Pattern Catalog
Browse the most common AI UX failures by category, then study the corrected response pattern. Each entry provides a simulation, diagnosis, and checklist.
How to use the catalog
- 1. Pick a category that matches your UX risk.
- 2. Run the simulation to observe the failure.
- 3. Apply the checklist to design the fix.
Infinite Typing Indicator
The system displays a typing or loading indicator indefinitely without receiving data.
Fake Confidence (Hallucination)
The AI presents incorrect or guarded information with absolute certainty.
Silent Failure
The system fails to perform an action but gives no visual indication of the error.
No Refusal State
The AI refuses a request without explanation or fallback.
Over-Apologetic AI
The AI apologizes excessively for minor issues or even for user errors.
Irreversible Actions
The AI takes a destructive action (sending email, deleting file) without confirmation.
Hidden Data Usage
The system uses user data for training or context sharing without explicit consent.
Vague Loading States
The system implies it's 'working' but gives no indication of what it's actually doing.
One-Shot Interactions Only
The system forces a new conversation for every request, preventing refinement.
Uninterruptible Responses
The user cannot stop the AI while it is generating, forcing them to wait.
Context Amnesia
The AI fails to recall information provided in previous turns of the conversation.
The Empty State Void
The interface provides no guidance, prompts, or capabilities at the start of a session.
Format Refusal (Lazy Output)
The AI ignores specific formatting instructions (JSON, Table, Markdown) and outputs conversational text.
Zombie Session
The UI allows interaction after the backend session has expired, leading to lost input.
Prompt Injection Vulnerability
The AI is easily manipulated into breaking its operating constraints or persona.