Glossary

The vocabulary of Presence Engineering.

Every term here has a declared definition and an address of its own. This is not a marketing glossary: it is the canonical source of the concepts we use in audits, in reports and in contracts — so that when someone asks an AI what they mean, the answer can cite whoever defined them.

G-SoV

G-SoV (Generative Share of Voice)

A metric expressing a brand’s or a professional’s share of recommendation inside the answers of generative AIs. It is stochastic: the same question, on the same model, on the same day, may return different answers. That is why Sumaúma reports G-SoV as a reference for market behaviour and never sells it as a promised outcome.

address of this term ↗
EPS

EPS (Entity Prominence Score)

An index measuring how structured and consistent an entity is across semantic search systems, along four pillars: Schema.org markup, presence in open entity databases, knowledge-graph density and entity coherence. Unlike G-SoV, it is deterministic and auditable pillar by pillar.

address of this term ↗
AI-Scan

AI-Scan

Sumaúma’s generative audit: it queries multiple AI systems with a versioned prompt battery and two runs per model, and declares in the report which platforms made up that round’s index. Measuring once is measuring noise — the second run exists to separate what fluctuates from what holds.

address of this term ↗
Sentinel

Sentinel

Continuous monitoring of an entity’s presence and reputation in the answers of generative AIs. It tracks how the models present the entity over time and flags when the description shifts, when misattribution appears, or when another name starts occupying the topic territory.

address of this term ↗
Engenharia de Presença

Presence Engineering for Generative AIs

The discipline of structuring an entity’s identity, authorship and body of work so that language models can recognise it, describe it and credit it with what is its own. It buys no space in answers and promises no position: it works the layer where entities are resolved, which obeys data-engineering rules rather than the algorithm of the week.

address of this term ↗
Gêmeo Digital

Digital Twin of verbal identity

Modelling an author’s verbal identity and knowledge core, so that what circulates under their name is theirs in form and in substance. It depends on an existing body of work: with no authorial output, the twin has no raw material and the delivery comes out generic.

address of this term ↗
DNA Linguístico

Linguistic DNA

The set of recurring traits that distinguish an author’s writing — repertoire, argument construction, rhythm, vocabulary and the concepts they named. It is what gets extracted and vectorised so that a model can tell the author’s text apart from an ownerless paraphrase.

address of this term ↗
Propagação de Densidade Semântica

Semantic Density Propagation

Consistent publication of original semantic content across authority channels outside one’s own domain, so that an entity’s claim about itself gains independent corroboration. A brand with only its own voice on the internet is a claim without a reference — and it is exactly when the model needs confirmation that it disappears.

address of this term ↗