AI UX DAILY
Saturday, September 12, 2026
5 stories · curated for designers
The stories
Today in AI Products
 Loop11 |
via TLDR Design |
A concrete method for testing the interface states AI prototypes skip
Loop11 published a process for auditing AI-generated UIs: build a state inventory covering empty, loading, permission, and overlap states, then measure comprehension and recovery rather than preference.
| “ |
Run a state inventory against your next AI-generated prototype before usability testing, checking each edge state has a defined trigger, message, and recovery path. — Designer's Takeaway |
| Nielsen Norman Group |
Sep 11 |
NNG: build fully interactive prototypes with AI to test complex flows earlier
NNG argues AI tools now make it feasible to build interactive prototypes of complex interfaces early enough to run real usability tests before committing to a design direction.
| “ |
Use an AI prototyping tool to generate a clickable version of your most complex flow this sprint, then schedule a usability session before the design is finalized. — Designer's Takeaway |
 OpenAI ChatGPT |
Sep 10 |
ChatGPT Work ships a Data agent that builds dashboards from natural language
OpenAI launched a Data agent inside ChatGPT Work. Users connect company data and generate interactive dashboards by describing what they want in plain language.
| “ |
Study the Data agent's flow to see how OpenAI handles ambiguous natural-language queries, then apply the same progressive clarification pattern to your own AI-driven dashboard or reporting features. — Designer's Takeaway |
| UX + AI | Sep 11 |
Researcher Saeideh Bakhshi on where AI actually helps in UX research and where it fails
Bakhshi describes specific failure modes when AI is applied to user research, and identifies the tasks where it genuinely helps versus where human judgment stays irreplaceable.
| “ |
Map your current research workflow against Bakhshi's failure modes to find the one step where human judgment should stay in the loop before you automate it. — Designer's Takeaway |
| Ref |
via TLDR Design |
Ref's team: AI design work turns to slop without upfront constraints and ruthless editing
The Ref team describes their AI design process: set tight constraints first, iterate in Figma rather than code, stress-test taste in group critiques, and strip every extra element agents add by default.
| “ |
Before your next AI-assisted design session, write explicit constraints for layout, component count, and copy length so the output needs editing rather than rebuilding. — Designer's Takeaway |
Today's Idea
AI generates drafts. Designers define the boundaries.
Every item today points at the same gap: AI tools ship outputs that look complete but skip edge states, ignore constraints, and miss what research actually requires. Define your constraints, state inventories, and research guardrails before the AI runs, not after.
Stop shipping AI slop
Turn your design into Claude skills
Drop a screenshot. See which of the 38 patterns you are missing and take them away as Claude Code skills. Free, no signup for the first audit.