At AI Tech Expo Global in London, a pattern became impossible to ignore. Across demos, pitches, and panel discussions, vendor after vendor was showing the same thing: a dashboard with a binary answer. Your brand is cited. Or it isn't. Green dot. Red dot. That's the model. And that model is fundamentally broken.
The false comfort of binary measurement
The appeal of binary citability measurement is obvious. It is simple to understand, simple to sell, and produces a clear deliverable. "You are cited by ChatGPT on this query." It feels like data. It feels like insight. It is neither.
Here is the structural problem: AI models do not produce consistent, deterministic outputs. Every response is the product of a probabilistic process influenced by user context, conversation history, memory settings, browsing state, system prompt, model version, and the specific phrasing of the query. Ask the same question twice and you may get two different answers — with different citations.
Measuring citation as an absolute fact ignores all of this. It produces a snapshot that is technically accurate at the moment of capture and meaningless as a strategic signal. Worse, it produces false positives at scale — and false positives are the most dangerous kind of measurement error, because they give you confidence you haven't earned.
Standardising citation as a binary outcome means systematically generating false positives. You're not measuring reality — you're measuring one sample of a probabilistic process.
Why you can't simply ask the AI "am I cited?"
A common approach — and a seductive one — is to query an AI model directly: "Which brands do you recommend in [category]?" and then check whether your brand appears. This is not citation measurement. It is citation sampling.
The problem is multidimensional. The AI's response depends on the exact phrasing, the platform, the user profile, the date, and dozens of other variables. Sampling once, or even ten times, tells you almost nothing about your actual citation probability across the full distribution of real queries users are asking.
To measure citation meaningfully, you need to map the semantic field — the full range of queries, contexts, and framings in which your brand could plausibly appear — and then measure probability of citation within that field. That is a statistical exercise, not a lookup.
The competitor problem: who defines the competitive set?
Most citability tools ask you to define your competitors upfront. You enter a list of five brands, and the tool measures how often each one is cited. This seems reasonable. It is not.
The critical insight is this: AI models define their own competitive sets. When an AI answers a query about, say, luxury villa rentals in Tuscany, it doesn't consult your pre-defined competitor list. It draws from its training data and real-time context to identify the brands it considers relevant and comparable. That set may include players you've never considered, and it may exclude direct competitors you've been benchmarking for years.
By defining the competitive set yourself, you are measuring a game that isn't being played. You are optimising for a comparison the AI isn't making. The correct approach is to let the AI reveal its own competitive framing — and then measure your citation probability within that AI-defined set.
What Citation Rate actually measures
The Citation Rate model was built to address exactly these limitations. It does not ask whether a brand is cited in a single query. It measures the probability of citation within a specific semantic field — a defined cluster of related queries, contexts, and user intents that map to a real market space.
The output is not a green dot or a red dot. It is a probabilistic score that reflects the structural likelihood of citation across the full range of relevant AI interactions. It is a signal you can track over time, benchmark against AI-defined competitors, and use to inform content and positioning decisions.
This distinction — probability versus occurrence, field versus query, AI-defined versus self-defined — is what separates a measurement model from a measurement illusion. The brands investing in the right model now will have a structural advantage that compounds. Those relying on binary dashboards will keep chasing a signal that doesn't mean what they think it means.