Evidence, not promises: how we measure AI visibility

Anyone can type a question into an AI assistant, screenshot the answer, and call it analysis. Measurement is something stricter: a defined method, applied the same way every time, with the evidence kept.

Here is how Logres measurement works, and why we hold it to rules that occasionally make our own reports less flashy.

Pinned questions, stored answers

Every measurement starts with a pinned prompt suite: a fixed, versioned set of questions that reflect what buyers in a market actually ask. Those questions are run against the AI assistants buyers use, and every response is stored with its date and sources.

Pinning matters because comparison is only meaningful under a stable method. When we compare two measurements, it is because the questions, scoring rules, and methodology version allow it — and when they do not, we say so instead of presenting a trend that is not there.

Four dimensions, never one number

A Logres baseline reports four dimensions, each on a 0–100 scale:

• Visibility — how often and how prominently the business was named in recorded recommendation-style answers.
• Authority — how often recorded decision-support answers cited the business's own public materials.
• Accuracy — how well recorded descriptions matched approved business facts, when the evidence supports scoring it.
• Influence — whether recorded answers helped the business enter a buyer's set of options and provided supportive context.

We deliberately never average these into a single overall score. A blended number would hide exactly what a diagnosis needs to expose: a business can be named often (Visibility) while being described inaccurately (Accuracy), or be well described but rarely cited as a source (Authority). Collapsing that into one figure makes the report cleaner and the decision worse.

Why "unavailable" is an honest answer

Some dimensions require specific evidence. Accuracy, for example, can only be scored against an approved fact sheet. When the required evidence does not exist, we mark the dimension unavailable with a reason — we do not estimate a number to fill the gap.

An unavailable result is more honest than a number that looks precise but is not supported. A score that exists only because a dashboard dislikes empty cells is not measurement; it is decoration.

Movement is an observation, not a verdict

When a score changes between pinned measurements, that is recorded fact. What caused it is a separate question. AI answers shift as models, sources, and public information change, so a changed score is an observation to review, not proof of cause.

The same discipline applies in the other direction: a before/after sequence around remediation work can document that changes were delivered, verified on the site, and later re-measured. It cannot, by itself, establish that the work caused the movement. We document the sequence and say plainly what it does and does not show.

What this buys a client

Evidence-led measurement is less exciting than a guarantee. What it provides instead is durable: a dated baseline that can be inspected, a diagnosis grounded in stored answers, and later measurements that mean something because the method stayed pinned. When someone promises you a specific AI ranking, ask to see the recorded evidence behind the claim. If there is none, you already know what the promise is worth.

Frequently asked questions

Why are the four dimensions never combined into one score?

Because a blended number hides the diagnosis. A business can be named often while being described inaccurately, or described well but rarely cited as a source. One average would make the report tidier and the decision worse.

What does it mean when a dimension is "unavailable"?

It means the evidence required to score that dimension responsibly was not there — for example, Accuracy requires an approved fact sheet. We report the reason instead of estimating a number. An unavailable result is more honest than a precise-looking figure that is not supported.

If my score moves between measurements, did the work cause it?

Not necessarily. AI answers shift as models, sources, and public information change. A changed score is an observation to review, not proof of cause. We document the sequence — what was delivered, what was verified live, what was later measured — and say plainly what it does and does not show.

This note is for information only. Logres measurements describe recorded AI responses for a defined set of questions at a point in time; they are not a guarantee that any AI system will name, cite, or recommend a business in the future, and they do not predict rankings, traffic, leads, or revenue.