were unique to each problem-solver. Then Giegerich’s working group announced: this is what good measurement looks like today. You can ask your vendors whether they meet these thresholds. Not just a score or a list, rather a framework of observability and provenance for measuring “visibility” — information integrity at the scale of persuasion.

Now, as marketers and agencies return from the slow month, they finally have a new definition of reliable measurement, one that gives the quest for brand visibility in AI what it has never had, an industry-wide framework to gauge progress. It will evolve. The framework may not fit every market or budget, but the principles are widely applicable. Agencies and their clients can certainly begin with questions to service providers: walk us through how your work fits all of this. In fairness, it will take time for some -- if not most -- measurement providers to adjust. And agencies and clients should surely give them room to do that. Also, any agency that advocates for measuring brand visibility in AI systems is already way ahead and a valuable partner. But watch out for offers to self-measure. As in many industries, measurers should not also be the makers.

It is a big moment in the wonky world of measuring brand visibility in AI systems. Niche, but every brand needs it. Measurement, of course, is not new. But what makes this application so interesting is that much of what matters for measuring brands in AI systems is in motion: sources change, organizations grow (or shrink), brands rebrand, the competition organizes, and AI technology improves. If that weren’t enough, AI systems are often wrong about the sources they say they use to synthesize an answer. In my experience, opening them often reveals pages that do not contain what the AI said they did. Brands and agencies that take those initial results at face value can easily end up running down dead ends, never understanding why their efforts have not paid off. A citation is not necessarily evidence. And the numbers change. The cadence of measurement cycles, the volume and intention of prompts, and the traceability of the numbers to the sources that provided them are crucial in establishing confidence. That hasn’t existed before, and that is why this is a big deal.