What does the AI Visibility Score measure?

How often and how prominently a company brand is cited by ChatGPT, Perplexity, Gemini and Claude when buyers ask category questions, on a 0-100 scale across five weighted dimensions: presence, rank, engine coverage, consistency and source authority.

Can a company pay to rank higher?

No. Rankings derive solely from observed engine outputs. No company can pay to appear or to improve its position.

How do you handle the randomness of LLM answers?

Each prompt is sampled multiple times across paraphrased variants. Stability across those runs is itself a scored dimension (consistency), and scores are aggregated to reduce single-run noise.

How often is it updated?

Monthly, with the methodology versioned in public.

Methodology · v2.1

How the score is built.

One number, 0–100, summarising whether an AI assistant puts a brand in front of a buyer. It combines five dimensions, each capturing a distinct way a brand can fail — or succeed — at being seen. No brand can pay to rank. Scores reflect our model and are revised monthly.

The five dimensions

Presence

Is the brand cited at all?

Across a fixed corpus of category prompts, how reliably the brand appears in the answer at all. A brand absent from the answer scores zero here — product quality is irrelevant if the model never names it.

Rank

How prominently?

Where the brand lands within the answer when it does appear. Named first is not the same as buried eighth; position is weighted accordingly.

Engine coverage

On how many engines?

How many of the four engines cite the brand. Broad coverage signals durable, structural visibility rather than a single-model artifact that disappears with the next update.

Consistency

Is it stable?

How stable citation is across paraphrased prompts and repeated runs. Because LLM outputs are non-deterministic, a brand that appears only intermittently has fragile visibility — and we score that fragility explicitly.

Source authority

Why is it cited?

The independence and quality of the sources the engines lean on to surface the brand. Citations grounded in third-party, editorially independent sources are weighted above self-referential ones.

What we publish, what we protect

Transparent on the what and why; the exact mechanics stay proprietary

Published

The five dimensions and their weights
The engines measured and the update cadence
That measurement uses multi-sample prompting
The version history of the methodology
The cited sources behind each brand's score

Proprietary

The exact prompt corpus per category
The sampling cadence and aggregation math
The source-authority scoring model
The normalisation and anti-noise pipeline
The internal weighting calibration

Version history

v2.1 · Jun 2026Engine coverage weighting refined; Vertical AI added as a tracked category.

v2.0 · Mar 2026Source authority introduced as a fifth dimension; prompt-variant sampling expanded.

v1.0 · Jan 2026First public edition: presence, rank, engine coverage and consistency.

Trust rules

No pay-to-rank: rankings derive only from observed engine outputs.
Every brand's score links to the sources behind it.
Scores are estimates from our model, revised monthly.
The methodology is versioned in public; changes are logged.
ZivRank is measured on the same basis as every other company.