METHODOLOGY

A repeatable measure of AI visibility

Fameproof measures whether AI systems mention an entity, where it appears and whether the answer describes it accurately. The same framework is applied to every company, public figure and club.

24prompts per model
3AI models
293tracked entities
Weeklymeasurement cycle
01

What we measure

The index captures unprompted visibility: how likely an entity is to appear when someone asks a broad, relevant question without naming that entity in advance. It is designed to compare entities within the same kind of market or cultural field.

The public boards currently cover 293 companies, public figures and football clubs. Each entity is assigned to one board and one peer group before scoring.

02

How the prompt panel works

Each category uses a panel of 24 natural-language questions, each asked 4 times per model, modelled on what a buyer, fan, journalist or curious user might ask. Prompts cover discovery, comparisons, recommendations, reputation and factual knowledge.

No target names in discovery prompts

A prompt such as “Which enterprise AI companies should I know?” can test discovery. “Tell me about OpenAI” cannot, because it guarantees a mention.

The same prompt wording is used for every model in a measurement cycle. This keeps the comparison consistent even though model behaviour can vary between runs.

03

Models and timing

The current panel covers Claude, ChatGPT, Gemini. Results are grouped into weekly measurement cycles; the current cycle begins 24 August 2026.

Model versions can change over time. Fameproof keeps the public labels at product level and records the model configuration with each internal run so historical comparisons can be interpreted in context.

04

How the 0–100 score is built

Every successful response is evaluated across four signals. Together they describe not only whether an entity appears, but the quality of that appearance.

Mention rate

How often the entity appears in relevant answers.

Prominence

How early and how clearly it appears when mentioned.

Accuracy

Whether key factual statements are supported and correct.

Sentiment

The overall framing of the entity in the response.

A separate score is calculated for each model. The Fameproof index is the equally weighted mean of those model scores, rounded to one decimal place. No one provider receives extra weight.

05

Rankings, movement and freshness

Entities are ranked only against others on the same board. A company’s rank is therefore not directly comparable with a musician’s or a football club’s rank.

The trend shows movement over recent measurement cycles, while the headline score shows the latest completed cycle. Fresh results can be reused within the same cycle to avoid paying for duplicate model queries without changing the comparison.

06

What the index does not claim

Fameproof is a structured sample of public AI output, not a census of every possible answer. Scores can be affected by model updates, geography, language, temporary service behaviour and the wording of a prompt.

The index does not measure universal fame, public approval, search demand or commercial value. It should be read as a comparative AI-visibility benchmark, supported by the underlying prompt evidence.

Questions about the methodology? Request an audit to see the prompt-level evidence and the specific gaps behind a score. Request an audit