Direct answer
The signals are the four things an answer engine checks before it repeats you: an identity it can pin down, corroboration from sources it already trusts, claims specific enough to verify, and content structured so it can lift a clean line. Call it the Citable Entity Model — identity, consistency, corroboration, extractability. Build them in that order. A gap in any one is usually why an engine describes a competitor instead of you.
The instinct is to think citability is earned by being the best supplier in the category. It isn’t. An answer engine can’t see who is best; it can only see who it can describe without guessing. Being excellent and being citable are different achievements, and the second is the one that gets you named in the answer.
Identity: the engine has to know which company you are
Before anything else, the machine has to resolve you to a single, stable entity — one company, one name, one set of facts — not three half-matching versions scattered across the web. If your name, category and core details read differently on your site, your directory listings and your profiles, the engine hedges. It can’t confidently say the company being asked about is the company it found. Identity is the floor: without it, nothing built on top counts.
Consistency: the facts have to agree with themselves
Identity holds only if it’s repeated the same way everywhere a machine can read it. One description of what you do, one spelling of the name, one address, one category — stated identically across every surface. Contradictory versions don’t average out to a confident answer; they cancel. This is the cheapest signal to fix and the one most often left broken, which is part of why AI tools don’t recommend your company even when the work is strong.
An engine can’t name the best supplier. It can only name the one it can describe without guessing.
Corroboration: someone other than you has to say it
Your own site claiming you’re a leader is data the engine discounts by default — you would say that. Corroboration is the same claim confirmed by a source the engine already trusts: an independent mention, a listing, a third party describing you the same way. That outside confirmation is what turns a claim into something a machine will repeat. It’s the signal that separates real authority from volume, the subject of whether authority building is an investment or a vanity cost.
| Signal | What it proves to a machine | How you build it |
|---|---|---|
| Identity | You are one resolvable company | One name, category and set of facts, everywhere |
| Consistency | The facts don’t contradict | State them identically across every surface |
| Corroboration | Someone trusted confirms you | Independent mentions, listings, third-party references |
| Extractability | Your claim can be lifted cleanly | Direct answers, definitions, self-contained lines |
Extractability: the claim has to lift out clean
The last signal is mechanical. Even a trusted, well-identified company goes unquoted if its best claim is buried in a paragraph that only makes sense in context. Extractable content states the answer plainly, defines its own terms, and puts the quotable line where a machine can take it without dragging in three sentences around it. The company that writes to be lifted gets lifted.
Working benchmark: across the diagnostics I’ve run, the line where engines start quoting a company tends to sit near 60 on a 0–100 citability score, while most firms first measure in the high 30s. The gap is almost never all four signals at once — it’s usually one weak pillar, most often corroboration or extractability. An internal benchmark from client work, not a published statistic.
Build them in order
The order matters because each signal rests on the one before it. Corroboration of an identity the engine can’t resolve is wasted; extractable content about a company nothing confirms is a clean quote the machine still won’t trust. Fix identity and consistency first, earn corroboration second, make the content extractable third. If you want the range that groundwork costs, it’s set out in what answer-engine optimisation actually costs.
- Search your own name and category — does the engine describe one company, or several half-versions?
- State your name, category and core facts identically on every surface a machine reads.
- Get at least one trusted third party to describe you the same way, independently.
- Put your strongest claim where it lifts out clean — a direct answer, not a buried sentence.
If you can’t tell which of the four signals is the one keeping you out of the answer, and you don’t want to guess:
Keep polishing the product while the engine can’t resolve who you are, and you’ll stay the best-kept secret in a category that names someone easier to describe.
Book a free 15-minute session See the proof