Co-Mentions
- Co-mentions are the brands named with yours — the competitive set the model assigns you.
- Identical mention rates can hide opposite positioning: leaders beside you, or discounters.
- Co-mentions are the denominator that makes Share of Voice and BRS meaningful.
- A shifting co-mention set signals category re-framing before volume metrics move.
Consensus definition
When an AI answer recommends or compares options, it names a set. Co-mention analysis records which brands appear together and how often, building — across a prompt set — an implied competitive graph: who the model treats as substitutes, who anchors a category, who is peripheral. The founding GEO work established that brand visibility in generative answers is a measurable, comparative surface;1 co-mentions are the relational layer of that surface. A brand is not positioned in isolation in an AI answer — it is positioned by the company it keeps.
rhinegold operator refinement
Rhinegold uses co-mentions to read positioning, not just presence. Two brands can have identical Mention Rate yet sit in entirely different competitive frames: one consistently named beside category leaders, the other beside discount or niche players. The co-mention set is also the denominator that makes Share of Voice and Brand Recommendation Share meaningful — both are relative to who else is in the answer. Watching how the co-mention set shifts over time is an early signal of category re-framing: when the names beside you change, the model's mental map of the market is changing, often before any volume metric moves.
Operational use
Build a co-mention matrix across the prompt set: for each answer, record the brand set, then aggregate to see your most frequent companions and the brands you are rarely or never named with. Use it to define the real competitor set for Share of Voice, to spot aspirational gaps (leaders you are absent from), and to detect mis-framing (being bucketed with the wrong tier).
Measurement boundary
Co-mention frequency is associative, not evaluative: appearing beside a leader is not the same as being ranked with it (Brand Rank carries order) and says nothing about direction (Negated Mention — you can be co-mentioned as the cautionary contrast). Co-mention graphs are also prompt-set-dependent and vary by provider, so the competitive map is relative to what was asked and where.
Distinct from
Against Share of Voice, which collapses the competitive set into one share number — co-mentions keep the relational structure. Against Brand Recommendation Share, which weights a brand's presence by how crowded the answer is, but does not record which competitors. Against Mention Rate, which counts the brand alone. Against the Competitive Mention Map, which is the structured visualisation built on co-mention data — segmented by query cluster and platform — that turns the raw pairs into a navigable view of the AI-projected peer set.
Operational note
Common mistakes
- Defining the competitor set from internal assumptions rather than from who the model actually names alongside the brand.
- Reading any co-mention as good — being named as the contrast case is co-mention with a negative sign.
- Ignoring absence — the leaders a brand is never co-mentioned with are often the sharpest positioning gap.
Where consensus is missing
There is no standard for co-mention measurement: window (same sentence, same list, same answer), weighting, and how to treat parent/subsidiary names all vary. No published benchmark exists for what co-mention density or composition implies for consideration, so the graph is read qualitatively.
Sources & deeper reading
- rhinegold co-mention matrix methodology — competitive-set construction and category-frame tracking across providers, shared with engaged clients and partners.