How we score AI visibility: the AIS Score, in plain English.

Your AIS Score is a single number out of 100. AIS stands for AI Search visibility — how present your brand is in the answers the engines give, and how favourably. One number is a strong claim, so this page explains exactly what sits behind it: the four signals, how they are weighted, and where we have deliberately chosen to underpromise.

We would rather hand you a number you can interrogate than a number you have to trust. So here is the whole of it.

The four signals

The AIS Score is a weighted composite of four measured signals. Each is computed from real engine responses — not estimates, not modelled traffic — gathered by polling the same prompts repeatedly across ChatGPT, Claude, Gemini, Perplexity and AI Overviews.

1. Visibility — 60%

The first question is the blunt one: when a buyer asks, are you named at all — and on how many of the engines? For every prompt in your set, across every engine and country, we record whether the answer mentions your brand, and how broadly that holds across ChatGPT, Claude, Gemini, Perplexity and AI Overviews. Visibility combines that mention rate with platform breadth.

It carries well over half the score because it is the most consequential signal by a distance. A brand the engines never name cannot benefit from a good position or favourable language — there is nothing for those signals to attach to. Visibility is the gate everything else passes through, and being named on four engines is a different reality from being named on one.

2. Position — 15%

Being named is not the same as being named first. When an engine lists three CRMs, the order is not cosmetic; buyers anchor on the first name and treat the rest as alternatives. Position measures where in the answer you tend to appear.

We score position on a decaying curve rather than discrete ranks. First mention scores near the top of the band; each subsequent position is worth meaningfully less, and we interpolate between positions rather than rounding to a rank — a brand that sits consistently between second and third should not be scored as if it were always third. The decay is steep on purpose: in an answer, second is much closer to invisible than it is to first.

3. Sentiment — 10%

Presence without favour can hurt you. An engine that names you as "the dated, expensive option" has named you — and lost you the deal. Sentiment reads how you are characterised in the answers that mention you, scored from clearly negative through neutral to clearly positive.

We are conservative here. Neutral mentions — "tools in this space include X, Y and Z" — score as neutral, not positive. We only credit favour when the language is genuinely favourable. Inflating sentiment would inflate the headline score, and the headline score is the one thing we refuse to flatter.

4. Share of voice — 15%

This signal is competitive. Share of voice asks: of all the brand mentions in answers across your category, what proportion are yours? It is the only signal that moves when your competitors move, which is precisely why it earns a place — your visibility is always relative to the field you are measured against.

Visibility tells you if you are in the room, and on how many of the engines. Share of voice tells you how much of the room is talking about you rather than someone else.

How the weights combine

The four signals are normalised to a common 0–100 scale, multiplied by their weights, and summed:

AIS = (Visibility × 0.60) + (Position × 0.15) + (Sentiment × 0.10) + (Share of voice × 0.15)

The result is your AIS Score. The weighting is fixed and public, and it reflects a deliberate order of priority: being named — and named widely — matters most, by a clear margin; where you are named comes next; how you are spoken of and how you compare share the remainder. Visibility dominates because the rest is meaningless until you are in the answer at all.

The decisions we made on purpose

A score is only as honest as its edge cases. Three choices are worth stating plainly.

• No row is not a zero. If we have not yet gathered enough responses for a prompt, that prompt reads as "awaiting data", not as a zero. A blank is not bad news; it is the absence of news, and we render it that way rather than dragging your score down with phantom failures.
• A real zero is a real signal. If we *have* polled a prompt and you were named in none of the answers, that is a genuine, scored zero — shown as "no visibility yet", neutrally, never as a red alarm. Scanned-and-invisible is information you need; we just decline to dramatise it.
• Confidence travels with the number. Every score carries the volume of evidence behind it — how many engines, how many responses, over what window. A 61 from 40 responses and a 61 from 600 are not the same claim, and we never pretend they are.

Why one number, then?

A composite invites the obvious objection: why collapse four good signals into one lossy figure? Because teams need a single thing to watch move. The four signals are always one click away — the breakdown shows each, with its weight and its evidence — but the headline exists so you can answer "are we getting more visible, or less" without a meeting.

Underpromise, then show the working. The AIS Score is the promise. This page, and the breakdown behind every score in the product, is the working. If a number ever looks too kind, open it up — we built it to be opened.