UX-Context Design: Using UX Knowledge to Inform AI-Generated Design

Summary

NN/g argues that as AI tools increasingly generate user interfaces, human-centric UX documentation (such as personas and journey maps) should be replaced by machine-readable, carefully curated context files. To this end, the article proposes a 'UX.md' format that directly packages research findings, interaction rules, and user models into the context consumed by AI coding and design tools.

Why it matters

This approach reframes the core output of the UX profession. Since AI has become the primary 'reader' of research, the format and level of detail in outputs should shift from persuading humans to being actionable by a model. Furthermore, this method democratizes design quality by enabling product managers and engineers using AI tools to access organizational UX knowledge, thereby reducing the risk of context-free and generic outputs across the organization.

Highlights

• Google Labs open-sourced DESIGN.md in April 2026. This file is presented as a concrete, working example of machine-readable design context that AI tools read at the point of generation.
• The proposed UX.md goes beyond visual identity to include research synthesis as hard constraints, interaction standards, domain dictionaries, and user models.
• Curation by a human expert remains critical. Feeding raw research transcripts to a model can lead to the loss of key insights or an excessive focus on noise.

Critical take

UX.md is clearly a hypothesis, not a working standard. The article also downplays the maintenance issue. As products, users, and research evolve, a static context file becomes outdated; however, the article does not propose a governance or update frequency. The text also carries a subtle commercial undertone (NN/g sells a course called 'AI for Design Workflows') and exaggerates the 'average output' problem, which is already less felt in teams that write detailed prompts.