User feedback is a rough signal, not a finished instruction — users are excellent reporters of experience but poor designers of solutions. Treat feedback as a starting point, not an end: separate observation from interpretation, look for emotion underneath complaints, triangulate stated preference vs. actual behavior vs. underlying need, and remember the silent majority who never speak up often hold the real truth
UX Benchmarks for AI-Based Chat Software (2026)
A 2026 UX benchmark study of ChatGPT, Claude, Gemini, and Grok (420 participants) found: ChatGPT led in perceived usability (SUS 81.5) but NPS dropped significantly from 2025 (now 7%), while Claude showed the biggest gain in usefulness and now has the highest NPS (28%). Common complaints across all products: inaccurate responses, slow performance, and limited capabilities/usage limits; Claude users reported slightly higher tech savviness than ChatGPT users
NNG: Design-System Maturity - A 6-Dimension Framework
An NN/g framework for design-system maturity across 6 dimensions: Organizational Alignment, Team Effectiveness, Infrastructure Robustness, Governance, Support, and Adoption — each scored 1–5 (Absent to Exceptional). Instead of linear progression, use a radar chart to identify shape patterns (symmetry, valleys/spikes, tension between dimensions) and run regular assessments with diverse evaluators (system team, product users, sponsors) to diagnose bottlenecks and plan interventions
Prototyping: Error messages in UX - how to make them effective and user-friendly
Error messages should clearly describe the problem, offer specific solutions, use visual cues (colors/icons), stay consistent, and avoid jargon — they must communicate the error, help users fix it, and educate them to prevent future mistakes. Avoid vague messages, accusatory tone, lack of solutions, and weak visuals
AI: Stop Calling It Empathy - AI Does Not Feel Anything
A direct challenge to calling AI output "empathy" — AI doesn't feel or understand; it pattern-matches and tells you what you want to hear (sycophancy), while genuine empathy is being changed by another person's experience through presence and human interpretation. Mislabeling this leads to real harm: research budgets cut, researchers replaced, and products built on foundations that have never touched a real human
Experience: There Is No “Traditional” Way to Do UX Research - What Automotive UX Taught Me
A UX researcher on automotive projects learned there's no "traditional" research — when direct user access isn't possible, insights come from reviews, help sections, support tickets, and stakeholder feedback, looking for patterns. The real skill isn't knowing the domain, but knowing how to learn and decide with available information
@uxdigest
