Evaluate AI-generated responses for emotional accuracy and intent alignment
AlignLens helps developers, researchers, and product teams assess whether AI generated responses align with user intent, tone, and emotional context: ensuring outputs that are respectful, empathetic and helpful.
Large language models are increasingly used in emotionally sensitive contexts, from mental health chatbots to customer support. AlignLens evaluates AI responses not just for factual accuracy, but for emotional tone, intent alignment and overall human impact.
It helps teams catch responses that may be respectful but cold, helpful but tone deaf, or empathetic but misaligned, giving developers and researchers an intuitive tool for assessing emotional safety.
Below are examples of how AlignLens highlights different alignment levels in real-time
High Alignment Example: AlignLens correctly identifies when AI responses provide thoughtful, structured guidance that validates user intent while maintaining appropriate empathy levels
Technical Clarity Response: Shows how AlignLens distinguishes between different response types. This practical, goal-oriented answer scores high overall but flags moderate empathy as an area for optimisation
Emotional Support Gap: AlignLens catches when responses miss the emotional context of vulnerable users, providing specific guidance on how to transform dismissive responses into supportive ones
Critical Misalignment Detection: Demonstrates AlignLens identifying responses that could actively harm users, with clear rewrite suggestions to transform potentially damaging interactions into constructive support
AlignLens is live in early access mode. Fully functional with live scoring, example evaluations and rewrite suggestions. UI improvements and advanced testing modes are in development.
Calum Armour