AI visibility score

An AI visibility score blends five weighted sub-metrics into a single 0–100 number:

• Mention frequency (typically 25% weight): how often your brand appears in AI-generated responses across a tracked prompt set.

• Citation rate (25%): how often your content is cited as a source in those responses.

• Ranking position (20%): where your brand or content sits in source lists or inline mentions.

• Sentiment (15%): polarity and intensity of how the model describes your brand.

• AI-referred traffic (15%): real visits clicked through from AI assistant citations.

Each sub-metric is normalized to 0–100 against your tracked prompt set and tracked competitors, then weighted and summed. Weights are configurable — agencies often emphasize mention and citation more heavily; in-house brand teams often weight sentiment higher.

The score is computed per platform and rolled up into a composite. Per-platform scores reveal which AI assistant to prioritize next; the composite is the executive headline.

The composite is for executives. The sub-metrics are for operators.

An AI visibility score of 62 tells the CMO whether the program is on track. It does not tell the SEO team what to fix. For that, the breakout matters: 62 with high mentions and low citations means an on-page extraction gap; 62 with high citations and low ranking means a first-mention prominence gap; 62 with high everything and low traffic means a click-through gap (poor titles, weak intent match).

Reporting only the composite hides the levers. Reporting only the sub-metrics overwhelms executives. A mature program publishes both, with the composite as the headline and a one-line interpretation of which sub-metric drove the change.

AI visibility score gives boards, executives, and cross-functional partners a single number to track over time without losing the operational nuance underneath. It makes the results of a GEO program legible to people who don't read citation logs.

It also forces consistency. A composite score requires a defined prompt set, a defined platform list, defined sub-metric weights, and a defined refresh cadence. The discipline of computing the score reliably is half the operational value — programs that compute it weekly tend to ship the underlying improvements weekly too.

Five steps to compute a credible AI visibility score:

1. Lock the prompt set. 100–500 buyer-language prompts that don't change between measurement periods. Adding prompts mid-quarter invalidates trend comparisons.

2. Run prompts on every tracked AI platform on a recurring schedule. Weekly is typical; daily for high-volume brands.

3. Compute each sub-metric per platform. Mentions (count), citations (rate), ranking (mean position), sentiment (−100 to +100 normalized to 0–100), AI-referred traffic (sessions over period).

4. Apply weights and aggregate. Default weights: mentions 25%, citations 25%, ranking 20%, sentiment 15%, traffic 15%. Adjust deliberately, document the change, restate prior periods.

5. Report on rolling 30-day windows. Single-day snapshots are too noisy. 30-day rolling makes the trend legible without overreacting to short-term variance.