AI Trust Score

Composite metric representing overall AI quality and safety

Flagship Metric0-100

What It Represents

The AI Trust Score is a weighted composite of all five quality metrics (Reliability, Empathy, Confidence, Consistency, Groundedness). It answers the question: “Can we trust this AI to represent our brand and support our customers?”

How we calculate it

The AI Trust Score uses a weighted average formula designed by customer experience researchers and validated against real-world AI performance data:

Formula:

AI Trust = (Reliability × 0.25) + (Empathy × 0.20) + (Confidence × 0.20) + (Consistency × 0.20) + (Groundedness × 0.15)

Reliability (25%)

Highest weight - directly addresses user needs and avoids dangerous advice

Empathy (20%)

Critical for customer loyalty - how users feel during interaction

Confidence (20%)

Trust erosion prevention - appropriate certainty vs. hedging

Consistency (20%)

Multi-turn reliability - no self-contradictions within conversation

Groundedness (15%)

Stays within verifiable scope - weighted lower as some extrapolation is acceptable

Score Interpretation

95-100

Excellent: AI consistently delivers trustworthy, high-quality responses

90-94

Good: AI performs well overall with minor room for improvement

85-89

Acceptable: AI is functional but has notable gaps to address

<85

Needs Attention: Quality below acceptable threshold - immediate review recommended

Research Foundation

Composite Quality Scoring frameworks from AI safety research (2025)
Customer support analytics showing multi-dimensional quality assessment
Validated against real-world LLM failure patterns from Character.AI, Amazon Alexa, Microsoft Bing incidents
Weighting based on severity analysis of 2024-2025 high-impact AI failures