What changed in the process
Formalizing the practice required building measurement infrastructure that did not exist off the shelf: tracking citation frequency across multiple AI assistants, distinguishing that signal from ordinary seasonal noise, and connecting it back to the channel-specific discovery features that a purely web-focused approach would have missed.
The team also says internal expectations shifted over the course of building the practice. What started as a narrow reporting fix became a recognition that AEO measurement needed its own standing discipline, connected to the agency’s existing organic and paid acquisition work but scoped separately.
The early tooling problem
Part of what slowed the initial investigation, the team says, was that no existing tool answered the specific question they had. Not how a site ranks, but whether an AI system mentions the brand when asked, and why. Building that answer meant querying multiple assistants directly and manually, on a repeated schedule, before anything resembling automated tracking existed. Some of that manual process still underpins the methodology today, even as parts of it have been automated. The team says the manual groundwork, tedious as it was, gave them an unusually granular early view of how citation behavior varied across assistants, one that off-the-shelf tools built later did not initially replicate.
Read the full article here
