GEO measurement is not enough

Marketing and communications leaders are used to managing a complex ecosystem of media, platforms, creators and audiences. They now need to manage another increasingly influential layer of discovery: large language models (LLMs).

AI-generated answers from platforms like ChatGPT, Claude, and Gemini are rapidly becoming a first point of contact between organisations and their audiences. These systems do not simply retrieve and surface information. They synthesise it, interpret it and shape how brands, issues and institutions are understood. How an organisation appears in those answers is quickly becoming a discoverability, reputation, and growth issue, affecting everything from awareness and consideration to stakeholder trust.

In AI discovery environments, visibility is increasingly shaped by authority signals. Those signals come from what is increasingly being defined as earned: not simply media coverage, but the broader ecosystem of independent voices, expert communities, institutional sources, industry publications, and peers that establish credibility before brands attempt to scale reach through paid channels.

AI systems are fundamentally earned environments. They rely on credible third-party information, editorial validation, and consistent digital signals to determine what information and brands are trustworthy enough to surface. As a result, earned is no longer just a communications and marketing discipline, but a core driver of discoverability, awareness and growth. That changes the role earned media, creator influence, communities, expert advocacy and other forms of earned authority play in discoverability itself.

AMEC's recently released GEO Principles represent a meaningful step forward for the industry, bringing needed rigor to a space that has, until now, operated largely without shared measurement standards or consistent evaluation approaches. GEO is increasingly being discussed as a channel strategy. But measurement, even good measurement, is not channel management. As GEO moves into the boardroom, that distinction matters.

What organisations can realistically measure today

Today, GEO measurement is primarily a framework for observing visibility, sentiment, competitive share of voice, citation share, referral traffic from AI platforms and accuracy across prompts, models and markets.

These are meaningful baselines. They tell you where you stand in the information environment AI systems are drawing from. They help you understand whether your brand is present, whether it is being characterised accurately and whether that picture is positive. They are the GEO equivalent of media monitoring, social listening or share-of-voice analysis.

What they do not tell you is how to change the outcome

That gap is where many organisations are currently overestimating the maturity of GEO.

In one recent audit for a skincare brand, the most useful insight was not simply whether the brand appeared in AI-generated answers. It was where AI systems appeared to be looking for confidence. Reddit significantly over-indexed as a cited source in hand cream recommendations, while dermatologist-led YouTube reviews emerged as an important secondary signal. That changed the strategic question from, “How visible are we?” to, “Which credible communities, experts and content formats do we need to influence to shape future recommendations?”

The attribution gap

The bigger challenge in GEO right now is not measurement. It is causality. While there is growing consensus around certain content and technical practices, including structured content, authoritative third-party validation and discoverable information architecture, the market still lacks reliable approaches for understanding which specific interventions consistently move outcomes in AI discovery environments over time.

Which content changes led to a shift in how your brand is represented? Which earned placements, expert endorsements, institutional references or content investments are influencing citation share and visibility? Which information architecture decisions are improving or degrading your discoverability? These are the questions that matter for building a durable GEO strategy, and they are the questions the industry is least equipped to answer right now.

This is not a critique of the frameworks being developed. It is a reflection of where the technology is. AI discovery environments are probabilistic, not deterministic. Outputs vary across models, prompts, geographies and moments in time. Attributing a shift in visibility to a specific action is genuinely hard, and any vendor or framework claiming otherwise deserves scrutiny.

In one alcohol-free beer engagement, the starting point was a practical consumer question: when people asked AI assistants for the best nonalcoholic beers for social occasions, did the brand show up as a leading recommendation? The work identified which sources and category narratives were shaping those answers, then translated that into an earned strategy designed to win the sources AI systems were already referencing.

Measurement and deployment are not the same thing

Here is the distinction organisations need to internalise: measurement tells you where you are. Deployment is how you change it.

GEO deployment is the operational work of shaping the information environment AI systems learn from, retrieve from, and reference: content built for AI-led discovery, earned media that builds source authority, technical readiness that makes information accessible and governance that keeps it accurate over time. In practice, communications, search, content, analytics and reputation teams can no longer operate as separate functions if organisations want to influence AI visibility at scale.

Marcomms teams would never treat media monitoring as a media strategy or social listening as an influencer program. The same logic applies to LLMs. A brand that treats GEO as a deployment and authority-building challenge, rather than a monitoring exercise, is building something that compounds.

The organisations doing this well are not the ones running the most sophisticated dashboards. They are the ones thinking seriously about what AI systems need to accurately and confidently represent them: high-quality, consistent, credible information published across the right sources, structured for interpretation, and supported by the kind of third-party authority signals that AI systems treat as trustworthy.

What organisations should operationalise now

Given the current state of the market, organisations should focus on three priorities:

• Establish your baseline. Use emerging GEO frameworks, including AMEC's principles, to understand visibility, citation share, sentiment and answer accuracy across relevant prompts and models.

• Understand the information environment. Asses the sources, content and authority signals AI systems appear to be relying on when representing your organisation.

• Treat measurement as intelligence, not proof. Use GEO metrics to identify opportunities and risks, while recognising that attribution remains immature and causal relationships remain difficult to prove. Organisations that delay this work until attribution frameworks become fully mature likely fall behind competitors already shaping the information environment AI systems rely on today.

What this looks like in practice

Over the past year, working with organisations across sectors through Edelman’s GEOsight capability has reinforced a consistent lesson: visibility tracking is only one part of the challenge.

The organisations making the most progress are not treating GEO as a dashboarding exercise. They are using measurement to understand which authority signals, sources and narratives AI systems appear to trust, the investing in the earned ecosystems that shape those outcomes over time.

What has become increasingly clear is that discoverability in AI environments is less about technical optimisation alone and more about building the kind of credible information presence that earns trust across the broader digital ecosystem.

The early results we are seeing across industries confirm what the theory suggests: visibility in AI-generated answers is not random. It is a reflection of the quality, authority, and reach of the information environment a brand has built over time. You cannot buy your way to the top of an AI-generated answer. But you can earn it.

The broader shift

GEO is not fundamentally a measurement problem. It’s an authority, discoverability and information readiness problem. The brands that win in AI discovery environments will not be the ones that tracked their visibility most diligently. They will be the ones that build the kind of credible, consistent, well-structured information presence that LLMs are designed to find, trust and recommend.

Measurement, done responsibly and with appropriate humility about current limitations, is an important part of navigating this shift. But measurement alone will not determine who wins visibility, trust and influence in AI discovery environments.

Mirza Germovic is head of strategic AI Advisory at Edelman Intelligence