| name | ai-ui-anti-patterns |
| description | The 10 most common AI UI failures, ranked by severity, with root cause analysis and concrete fixes. Covers the chatbot trap, dashboard fatigue, perpetual shimmer, hidden AI activity, read-only output, missing kill switches, and more. Use when debugging AI interface problems, reviewing an AI app for UX issues, or preventing common mistakes in AI-native design. Triggers on: "AI UI mistakes", "common AI UI failures", "anti-patterns", "what not to do", "AI UI pitfalls", "fix this AI interface".
|
AI UI Anti-Patterns
The 10 most common failures in AI-native interface design, ranked by severity. Each includes the root cause, the user impact, and a concrete fix.
#1: User Can't Tell If AI Is Working
Root cause: No visible system status. The AI is processing but there's no indicator.
User impact: User thinks the app is frozen. They refresh, resubmit, or abandon.
Fix: Add ambient activity indicator — subtle glow on the element being worked on, or a small pulsing dot in the corner. See calm-technology-ui for patterns.
#2: AI Output Is Read-Only
Root cause: Generated content is rendered as static text/images, not editable elements.
User impact: User sees a mistake but can't fix it inline. They must regenerate entirely or copy-paste to another tool.
Fix: Make every AI-generated element editable by default. Click to edit. Show "regenerate this section" on any element.
#3: No Way to Stop AI
Root cause: No stop/cancel mechanism. Once started, the AI runs to completion.
User impact: User watches helplessly as AI generates wrong content, wasting time and tokens.
Fix: Prominent stop button. Kill switch within 2 taps. Scoped control — stop individual agents, not just everything.
#4: Dashboard Overwhelms User
Root cause: Too much chrome — agent cards, progress bars, metric panels, sidebars — squeezing the actual output into a corner.
User impact: Cognitive overload. User can't focus on the content because they're monitoring 12 UI elements.
Fix: Output takes 80% of screen. Move chrome to hover/gesture. One primary action. See ai-native-ui-design for the generative canvas pattern.
#5: AI Failures Are Hidden
Root cause: Errors are logged to console or hidden in settings. User sees nothing wrong but gets bad output.
User impact: User trusts bad output. Mistakes propagate downstream.
Fix: Show errors immediately in plain language with recovery path. "The AI couldn't process your image because it was too large (max 5MB). Try compressing it."
#6: Chat Is the Only Interface
Root cause: The "chatbot trap" — every interaction goes through a chat conversation.
User impact: Slow, tedious interaction. Users must type what they want instead of clicking or gesturing.
Fix: Use buttons for common actions. Direct manipulation for editing. Chat for open-ended requests only. See ai-native-ui-design for the control transfer loop pattern.
#7: No Confidence Indication
Root cause: AI outputs are presented as fact with no indication of reliability.
User impact: User can't distinguish high-confidence output from speculative guesswork.
Fix: Per-element confidence scores. Inline badges: "Your headline here" [94%]. Hover → "3 iterations, strong test results."
#8: Perpetual Loading Spinners
Root cause: Using generic spinners for all loading states instead of streaming content.
User impact: User stares at a spinner for 30 seconds with no sense of progress.
Fix: Stream content as it's generated. Use skeleton → content morph. Show partial results immediately.
#9: Inconsistent AI Behavior
Root cause: Same prompt produces wildly different results. No explanation of why.
User impact: User can't build mental model of what the AI can and can't do. Trust erodes.
Fix: Show which model and settings were used. Use deterministic parameters when consistency matters. If variation is expected, tell the user.
#10: App Breaks Without AI
Root cause: The interface is a shell that only works when AI is available.
User impact: App becomes completely unusable when AI service is down or user is offline.
Fix: Graceful degradation. The pre-AI state must be a complete, usable product. Every feature should have a manual fallback.
Quick Checklist
Before shipping an AI feature, verify:
Related Skills
- For calm technology principles → use
calm-technology-ui
- For Nielsen's heuristics → use
ai-ui-heuristics
- For generative UI patterns → use
ai-native-ui-design
- For multi-agent UI → use
agentic-ui-patterns