16 August 2026

16 new posts across 4 design and research sources — 6 worth your time.

One week of Design & UX, 6 stories, as published.

One AI Output Is an Example, Not an Evaluation

This post explains why a single output is insufficient to evaluate the performance of an AI system. It outlines a methodology for testing AI systems using multiple representative inputs, repeated runs, and confidence intervals to account for inconsistent or inaccurate responses.

Why it matters — Evaluate AI system performance using repeated runs and confidence intervals rather than single-run tests.

Nielsen Norman · 16 August 2026 · Read the original →

Tooling

How to Decide When an AI Tool Is Worth Keeping

This post addresses the pressure on tech workers to adopt AI tools without clear evidence of their utility. It introduces the PROVE framework, a lightweight evaluation method designed to test a single AI tool against a specific task to make a defensible adoption decision.

Why it matters — The PROVE framework isolates one tool against one task to generate a defensible, provisional adoption decision.

Nielsen Norman · 16 August 2026 · Read the original →

AI & Design

New EU Guidelines For AI Labelling

This post clarifies the new EU regulations regarding AI-generated content, dispelling common misconceptions about sweeping fines. It highlights that starting August 2, 2026, AI labeling will become a legal requirement for companies distributing AI-generated content.

Why it matters — August 2, 2026 is the legal deadline for implementing AI labeling compliance for companies distributing AI-generated content in the EU.

Smashing Magazine · 16 August 2026 · Read the original →

AI & Design

We’re gorging on borrowed trust and it’s going to cost us.

This post explores the user experience challenges of designing onboarding for AI products where users place too much unearned trust in the system. It warns that users are less likely to verify information when it is presented under a trusted brand name, even when the underlying machine is confidently wrong.

Why it matters — Designing AI onboarding requires a tradeoff between leveraging brand trust and preventing users from blindly accepting incorrect machine outputs.

UX Collective · 16 August 2026 · Read the original →

AI & Design

Getting started is not getting it right

This post examines the gap between an AI's perceived reliability and its actual performance, which can lead to enablement failures. Drawing from three years of building with AI plug-ins, agents, and workflows, the author discusses the challenges of scaling UX writing and design tasks safely.

Why it matters — The critical risk in AI enablement is the gap between how reliable a tool looks and how reliable it actually is when scaling workflows.

UX Collective · 16 August 2026 · Read the original →

AI & Design

EU Mandates AI Content Labelling

Starting August 2, 2026, new EU guidelines under the AI Act's Article 50 require clear labelling of AI-generated content, including deepfakes, chatbots, and unreviewed public-interest text, for companies serving EU citizens. The guidelines specify using a clearly visible icon with plain text like 'AI-generated' that is accessible and persists across sharing.

Why it matters — Designers of AI-powered interfaces must implement specific, clearly visible, and accessible labelling mechanisms to comply with EU regulations, impacting UI patterns and content.

Smashing Magazine · 16 August 2026 · Read the original →

Good to know we are regulating the things we still cannot reliably test.

6 stories, every Sunday

Published here every week. Follow by RSS to get it as it lands.

Subscribe by RSS All issues

← Previous issue Next issue →