AI Visibility Score Explained
Understand how AI recommends your brand when customers research your market
What is AI Response Stability?
When you ask an AI "What are the best CRM tools?", the answer varies each time due to probabilistic sampling. This variability matters for brand visibility:
Stable Mentions
Your brand is consistently recommended across multiple runs. This indicates strong AI perception.
Variable Mentions
Your brand appears sometimes but not reliably. There's room for improvement in content strategy.
Missing Mentions
Your brand doesn't appear at all. This represents an opportunity gap to address.
Consistency Score
A percentage (0-100) that tells you how reliable AI's answers are about a topic.
AI gives almost the same answer every time. The story is settled.
AI's answer varies moderately. Core narrative exists but is malleable.
No clear narrative has formed. Move fast with clear, authoritative content.
What to do with this score
- High consistency, and you're included? Protect your position with ongoing content.
- High consistency, and you're missing? Harder to break in. Focus on differentiation.
- Low consistency? The field is open. The right content strategy could shape the narrative.
Understanding AI's Answers
Dice Roller classifies what AI says into three categories. Each tells you something different about your brand strategy.
What AI Always Says (Core Messages)
Messages that appear in 80%+ of responses. These are things AI has "decided" about the topic.
What AI Sometimes Says (Variable Messages)
Messages that appear in 30-80% of responses. AI includes them sometimes, but not consistently.
What's Missing (Opportunity Gaps)
Topics, brands, or attributes that appeared only once or never.
Statistical Metrics (Experiment Mode)
Cross-model experiments provide additional statistical insights:
Shannon Entropy
Measures diversity of brand recommendations. Higher entropy means more brands are recommended equally; lower means few dominant brands.
Few dominant brands
Balanced recommendations
Many brands equally
Gini Coefficient
Measures concentration inequality. 0 means perfect equality (all brands mentioned equally), 1 means perfect inequality (one brand dominates).
Even brand distribution
Some concentration
Few brands dominate
Jaccard Index
Measures cross-model overlap. 1 means perfect overlap (models recommend same brands), 0 means no overlap.
Models agree on brands
Some shared brands
Different recommendations
Best Practices
1. Run Sufficient Iterations
Use at least 5 iterations for statistically meaningful results. For critical decisions, consider 7-10 iterations.
2. Compare Memory vs Search Mode
Run analysis in both modes. If your brand appears in search but not memory, focus on content that gets into AI training data. If it appears in memory but not search, ensure your web presence is optimized.
3. Test Multiple Prompts
Don't rely on a single prompt. Test variations like "best X for Y", "top X solutions", "X recommendations for Z industry" to understand your visibility across different query patterns.
4. Track Competitors
Run the same prompts with competitor brand names. Compare your mention rate and sentiment against competitors to identify positioning opportunities.
5. Monitor Over Time
AI models update periodically. Run stability analysis monthly or quarterly to track changes in your brand's AI perception.
Common Pitfalls
Over-interpreting Small Samples
3 iterations isn't enough. Random variance can make results misleading. Always use 5+ iterations.
Ignoring Sentiment Context
A brand mention isn't always positive. Check the sentiment of mentions - being cited as "expensive alternative" isn't ideal.
Not Accounting for Model Differences
Different AI models have different training data. A brand might be well-known to Gemini but not GPT, or vice versa.
Confusing Stability with Visibility
High consistency doesn't mean high visibility. A brand could be consistently NOT mentioned. Check mention rates alongside consistency scores.
Use Cases
Dice Roller helps marketing teams answer a question that was previously impossible to answer: What does AI actually believe about our brand?
Validate Your Brand Messaging
The Problem: You've crafted your positioning. Your website says one thing. But does AI agree?
How Dice Roller Helps: Run your core brand questions. If AI consistently mentions your key differentiators, your messaging is landing. If AI says something different, you have a gap between what you say and what AI believes.
Pre-Launch Message Testing
The Problem: You're about to launch a campaign with new messaging. Traditional testing tells you if humans like it. But will AI pick it up?
How Dice Roller Helps: Before launch, run category prompts to establish a baseline. After launch, run again. Track whether your new messaging appears in AI's responses.
Competitive Intelligence
The Problem: You know your competitors' marketing. But you don't know how AI positions them relative to you.
How Dice Roller Helps: Run the same category question and track which brands AI mentions consistently, which appear sometimes, and which are missing entirely.
Find Content Gaps
The Problem: AI recommends brands that have the clearest, most consistent content about a topic. If your content is thin or scattered, AI ignores you.
How Dice Roller Helps: When you're mentioned inconsistently or missing from responses, that's a content signal. AI doesn't have enough to go on.
Multi-Model Consistency Check
The Problem: ChatGPT says one thing about your brand. Gemini says another. Grok says something different. Which story is AI telling?
How Dice Roller Helps: Run the same question across multiple AI models. See where they agree and where they diverge. Consensus means your brand narrative is solid.
Training Data vs. Live Search
The Problem: AI has two sources of information: what it learned during training and what it finds via web search. You need to know which one is working for you.
How Dice Roller Helps: Run the same question in "memory mode" (training data only) and "search mode" (live web search). Compare the results.