AI Visibility
Track how your brand appears in AI answers.
The AI Visibility page gives a full breakdown of your brand's presence across all tracked AI engines over time.
Visibility Score
The Visibility Score is a 0–100 index calculated per scan batch:
score = round((sum of tier_weights[best_tier] for each prompt result) / total_results × 100)
Tier weights: Tier 1 = 1.0 | Tier 2 = 0.6 | Tier 3 = 0.3
If your brand is not mentioned in a result, that result contributes 0 to the weight sum. A score of 100 means every AI response cited your brand at Tier 1.
Timeline chart
The timeline chart shows your Visibility Score over time (7d / 30d / 90d). Each data point represents one scan batch. When the chart has fewer than 7 data points, anomaly detection is disabled. With 7+ days of data, spikes and drops are flagged when the z-score exceeds 2.
Mention Quality tiers
Every brand mention in a scan result is assigned a tier:
| Tier | Meaning | Weight |
|---|---|---|
| Tier 1 | Primary recommendation — brand is the main answer | 1.0 |
| Tier 2 | Secondary mention — brand appears but is not the top recommendation | 0.6 |
| Tier 3 | Footnote — brand appears only peripherally | 0.3 |
The best tier for each prompt result is used in the score calculation.
Share of Voice by model
The Competitors tab shows Share of Voice broken down by AI model (ChatGPT, Perplexity, Claude, Gemini, etc.). Each model shows your brand's average visibility score, mention count, and average sentiment on that model.
Sentiment
Sentiment is scored per AI response as positive, neutral, or negative. The aggregate Sentiment Score is the average of all response scores:
sentiment_score = round((positive×100 + neutral×50 + negative×0) / total_results)
Sentiment is averaged across results — not summed. A score of 50 means the average response is neutral.
Hallucination Monitor
The Hallucination Monitor checks AI responses against your Brand Profile ground truth (pricing, founders, headquarters, features). When an AI response contradicts a known fact, the system logs an inaccurate check. You can review inaccurate claims and see which model produced them and in which prompt context.
Signals tab
The Signals tab lists all detected anomaly events with their type, severity, title, description, and current status. Each signal links to the relevant action URL for follow-up.
Anomaly detection requires at least 7 days of scan history. Signals will not appear before that threshold.
Next steps
- Prompts — manage which queries are tracked
- Sources — see what domains AI models cite
- Optimize Actions — act on visibility gaps