The 6-metric framework
VectorGap measures AI perception across six distinct dimensions. Each captures a different aspect of how AI represents your brand.
1. Accuracy (0-100)
What it measures: Whether AI states factually correct information about your brand.
How it's calculated: We compare specific claims in AI responses against your knowledge base. Each claim is marked as accurate, inaccurate, or unverifiable.
What affects it: Outdated information, hallucinations, entity confusion.
How to improve: Build a comprehensive knowledge base. Create structured content with clear, verifiable facts.
2. Sentiment (0-100)
What it measures: The emotional tone AI uses when discussing your brand.
Score interpretation: 0-40 (predominantly negative), 41-60 (neutral/mixed), 61-80 (positive), 81-100 (very positive).
What affects it: Customer reviews, press coverage, competitor comparison pages.
3. Visibility (0-100)
What it measures: How prominently you appear in AI responses to relevant queries.
How it's calculated: First mention (100 points), second mention (80 points), third mention (60 points), lower or not mentioned (decreasing points).
What affects it: Brand authority, content volume, structured data, citation frequency.
4. Coverage (0-100)
What it measures: Whether AI mentions your key features, differentiators, and use cases.
Example: If your knowledge base lists 10 key features and AI mentions 6, your coverage is 60%.
5. Credibility (0-100)
What it measures: How authoritative AI sounds when discussing your brand.
Factors: Does AI cite sources? Use confident vs. hedging language? Present you as established or unknown?
6. Recommendation (0-100)
What it measures: Whether AI actively recommends your product to users.
The difference: Mention says "Company X exists" while Recommendation says "Company X is a good choice for [use case]".
This metric correlates most directly with AI-driven conversions.
How to interpret scores without lying to yourself
The score is an initial signal, not the business outcome. A score can improve because a model says nicer things, while the commercially important answer still omits your differentiator.
Read the evidence snippets before celebrating. The important question is: would a buyer trust this answer enough to keep you in the consideration set?
Priority order
Fix accuracy first, then coverage, then recommendation. Positive sentiment built on incorrect facts is dangerous. Visibility without the right positioning creates low-quality demand.
For agencies, turn the score into a client narrative: what AI currently believes, why that belief exists, and which public fixes are most likely to change it.