The Comms Industry Just Got a Rulebook for Measuring AI Visibility. Marketing Should Steal It.
AMEC has published a three-part framework for measuring brand presence inside AI search and chatbots. The principles target communications teams but the methodology applies to any marketer trying to prove ROI on generative engine optimisation.
Vanity dashboards that count brand mentions inside a single AI tool on a single day are not measurement. They are theatre. AMEC has finally put that distinction in writing.
AMEC, the global body for communications measurement, has launched a formal set of principles for measuring brand presence inside AI search and generative engines. The framework, called the AMEC GEO Principles, splits AI visibility into three connected layers and introduces baseline evidence standards that any practitioner is expected to meet before claiming a measurement result. A companion document, A Practitioner's Guide to GEO Measurement, ships alongside the principles.
The three layers are the useful part. Upstream reputation signals cover earned coverage, third-party commentary, reviews, expert content and owned assets that AI models train on or retrieve from. Search and content readiness covers whether the digital footprint is credible, accessible and structured enough for AI systems to interpret. Downstream AI outputs cover how the brand actually appears in AI-generated answers, including citations, framing, omissions and reputational risk.
The evidence standards matter as much as the framework. AMEC has spelled out that any GEO measurement claim needs repeatable prompts, documented methods, transparent assumptions and clear limitations. The intent is to stop the current practice of running a few prompts in ChatGPT and presenting the screenshots as a measurement system.
Why it matters
Australian businesses spending on GEO right now are operating without standards. Some teams track brand visibility in ChatGPT. Others track Perplexity. Most track nothing systematically and rely on anecdotes from clients or executives who searched for the brand and did or did not see it. The result is a market full of GEO services with no agreed way to compare performance.
AMEC's framework changes the conversation. A marketing director can now ask a GEO vendor or agency three questions and learn most of what they need to know. Which layer are you measuring? What is your prompt methodology? What are the documented limitations of your data?
The upstream and readiness layers also matter for SEO teams. The same content depth, citation profile and structural cleanliness that wins AI citations is the foundation of every other organic visibility play. The AMEC framework just gives the work a shared vocabulary.
The AMEC GEO Principles split AI visibility into upstream reputation signals, search and content readiness, and downstream AI outputs
What to do about it
Request a written GEO measurement methodology from any agency or vendor pitching AI visibility services. If they cannot tell you which AMEC layer they are measuring or share their prompt set, treat the engagement as exploratory not strategic.
Map your existing measurement coverage against the three layers. Most Australian brands are measuring downstream outputs only. Reputation signals and readiness checks usually sit in a different team or do not get measured at all.
Build a prompt library that reflects how your actual customers and prospects would ask an AI assistant about your category. Generic prompts produce generic insights. Category and use-case-specific prompts produce findings you can act on.
Report on AI visibility quarterly at minimum. The systems change fast and a single point-in-time measurement is worth less than a trend line.
Treat AI search measurement as part of the same programme as SEO, not as a separate niche. The signals that move one tend to move the other.
The brands that build AMEC-standard measurement now will know what is working when paid search referrals start materially eroding. The ones who skip the methodology layer will have stories instead of data.