G-SoV: the metric that decides whose name AI recommends (and whose it forgets)
Generative Share of Voice is the new share of voice. What it is, how it is measured, and why it is already redistributing authority between brands — quietly.
For twenty years, the marketing question was "how do I show up on Google?". The question has changed. Today, millions of buying decisions begin with a prompt — "what is the best tool for…", "who should I hire to…" — and the answer arrives finished, in prose, from a generative AI. In that answer, your name is either cited or it simply does not exist.
This is where a new metric comes in, one we at Sumaúma call G-SoV — Generative Share of Voice. It measures the share of recommendation and authority your name holds inside language models. It is not about traffic. It is about presence in the answer.
What Generative Share of Voice is
Traditional share of voice measured how much of a medium a brand occupied — how many ads, how many mentions, how much space. G-SoV translates that idea into the age of AI: out of every 100 times someone asks something relevant to your field, in how many is your name cited, recommended or used as a reference by the AI?
While one expert is cited in 86 out of 100 prompts, another is cited in one or two. The first has become the market's default answer. The second has entered what we call semantic extinction: it still exists in the real world, but it has disappeared from the place where decisions are now being made.
Why the click stopped being the metric
The old model was simple: you appeared in a list of links, the person clicked, you measured the visit. Generative AI breaks that chain. It does not return ten blue links — it returns one synthesised answer, and that is where your name needs to be.
This means a large part of the discovery journey now happens without a single click on your site. You can be losing share every day without any obvious drop in analytics — because the contest has moved somewhere your traffic dashboard cannot see.
We do not measure clicks. We measure algorithmic sovereignty.
How G-SoV is measured
Measuring presence in AI requires leaving guesswork behind. The diagnosis combines three readings:
- Prompt coverage — we map the real questions of your field and check, across the main models, in how many your name appears.
- Citation quality — being mentioned is not enough. Does the AI recommend you, mention you in passing, or associate your name with something negative? Position and context matter.
- Consistency across models — being strong in one model and absent in another is a weakness. Real sovereignty means appearing reliably across all of them.
The result is a comparable number — your G-SoV — and, beside it, your competitors'. It is a photograph of who dominates the AI conversation in your sector.
Dominance on one side, extinction on the other
What makes G-SoV urgent is that it concentrates. Language models tend to reinforce whoever is already cited as a reference: the more a name is recommended, the more material exists about it for the model to learn from, and the more it gets recommended again. It is a compounding effect — much like interest — and it separates quickly those who built presence from those who stood still.
In plain terms: every quarter without structuring your presence is not a draw. It is ground conceded to whoever is structuring theirs right now.
What to do about it
Improving G-SoV is not "posting more". It is presence engineering, across four fronts that feed one another:
- Graph structuring — making the AI understand that you exist, who you are and where your authority comes from, as a recognised entity.
- Digital twin — cloning your voice and your authorial knowledge, so the AI produces authentic and consistent content in its name.
- Semantic propagation — distributing content density, anchored in current conversations, across the authority axes the models consult.
- Monitoring — tracking G-SoV, sentiment and reputation over time, adjusting the strategy.
You do not need to start with everything. You need to start by knowing where you stand — and that is exactly what a diagnosis reveals.
The starting point
The question that opens this text — "what is my presence in AI?" — now has an objective answer. G-SoV exists, it is measurable, and it is already redistributing authority as you read. The only real choice is whether you measure yours before or after your competitor measures theirs.
Discover your name's real G-SoV.
Request the AI-Scan Audit and see whether your name shows up — or disappears — in AI answers.
Request AI-Scan