Methodology

How we measure AI visibility.

Everything behind the score: which signals we measure, how we weight them, which engines we query, and how the pipeline runs.

What the score measures

The AI Visibility Score (0–100) answers one question: when a potential customer asks an AI engine about businesses like yours, how well does your business perform across every relevant query? A score of 100 means your business is cited first, positively, in every response. A score of 0 means you don't appear at all.

The score is not a ranking within Clartiv. It is a measurement of your real-world AI recommendation rate, aggregated across four engines and dozens of semantically distinct queries.

The three signals

Visibility Rate

45%

The percentage of queries across all engines where your business appears in the response at all, whether cited directly, mentioned, or included in a recommended list.

Position Score

35%

Where you appear when you do show up. First position counts fully; each subsequent position is discounted. This reflects how prominently AI engines recommend you versus competitors.

Sentiment Score

20%

The tone of the language used to describe your business. We classify each mention as positive, neutral, or negative using a second-pass AI analysis step.

Score = (0.45 × VisibilityRate) + (0.35 × PositionScore) + (0.20 × SentimentScore)

Which engines we query

We query four AI platforms via their official APIs. We never scrape.

ChatGPT

OpenAI Chat Completions API

Gemini

Google Generative Language API

Perplexity

Perplexity API

Google AI Overviews

Licensed SERP data provider

Query fan-out

For each business, we generate 12–24 semantically varied queries representing how real customers search for businesses in that category and location. Examples include direct queries ("best accountant in Manchester"), comparison queries ("accountant vs bookkeeper in Manchester"), and recommendation-style queries ("which accountant should I use in Manchester").

Each query set is run twice on a 15-minute interval. We take the mean of both runs to smooth out per-request variation inherent in probabilistic language models.

Computation pipeline

  1. 1

    Generate query fan-out (12–24 queries) for category + location.

  2. 2

    Run queries in parallel against all four engine APIs with temperature set to 0.

  3. 3

    Parse each response for mentions of the target business name (exact + fuzzy match).

  4. 4

    Extract position: first mention = position 1, each subsequent = position N.

  5. 5

    Run second-pass sentiment classification on each mention.

  6. 6

    Compute VisibilityRate, PositionScore, and SentimentScore per engine.

  7. 7

    Average scores across all four engines, weighted by engine market share.

  8. 8

    Apply the composite formula to produce the final 0–100 score.

What the free check does

The free check at clartiv.com/check runs the full computation pipeline on a single snapshot: your business name, category, and city. It queries all four engines, computes all three signals, and returns your composite score within about 30 seconds. No account required.

Paid plans run the same pipeline on a weekly cadence, track score changes over time, add competitor benchmarking, and give you access to the full query log.

Frequently asked

See your score.

Free check, 60 seconds, no account required.

Run a free check